5 files changed, 20 insertions, 132 deletions

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index e02bafdd3682..93f4abd990a9 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -668,7 +668,7 @@ int mlx4_en_start_port(struct net_device *dev)
         queue_work(mdev->workqueue, &priv->mcast_task);
 
         priv->port_up = true;
-        netif_start_queue(dev);
+        netif_tx_start_all_queues(dev);
         return 0;
 
 mac_err:
@@ -700,14 +700,14 @@ void mlx4_en_stop_port(struct net_device *dev)
                 en_dbg(DRV, priv, "stop port called while port already down\n");
                 return;
         }
-        netif_stop_queue(dev);
 
         /* Synchronize with tx routine */
         netif_tx_lock_bh(dev);
-        priv->port_up = false;
+        netif_tx_stop_all_queues(dev);
         netif_tx_unlock_bh(dev);
 
         /* close port*/
+        priv->port_up = false;
         mlx4_CLOSE_PORT(mdev->dev, priv->port);
 
         /* Unregister Mac address for the port */
@@ -881,7 +881,6 @@ void mlx4_en_destroy_netdev(struct net_device *dev)
                 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
 
         cancel_delayed_work(&priv->stats_task);
-        cancel_delayed_work(&priv->refill_task);
         /* flush any pending task for this netdev */
         flush_workqueue(mdev->workqueue);
 
@@ -986,7 +985,6 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
         spin_lock_init(&priv->stats_lock);
         INIT_WORK(&priv->mcast_task, mlx4_en_do_set_multicast);
         INIT_WORK(&priv->mac_task, mlx4_en_do_set_mac);
-        INIT_DELAYED_WORK(&priv->refill_task, mlx4_en_rx_refill);
         INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
         INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
         INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 5a14899c1e25..91bdfdfd431f 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -269,31 +269,6 @@ reduce_rings:
         return 0;
 }
 
-static int mlx4_en_fill_rx_buf(struct net_device *dev,
-                               struct mlx4_en_rx_ring *ring)
-{
-        struct mlx4_en_priv *priv = netdev_priv(dev);
-        int num = 0;
-        int err;
-
-        while ((u32) (ring->prod - ring->cons) < ring->actual_size) {
-                err = mlx4_en_prepare_rx_desc(priv, ring, ring->prod &
-                                              ring->size_mask);
-                if (err) {
-                        if (netif_msg_rx_err(priv))
-                                en_warn(priv, "Failed preparing rx descriptor\n");
-                        priv->port_stats.rx_alloc_failed++;
-                        break;
-                }
-                ++num;
-                ++ring->prod;
-        }
-        if ((u32) (ring->prod - ring->cons) == ring->actual_size)
-                ring->full = 1;
-
-        return num;
-}
-
 static void mlx4_en_free_rx_buf(struct mlx4_en_priv *priv,
                                 struct mlx4_en_rx_ring *ring)
 {
@@ -312,42 +287,6 @@ static void mlx4_en_free_rx_buf(struct mlx4_en_priv *priv,
         }
 }
 
-
-void mlx4_en_rx_refill(struct work_struct *work)
-{
-        struct delayed_work *delay = to_delayed_work(work);
-        struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
-                                                 refill_task);
-        struct mlx4_en_dev *mdev = priv->mdev;
-        struct net_device *dev = priv->dev;
-        struct mlx4_en_rx_ring *ring;
-        int need_refill = 0;
-        int i;
-
-        mutex_lock(&mdev->state_lock);
-        if (!mdev->device_up || !priv->port_up)
-                goto out;
-
-        /* We only get here if there are no receive buffers, so we can't race
-         * with Rx interrupts while filling buffers */
-        for (i = 0; i < priv->rx_ring_num; i++) {
-                ring = &priv->rx_ring[i];
-                if (ring->need_refill) {
-                        if (mlx4_en_fill_rx_buf(dev, ring)) {
-                                ring->need_refill = 0;
-                                mlx4_en_update_rx_prod_db(ring);
-                        } else
-                                need_refill = 1;
-                }
-        }
-        if (need_refill)
-                queue_delayed_work(mdev->workqueue, &priv->refill_task, HZ);
-
-out:
-        mutex_unlock(&mdev->state_lock);
-}
-
-
 int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,
                            struct mlx4_en_rx_ring *ring, u32 size, u16 stride)
 {
@@ -457,9 +396,6 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
                         ring_ind--;
                         goto err_allocator;
                 }
-
-                /* Fill Rx buffers */
-                ring->full = 0;
         }
         err = mlx4_en_fill_rx_buffers(priv);
         if (err)
@@ -647,33 +583,6 @@ static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,
         return skb;
 }
 
-static void mlx4_en_copy_desc(struct mlx4_en_priv *priv,
-                              struct mlx4_en_rx_ring *ring,
-                              int from, int to, int num)
-{
-        struct skb_frag_struct *skb_frags_from;
-        struct skb_frag_struct *skb_frags_to;
-        struct mlx4_en_rx_desc *rx_desc_from;
-        struct mlx4_en_rx_desc *rx_desc_to;
-        int from_index, to_index;
-        int nr, i;
-
-        for (i = 0; i < num; i++) {
-                from_index = (from + i) & ring->size_mask;
-                to_index = (to + i) & ring->size_mask;
-                skb_frags_from = ring->rx_info + (from_index << priv->log_rx_info);
-                skb_frags_to = ring->rx_info + (to_index << priv->log_rx_info);
-                rx_desc_from = ring->buf + (from_index << ring->log_stride);
-                rx_desc_to = ring->buf + (to_index << ring->log_stride);
-
-                for (nr = 0; nr < priv->num_frags; nr++) {
-                        skb_frags_to[nr].page = skb_frags_from[nr].page;
-                        skb_frags_to[nr].page_offset = skb_frags_from[nr].page_offset;
-                        rx_desc_to->data[nr].addr = rx_desc_from->data[nr].addr;
-                }
-        }
-}
-
 
 int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int budget)
 {
@@ -821,11 +730,6 @@ out:
         wmb(); /* ensure HW sees CQ consumer before we post new buffers */
         ring->cons = cq->mcq.cons_index;
         ring->prod += polled; /* Polled descriptors were realocated in place */
-        if (unlikely(!ring->full)) {
-                mlx4_en_copy_desc(priv, ring, ring->cons - polled,
-                                  ring->prod - polled, polled);
-                mlx4_en_fill_rx_buf(dev, ring);
-        }
         mlx4_en_update_rx_prod_db(ring);
         return polled;
 }
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index 5dc7466ad035..08c43f2ae72b 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -515,16 +515,9 @@ static int get_real_size(struct sk_buff *skb, struct net_device *dev,
                         else {
                                 if (netif_msg_tx_err(priv))
                                         en_warn(priv, "Non-linear headers\n");
-                                dev_kfree_skb_any(skb);
                                 return 0;
                         }
                 }
-                if (unlikely(*lso_header_size > MAX_LSO_HDR_SIZE)) {
-                        if (netif_msg_tx_err(priv))
-                                en_warn(priv, "LSO header size too big\n");
-                        dev_kfree_skb_any(skb);
-                        return 0;
-                }
         } else {
                 *lso_header_size = 0;
                 if (!is_inline(skb, NULL))
@@ -616,13 +609,9 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
         int lso_header_size;
         void *fragptr;
 
-        if (unlikely(!skb->len)) {
-                dev_kfree_skb_any(skb);
-                return NETDEV_TX_OK;
-        }
         real_size = get_real_size(skb, dev, &lso_header_size);
         if (unlikely(!real_size))
-                return NETDEV_TX_OK;
+                goto tx_drop;
 
         /* Allign descriptor to TXBB size */
         desc_size = ALIGN(real_size, TXBB_SIZE);
@@ -630,8 +619,7 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
         if (unlikely(nr_txbb > MAX_DESC_TXBBS)) {
                 if (netif_msg_tx_err(priv))
                         en_warn(priv, "Oversized header or SG list\n");
-                dev_kfree_skb_any(skb);
-                return NETDEV_TX_OK;
+                goto tx_drop;
         }
 
         tx_ind = skb->queue_mapping;
@@ -653,14 +641,6 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
                 return NETDEV_TX_BUSY;
         }
 
-        /* Now that we know what Tx ring to use */
-        if (unlikely(!priv->port_up)) {
-                if (netif_msg_tx_err(priv))
-                        en_warn(priv, "xmit: port down!\n");
-                dev_kfree_skb_any(skb);
-                return NETDEV_TX_OK;
-        }
-
         /* Track current inflight packets for performance analysis */
         AVG_PERF_COUNTER(priv->pstats.inflight_avg,
                          (u32) (ring->prod - ring->cons - 1));
@@ -785,5 +765,10 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
         mlx4_en_xmit_poll(priv, tx_ind);
 
         return 0;
+
+tx_drop:
+        dev_kfree_skb_any(skb);
+        priv->stats.tx_dropped++;
+        return NETDEV_TX_OK;
 }
 
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index d43a9e4c2aea..c7c5e86804ff 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -99,7 +99,6 @@
 #define RSS_FACTOR              2
 #define TXBB_SIZE               64
 #define HEADROOM                (2048 / TXBB_SIZE + 1)
-#define MAX_LSO_HDR_SIZE        92
 #define STAMP_STRIDE            64
 #define STAMP_DWORDS            (STAMP_STRIDE / 4)
 #define STAMP_SHIFT             31
@@ -296,8 +295,6 @@ struct mlx4_en_rx_ring {
         u32 prod;
         u32 cons;
         u32 buf_size;
-        int need_refill;
-        int full;
         void *buf;
         void *rx_info;
         unsigned long bytes;
@@ -495,7 +492,6 @@ struct mlx4_en_priv {
         struct mlx4_en_cq rx_cq[MAX_RX_RINGS];
         struct work_struct mcast_task;
         struct work_struct mac_task;
-        struct delayed_work refill_task;
         struct work_struct watchdog_task;
         struct work_struct linkstate_task;
         struct delayed_work stats_task;
@@ -565,7 +561,6 @@ void mlx4_en_set_default_rss_map(struct mlx4_en_priv *priv,
 int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv);
 void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv);
 int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring);
-void mlx4_en_rx_refill(struct work_struct *work);
 void mlx4_en_rx_irq(struct mlx4_cq *mcq);
 
 int mlx4_SET_MCAST_FLTR(struct mlx4_dev *dev, u8 port, u64 mac, u64 clear, u8 mode);
diff --git a/drivers/net/mlx4/mr.c b/drivers/net/mlx4/mr.c
index 5887e4764d22..f96948be0a44 100644
--- a/drivers/net/mlx4/mr.c
+++ b/drivers/net/mlx4/mr.c
@@ -399,11 +399,14 @@ static int mlx4_write_mtt_chunk(struct mlx4_dev *dev, struct mlx4_mtt *mtt,
         if (!mtts)
                 return -ENOMEM;
 
+        dma_sync_single_for_cpu(&dev->pdev->dev, dma_handle,
+                                npages * sizeof (u64), DMA_TO_DEVICE);
+
         for (i = 0; i < npages; ++i)
                 mtts[i] = cpu_to_be64(page_list[i] | MLX4_MTT_FLAG_PRESENT);
 
-        dma_sync_single_for_cpu(&dev->pdev->dev, dma_handle,
-                                npages * sizeof (u64), DMA_TO_DEVICE);
+        dma_sync_single_for_device(&dev->pdev->dev, dma_handle,
+                                   npages * sizeof (u64), DMA_TO_DEVICE);
 
         return 0;
 }
@@ -547,11 +550,14 @@ int mlx4_map_phys_fmr(struct mlx4_dev *dev, struct mlx4_fmr *fmr, u64 *page_list
         /* Make sure MPT status is visible before writing MTT entries */
         wmb();
 
+        dma_sync_single_for_cpu(&dev->pdev->dev, fmr->dma_handle,
+                                npages * sizeof(u64), DMA_TO_DEVICE);
+
         for (i = 0; i < npages; ++i)
                 fmr->mtts[i] = cpu_to_be64(page_list[i] | MLX4_MTT_FLAG_PRESENT);
 
-        dma_sync_single_for_cpu(&dev->pdev->dev, fmr->dma_handle,
-                                npages * sizeof(u64), DMA_TO_DEVICE);
+        dma_sync_single_for_device(&dev->pdev->dev, fmr->dma_handle,
+                                   npages * sizeof(u64), DMA_TO_DEVICE);
 
         fmr->mpt->key    = cpu_to_be32(key);
         fmr->mpt->lkey   = cpu_to_be32(key);