diff options
Diffstat (limited to 'drivers/net/mlx4/en_tx.c')
| -rw-r--r-- | drivers/net/mlx4/en_tx.c | 74 |
1 files changed, 53 insertions, 21 deletions
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c index a680cd4..b229acf 100644 --- a/drivers/net/mlx4/en_tx.c +++ b/drivers/net/mlx4/en_tx.c @@ -44,6 +44,7 @@ enum { MAX_INLINE = 104, /* 128 - 16 - 4 - 4 */ + MAX_BF = 256, }; static int inline_thold __read_mostly = MAX_INLINE; @@ -52,7 +53,7 @@ module_param_named(inline_thold, inline_thold, int, 0444); MODULE_PARM_DESC(inline_thold, "threshold for using inline data"); int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv, - struct mlx4_en_tx_ring *ring, u32 size, + struct mlx4_en_tx_ring *ring, int qpn, u32 size, u16 stride) { struct mlx4_en_dev *mdev = priv->mdev; @@ -103,23 +104,25 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv, "buf_size:%d dma:%llx\n", ring, ring->buf, ring->size, ring->buf_size, (unsigned long long) ring->wqres.buf.direct.map); - err = mlx4_qp_reserve_range(mdev->dev, 1, 1, &ring->qpn); - if (err) { - en_err(priv, "Failed reserving qp for tx ring.\n"); - goto err_map; - } - + ring->qpn = qpn; err = mlx4_qp_alloc(mdev->dev, ring->qpn, &ring->qp); if (err) { en_err(priv, "Failed allocating qp %d\n", ring->qpn); - goto err_reserve; + goto err_map; } ring->qp.event = mlx4_en_sqp_event; + err = mlx4_bf_alloc(mdev->dev, &ring->bf); + if (err) { + en_dbg(DRV, priv, "working without blueflame (%d)", err); + ring->bf.uar = &mdev->priv_uar; + ring->bf.uar->map = mdev->uar_map; + ring->bf_enabled = false; + } else + ring->bf_enabled = true; + return 0; -err_reserve: - mlx4_qp_release_range(mdev->dev, ring->qpn, 1); err_map: mlx4_en_unmap_buffer(&ring->wqres.buf); err_hwq_res: @@ -139,6 +142,8 @@ void mlx4_en_destroy_tx_ring(struct mlx4_en_priv *priv, struct mlx4_en_dev *mdev = priv->mdev; en_dbg(DRV, priv, "Destroying tx ring, qpn: %d\n", ring->qpn); + if (ring->bf_enabled) + mlx4_bf_free(mdev->dev, &ring->bf); mlx4_qp_remove(mdev->dev, &ring->qp); mlx4_qp_free(mdev->dev, &ring->qp); mlx4_qp_release_range(mdev->dev, ring->qpn, 1); @@ -171,6 +176,8 @@ int mlx4_en_activate_tx_ring(struct mlx4_en_priv *priv, mlx4_en_fill_qp_context(priv, ring->size, ring->stride, 1, 0, ring->qpn, ring->cqn, &ring->context); + if (ring->bf_enabled) + ring->context.usr_page = cpu_to_be32(ring->bf.uar->index); err = mlx4_qp_to_ready(mdev->dev, &ring->wqres.mtt, &ring->context, &ring->qp, &ring->qp_state); @@ -591,6 +598,11 @@ u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb) return skb_tx_hash(dev, skb); } +static void mlx4_bf_copy(unsigned long *dst, unsigned long *src, unsigned bytecnt) +{ + __iowrite64_copy(dst, src, bytecnt / 8); +} + netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev) { struct mlx4_en_priv *priv = netdev_priv(dev); @@ -609,12 +621,13 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev) int desc_size; int real_size; dma_addr_t dma; - u32 index; + u32 index, bf_index; __be32 op_own; u16 vlan_tag = 0; int i; int lso_header_size; void *fragptr; + bool bounce = false; if (!priv->port_up) goto tx_drop; @@ -623,7 +636,7 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev) if (unlikely(!real_size)) goto tx_drop; - /* Allign descriptor to TXBB size */ + /* Align descriptor to TXBB size */ desc_size = ALIGN(real_size, TXBB_SIZE); nr_txbb = desc_size / TXBB_SIZE; if (unlikely(nr_txbb > MAX_DESC_TXBBS)) { @@ -657,13 +670,16 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev) /* Packet is good - grab an index and transmit it */ index = ring->prod & ring->size_mask; + bf_index = ring->prod; /* See if we have enough space for whole descriptor TXBB for setting * SW ownership on next descriptor; if not, use a bounce buffer. */ if (likely(index + nr_txbb <= ring->size)) tx_desc = ring->buf + index * TXBB_SIZE; - else + else { tx_desc = (struct mlx4_en_tx_desc *) ring->bounce_buf; + bounce = true; + } /* Save skb in tx_info ring */ tx_info = &ring->tx_info[index]; @@ -768,21 +784,37 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev) ring->prod += nr_txbb; /* If we used a bounce buffer then copy descriptor back into place */ - if (tx_desc == (struct mlx4_en_tx_desc *) ring->bounce_buf) + if (bounce) tx_desc = mlx4_en_bounce_to_desc(priv, ring, index, desc_size); /* Run destructor before passing skb to HW */ if (likely(!skb_shared(skb))) skb_orphan(skb); - /* Ensure new descirptor hits memory - * before setting ownership of this descriptor to HW */ - wmb(); - tx_desc->ctrl.owner_opcode = op_own; + if (ring->bf_enabled && desc_size <= MAX_BF && !bounce && !vlan_tag) { + *(u32 *) (&tx_desc->ctrl.vlan_tag) |= ring->doorbell_qpn; + op_own |= htonl((bf_index & 0xffff) << 8); + /* Ensure new descirptor hits memory + * before setting ownership of this descriptor to HW */ + wmb(); + tx_desc->ctrl.owner_opcode = op_own; - /* Ring doorbell! */ - wmb(); - writel(ring->doorbell_qpn, mdev->uar_map + MLX4_SEND_DOORBELL); + wmb(); + + mlx4_bf_copy(ring->bf.reg + ring->bf.offset, (unsigned long *) &tx_desc->ctrl, + desc_size); + + wmb(); + + ring->bf.offset ^= ring->bf.buf_size; + } else { + /* Ensure new descirptor hits memory + * before setting ownership of this descriptor to HW */ + wmb(); + tx_desc->ctrl.owner_opcode = op_own; + wmb(); + writel(ring->doorbell_qpn, ring->bf.uar->map + MLX4_SEND_DOORBELL); + } /* Poll CQ here */ mlx4_en_xmit_poll(priv, tx_ind); |