[PATCH 1/5] mlx4_en: Fix error handling while activating RX rings

[PATCH 1/5] mlx4_en: Fix error handling while activating RX rings

[Yevgeny Petrilin]

In case of failure of either srq creation or page allocation,
the cleanup code handled the failed ring as well, and tried
to destroy resources that where not allocated.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_rx.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 7e40741..8673008 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -436,8 +436,9 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 		/* Initialize page allocators */
 		err = mlx4_en_init_allocator(priv, ring);
 		if (err) {
-			 mlx4_err(mdev, "Failed initializing ring allocator\n");
-			 goto err_allocator;
+			mlx4_err(mdev, "Failed initializing ring allocator\n");
+			ring_ind--;
+			goto err_allocator;
 		}

 		/* Fill Rx buffers */
@@ -467,6 +468,7 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 				     ring->wqres.db.dma, &ring->srq);
 		if (err){
 			mlx4_err(mdev, "Failed to allocate srq\n");
+			ring_ind--;
 			goto err_srq;
 		}
 		ring->srq.event = mlx4_en_srq_event;
-- 
1.5.4

Re: [PATCH 2/5]mlx4_en: Fix a race at restart task

[Yevgeny Petrilin]

The query whether the port is up or not should be done at
the execution of the restart task and not when it is queued.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_netdev.c |   19 +++++++++++--------
 1 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 303c23d..09fb7cf 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -348,11 +348,9 @@ static void mlx4_en_tx_timeout(struct net_device *dev)
 	if (netif_msg_timer(priv))
 		mlx4_warn(mdev, "Tx timeout called on port:%d\n", priv->port);

-	if (netif_carrier_ok(dev)) {
-		priv->port_stats.tx_timeout++;
-		mlx4_dbg(DRV, priv, "Scheduling watchdog\n");
-		queue_work(mdev->workqueue, &priv->watchdog_task);
-	}
+	priv->port_stats.tx_timeout++;
+	mlx4_dbg(DRV, priv, "Scheduling watchdog\n");
+	queue_work(mdev->workqueue, &priv->watchdog_task);
 }


@@ -761,9 +759,14 @@ static void mlx4_en_restart(struct work_struct *work)
 	struct net_device *dev = priv->dev;

 	mlx4_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
-	mlx4_en_stop_port(dev);
-	if (mlx4_en_start_port(dev))
-	    mlx4_err(mdev, "Failed restarting port %d\n", priv->port);
+
+	mutex_lock(&mdev->state_lock);
+	if (priv->port_up) {
+		mlx4_en_stop_port(dev);
+		if (mlx4_en_start_port(dev))
+			mlx4_err(mdev, "Failed restarting port %d\n", priv->port);
+	}
+	mutex_unlock(&mdev->state_lock);
 }


-- 
1.5.4

[PATCH] mlx4_en: Fix cleanup if workqueue create in mlx4_en_add() fails

[Roland Dreier]

If creating a workqueue fails, don't jump to the error path where that
same workqueue is destroyed, since destroy_workqueue() can't handle a
NULL pointer.

This was spotted by the Coverity checker (CID 2617).

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/net/mlx4/en_main.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/net/mlx4/en_main.c b/drivers/net/mlx4/en_main.c
index eda72dd..510633f 100644
--- a/drivers/net/mlx4/en_main.c
+++ b/drivers/net/mlx4/en_main.c
@@ -181,7 +181,7 @@ static void *mlx4_en_add(struct mlx4_dev *dev)
 	mdev->workqueue = create_singlethread_workqueue("mlx4_en");
 	if (!mdev->workqueue) {
 		err = -ENOMEM;
-		goto err_close_nic;
+		goto err_mr;
 	}
 
 	/* At this stage all non-port specific tasks are complete:
@@ -214,9 +214,8 @@ err_free_netdev:
 	flush_workqueue(mdev->workqueue);
 
 	/* Stop event queue before we drop down to release shared SW state */
-
-err_close_nic:
 	destroy_workqueue(mdev->workqueue);
+
 err_mr:
 	mlx4_mr_free(dev, &mdev->mr);
 err_uar:

Re: [PATCH] mlx4_en: Fix cleanup if workqueue create in mlx4_en_add() fails

[David Miller]

From: Roland Dreier <rdreier@cisco.com>
Date: Mon, 20 Apr 2009 21:32:08 -0700

> If creating a workqueue fails, don't jump to the error path where that
> same workqueue is destroyed, since destroy_workqueue() can't handle a
> NULL pointer.
> 
> This was spotted by the Coverity checker (CID 2617).
> 
> Signed-off-by: Roland Dreier <rolandd@cisco.com>

Applied.

Re: [PATCH 2/5]mlx4_en: Fix a race at restart task

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 20 Apr 2009 17:26:05 +0300

> The query whether the port is up or not should be done at
> the execution of the restart task and not when it is queued.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

[PATCH 3/5] mlx4_en: Assign dummy event handler for TX queue

[Yevgeny Petrilin]

The low level driver always assumes this handler exists.
The lack of it could cause kernel panic

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_resources.c |    6 ++++++
 drivers/net/mlx4/en_rx.c        |    6 ------
 drivers/net/mlx4/en_tx.c        |    1 +
 drivers/net/mlx4/mlx4_en.h      |    1 +
 4 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/net/mlx4/en_resources.c b/drivers/net/mlx4/en_resources.c
index a054520..65ca706 100644
--- a/drivers/net/mlx4/en_resources.c
+++ b/drivers/net/mlx4/en_resources.c
@@ -94,3 +94,9 @@ void mlx4_en_unmap_buffer(struct mlx4_buf *buf)

 	vunmap(buf->direct.buf);
 }
+
+void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event)
+{
+    return;
+}
+
diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 8673008..0cbb78c 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -928,12 +928,6 @@ void mlx4_en_set_default_rss_map(struct mlx4_en_priv *priv,
 	}
 }

-static void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event)
-{
-    return;
-}
-
-
 static int mlx4_en_config_rss_qp(struct mlx4_en_priv *priv,
 				 int qpn, int srqn, int cqn,
 				 enum mlx4_qp_state *state,
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index 4afd599..ac6fc49 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -112,6 +112,7 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
 		mlx4_err(mdev, "Failed allocating qp %d\n", ring->qpn);
 		goto err_reserve;
 	}
+	ring->qp.event = mlx4_en_sqp_event;

 	return 0;

diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index e9af32d..ef840ab 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -538,6 +538,7 @@ int mlx4_en_poll_rx_cq(struct napi_struct *napi, int budget);
 void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
 			     int is_tx, int rss, int qpn, int cqn, int srqn,
 			     struct mlx4_qp_context *context);
+void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event);
 int mlx4_en_map_buffer(struct mlx4_buf *buf);
 void mlx4_en_unmap_buffer(struct mlx4_buf *buf);

-- 
1.5.4

Re: [PATCH 3/5] mlx4_en: Assign dummy event handler for TX queue

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 20 Apr 2009 17:30:03 +0300

> The low level driver always assumes this handler exists.
> The lack of it could cause kernel panic
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

[PATCH 4/5] mlx4_en: use NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM for tx csum at initialization

[Yevgeny Petrilin]

The former usage was to set the NETIF_F_HW_CSUM flag which is not used
in get_tx_csum. It caused Ethtool to show tx checksum as "on" even
though it was turned off in previous operation.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_netdev.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 09fb7cf..438678a 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -1057,7 +1057,7 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	 * Set driver features
 	 */
 	dev->features |= NETIF_F_SG;
-	dev->features |= NETIF_F_HW_CSUM;
+	dev->features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
 	dev->features |= NETIF_F_HIGHDMA;
 	dev->features |= NETIF_F_HW_VLAN_TX |
 			 NETIF_F_HW_VLAN_RX |
-- 
1.5.4

[PATCH 5/5] mlx4_en: Move to SW counters for total bytes and packets

[Yevgeny Petrilin]

The per ring counters are implemented in SW. Now moving to have the total
counters as the sum of all rings. This way the numbers will always be consistent
and we no longer depend on HW buffer size limitations for those counters
that can be insufficient in some cases.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_port.c |   45 ++++++++++++-------------------------------
 1 files changed, 13 insertions(+), 32 deletions(-)

diff --git a/drivers/net/mlx4/en_port.c b/drivers/net/mlx4/en_port.c
index c5a4c03..a29abe8 100644
--- a/drivers/net/mlx4/en_port.c
+++ b/drivers/net/mlx4/en_port.c
@@ -151,6 +151,7 @@ int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset)
 	struct mlx4_cmd_mailbox *mailbox;
 	u64 in_mod = reset << 8 | port;
 	int err;
+	int i;

 	mailbox = mlx4_alloc_cmd_mailbox(mdev->dev);
 	if (IS_ERR(mailbox))
@@ -165,38 +166,18 @@ int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset)

 	spin_lock_bh(&priv->stats_lock);

-	stats->rx_packets = be32_to_cpu(mlx4_en_stats->RTOTFRMS) -
-			    be32_to_cpu(mlx4_en_stats->RDROP);
-	stats->tx_packets = be64_to_cpu(mlx4_en_stats->TTOT_prio_0) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_1) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_2) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_3) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_4) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_5) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_6) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_prio_7) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_novlan) +
-			    be64_to_cpu(mlx4_en_stats->TTOT_loopbk);
-	stats->rx_bytes = be64_to_cpu(mlx4_en_stats->ROCT_prio_0) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_1) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_2) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_3) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_4) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_5) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_6) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_prio_7) +
-			  be64_to_cpu(mlx4_en_stats->ROCT_novlan);
-
-	stats->tx_bytes = be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_0) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_1) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_2) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_3) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_4) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_5) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_6) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_prio_7) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_novlan) +
-			  be64_to_cpu(mlx4_en_stats->TTTLOCT_loopbk);
+	stats->rx_packets = 0;
+	stats->rx_bytes = 0;
+	for (i = 0; i < priv->rx_ring_num; i++) {
+		stats->rx_packets += priv->rx_ring[i].packets;
+		stats->rx_bytes += priv->rx_ring[i].bytes;
+	}
+	stats->tx_packets = 0;
+	stats->tx_bytes = 0;
+	for (i = 0; i <= priv->tx_ring_num; i++) {
+		stats->tx_packets += priv->tx_ring[i].packets;
+		stats->tx_bytes += priv->tx_ring[i].bytes;
+	}

 	stats->rx_errors = be64_to_cpu(mlx4_en_stats->PCS) +
 			   be32_to_cpu(mlx4_en_stats->RdropLength) +
-- 
1.5.4

Re: [PATCH 5/5] mlx4_en: Move to SW counters for total bytes and packets

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 20 Apr 2009 17:34:38 +0300

> The per ring counters are implemented in SW. Now moving to have the total
> counters as the sum of all rings. This way the numbers will always be consistent
> and we no longer depend on HW buffer size limitations for those counters
> that can be insufficient in some cases.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

Re: [PATCH 4/5] mlx4_en: use NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM for tx csum at initialization

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 20 Apr 2009 17:33:15 +0300

> The former usage was to set the NETIF_F_HW_CSUM flag which is not used
> in get_tx_csum. It caused Ethtool to show tx checksum as "on" even
> though it was turned off in previous operation.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

[PATCH 2/2] mlx4_en: Handle page allocation failure during receive

[Yevgeny Petrilin]

If we failed to allocate new fragments for receive buffer,
the packet should be dropped and packets should be reused.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_rx.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 0cbb78c..7942c4d 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -610,6 +610,10 @@ static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,
 		used_frags = mlx4_en_complete_rx_desc(priv, rx_desc, skb_frags,
 						      skb_shinfo(skb)->frags,
 						      page_alloc, length);
+		if (unlikely(!used_frags)) {
+			kfree_skb(skb);
+			return NULL;
+		}
 		skb_shinfo(skb)->nr_frags = used_frags;

 		/* Copy headers into the skb linear buffer */
-- 

Re: [PATCH 2/2] mlx4_en: Handle page allocation failure during receive

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 27 Apr 2009 09:42:57 +0300

> If we failed to allocate new fragments for receive buffer,
> the packet should be dropped and packets should be reused.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

Re: [PATCH 1/5] mlx4_en: Fix error handling while activating RX rings

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 20 Apr 2009 17:24:28 +0300

> In case of failure of either srq creation or page allocation,
> the cleanup code handled the failed ring as well, and tried
> to destroy resources that where not allocated.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

[PATCH 1/2] mlx4_en: Fix cleanup flow on cq activation

[Yevgeny Petrilin]

In case of mlx4_en_activate_cq() failure, the cleanup
code would go to rx_err and try to disable unactivated rings.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_netdev.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 438678a..7bcc49d 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -583,7 +583,7 @@ int mlx4_en_start_port(struct net_device *dev)
 		err = mlx4_en_activate_cq(priv, cq);
 		if (err) {
 			mlx4_err(mdev, "Failed activating Rx CQ\n");
-			goto rx_err;
+			goto cq_err;
 		}
 		for (j = 0; j < cq->size; j++)
 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
--

Re: [PATCH 1/2] mlx4_en: Fix cleanup flow on cq activation

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 27 Apr 2009 09:41:34 +0300

> In case of mlx4_en_activate_cq() failure, the cleanup
> code would go to rx_err and try to disable unactivated rings.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied.

[PATCH] mlx4_en: Fix not deleted napi structures

[Yevgeny Petrilin]

Napi structures are being created each time we open a port, but when
the port is closed the napi structure is only disabled but not removed.
This bug caused hang while removing the driver.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_cq.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/drivers/net/mlx4/en_cq.c b/drivers/net/mlx4/en_cq.c
index 91f50de..a276125 100644
--- a/drivers/net/mlx4/en_cq.c
+++ b/drivers/net/mlx4/en_cq.c
@@ -125,8 +125,10 @@ void mlx4_en_deactivate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq)

 	if (cq->is_tx)
 		del_timer(&cq->timer);
-	else
+	else {
 		napi_disable(&cq->napi);
+		netif_napi_del(&cq->napi);
+	}

 	mlx4_cq_free(mdev->dev, &cq->mcq);
 }
-- 
1.6.0

Re: [PATCH] mlx4_en: Fix not deleted napi structures

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Wed, 13 May 2009 14:47:22 +0300

> Napi structures are being created each time we open a port, but when
> the port is closed the napi structure is only disabled but not removed.
> This bug caused hang while removing the driver.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied, thanks!

[PATCH 1/2] mlx4_en: Removed redundant stride variable

[Yevgeny Petrilin]

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_netdev.c |    3 ---
 1 files changed, 0 insertions(+), 3 deletions(-)

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 7bcc49d..be487fa 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -559,7 +559,6 @@ int mlx4_en_start_port(struct net_device *dev)
 	struct mlx4_en_rx_ring *rx_ring;
 	int rx_index = 0;
 	int tx_index = 0;
-	u16 stride;
 	int err = 0;
 	int i;
 	int j;
@@ -573,8 +572,6 @@ int mlx4_en_start_port(struct net_device *dev)
 	dev->mtu = min(dev->mtu, priv->max_mtu);
 	mlx4_en_calc_rx_buf(dev);
 	mlx4_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
-	stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
-				    DS_SIZE * priv->num_frags);
 	/* Configure rx cq's and rings */
 	for (i = 0; i < priv->rx_ring_num; i++) {
 		cq = &priv->rx_cq[i];
-- 
1.6.0

Re: [PATCH 1/2] mlx4_en: Removed redundant stride variable

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Sun, 24 May 2009 16:16:51 +0300

> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied to net-next-2.6

[PATCH 2/2] mlx4_en: Fix partial rings feature

[Yevgeny Petrilin]

In case of allocation failure, the actual ring size is rounded down to
nearest power of 2. The remaining descriptors are freed.
The CQ and SRQ are allocated with the actual size and the mask is updated.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_cq.c     |    3 ++
 drivers/net/mlx4/en_netdev.c |   23 ++++++--------
 drivers/net/mlx4/en_rx.c     |   71 ++++++++++++++++++++++++++---------------
 3 files changed, 58 insertions(+), 39 deletions(-)

diff --git a/drivers/net/mlx4/en_cq.c b/drivers/net/mlx4/en_cq.c
index a276125..21786ad 100644
--- a/drivers/net/mlx4/en_cq.c
+++ b/drivers/net/mlx4/en_cq.c
@@ -89,6 +89,9 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq)
 	*cq->mcq.arm_db    = 0;
 	memset(cq->buf, 0, cq->buf_size);

+	if (!cq->is_tx)
+		cq->size = priv->rx_ring[cq->ring].actual_size;
+
 	err = mlx4_cq_alloc(mdev->dev, cq->size, &cq->wqres.mtt, &mdev->priv_uar,
 			    cq->wqres.db.dma, &cq->mcq, cq->vector, cq->is_tx);
 	if (err)
diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index be487fa..0cd185a 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -556,7 +556,6 @@ int mlx4_en_start_port(struct net_device *dev)
 	struct mlx4_en_dev *mdev = priv->mdev;
 	struct mlx4_en_cq *cq;
 	struct mlx4_en_tx_ring *tx_ring;
-	struct mlx4_en_rx_ring *rx_ring;
 	int rx_index = 0;
 	int tx_index = 0;
 	int err = 0;
@@ -572,10 +571,15 @@ int mlx4_en_start_port(struct net_device *dev)
 	dev->mtu = min(dev->mtu, priv->max_mtu);
 	mlx4_en_calc_rx_buf(dev);
 	mlx4_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
+
 	/* Configure rx cq's and rings */
+	err = mlx4_en_activate_rx_rings(priv);
+	if (err) {
+		mlx4_err(mdev, "Failed to activate RX rings\n");
+		return err;
+	}
 	for (i = 0; i < priv->rx_ring_num; i++) {
 		cq = &priv->rx_cq[i];
-		rx_ring = &priv->rx_ring[i];

 		err = mlx4_en_activate_cq(priv, cq);
 		if (err) {
@@ -591,20 +595,14 @@ int mlx4_en_start_port(struct net_device *dev)
 			goto cq_err;
 		}
 		mlx4_en_arm_cq(priv, cq);
-
+		priv->rx_ring[i].cqn = cq->mcq.cqn;
 		++rx_index;
 	}

-	err = mlx4_en_activate_rx_rings(priv);
-	if (err) {
-		mlx4_err(mdev, "Failed to activate RX rings\n");
-		goto cq_err;
-	}
-
 	err = mlx4_en_config_rss_steer(priv);
 	if (err) {
 		mlx4_err(mdev, "Failed configuring rss steering\n");
-		goto rx_err;
+		goto cq_err;
 	}

 	/* Configure tx cq's and rings */
@@ -691,12 +689,11 @@ tx_err:
 	}

 	mlx4_en_release_rss_steer(priv);
-rx_err:
-	for (i = 0; i < priv->rx_ring_num; i++)
-		mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
 cq_err:
 	while (rx_index--)
 		mlx4_en_deactivate_cq(priv, &priv->rx_cq[rx_index]);
+	for (i = 0; i < priv->rx_ring_num; i++)
+		mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);

 	return err; /* need to close devices */
 }
diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 9ee873e..6bfab6e 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -202,12 +202,35 @@ static inline void mlx4_en_update_rx_prod_db(struct mlx4_en_rx_ring *ring)
 	*ring->wqres.db.db = cpu_to_be32(ring->prod & 0xffff);
 }

+static void mlx4_en_free_rx_desc(struct mlx4_en_priv *priv,
+				 struct mlx4_en_rx_ring *ring,
+				 int index)
+{
+	struct mlx4_en_dev *mdev = priv->mdev;
+	struct skb_frag_struct *skb_frags;
+	struct mlx4_en_rx_desc *rx_desc = ring->buf + (index << ring->log_stride);
+	dma_addr_t dma;
+	int nr;
+
+	skb_frags = ring->rx_info + (index << priv->log_rx_info);
+	for (nr = 0; nr < priv->num_frags; nr++) {
+		mlx4_dbg(DRV, priv, "Freeing fragment:%d\n", nr);
+		dma = be64_to_cpu(rx_desc->data[nr].addr);
+
+		mlx4_dbg(DRV, priv, "Unmaping buffer at dma:0x%llx\n", (u64) dma);
+		pci_unmap_single(mdev->pdev, dma, skb_frags[nr].size,
+				 PCI_DMA_FROMDEVICE);
+		put_page(skb_frags[nr].page);
+	}
+}
+
 static int mlx4_en_fill_rx_buffers(struct mlx4_en_priv *priv)
 {
 	struct mlx4_en_dev *mdev = priv->mdev;
 	struct mlx4_en_rx_ring *ring;
 	int ring_ind;
 	int buf_ind;
+	int new_size;

 	for (buf_ind = 0; buf_ind < priv->prof->rx_ring_size; buf_ind++) {
 		for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {
@@ -220,18 +243,30 @@ static int mlx4_en_fill_rx_buffers(struct mlx4_en_priv *priv)
 						       "enough rx buffers\n");
 					return -ENOMEM;
 				} else {
-					if (netif_msg_rx_err(priv))
-						mlx4_warn(mdev,
-							  "Only %d buffers allocated\n",
-							  ring->actual_size);
-					goto out;
+					new_size = rounddown_pow_of_two(ring->actual_size);
+					mlx4_warn(mdev, "Only %d buffers allocated "
+							"reducing ring size to %d",
+						  ring->actual_size, new_size);
+					goto reduce_rings;
 				}
 			}
 			ring->actual_size++;
 			ring->prod++;
 		}
 	}
-out:
+	return 0;
+
+reduce_rings:
+	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {
+		ring = &priv->rx_ring[ring_ind];
+		while (ring->actual_size > new_size) {
+			ring->actual_size--;
+			ring->prod--;
+			mlx4_en_free_rx_desc(priv, ring, ring->actual_size);
+		}
+		ring->size_mask = ring->actual_size - 1;
+	}
+
 	return 0;
 }

@@ -255,7 +290,7 @@ static int mlx4_en_fill_rx_buf(struct net_device *dev,
 		++num;
 		++ring->prod;
 	}
-	if ((u32) (ring->prod - ring->cons) == ring->size)
+	if ((u32) (ring->prod - ring->cons) == ring->actual_size)
 		ring->full = 1;

 	return num;
@@ -264,33 +299,17 @@ static int mlx4_en_fill_rx_buf(struct net_device *dev,
 static void mlx4_en_free_rx_buf(struct mlx4_en_priv *priv,
 				struct mlx4_en_rx_ring *ring)
 {
-	struct mlx4_en_dev *mdev = priv->mdev;
-	struct skb_frag_struct *skb_frags;
-	struct mlx4_en_rx_desc *rx_desc;
-	dma_addr_t dma;
 	int index;
-	int nr;

 	mlx4_dbg(DRV, priv, "Freeing Rx buf - cons:%d prod:%d\n",
 			ring->cons, ring->prod);

 	/* Unmap and free Rx buffers */
-	BUG_ON((u32) (ring->prod - ring->cons) > ring->size);
+	BUG_ON((u32) (ring->prod - ring->cons) > ring->actual_size);
 	while (ring->cons != ring->prod) {
 		index = ring->cons & ring->size_mask;
-		rx_desc = ring->buf + (index << ring->log_stride);
-		skb_frags = ring->rx_info + (index << priv->log_rx_info);
 		mlx4_dbg(DRV, priv, "Processing descriptor:%d\n", index);
-
-		for (nr = 0; nr < priv->num_frags; nr++) {
-			mlx4_dbg(DRV, priv, "Freeing fragment:%d\n", nr);
-			dma = be64_to_cpu(rx_desc->data[nr].addr);
-
-			mlx4_dbg(DRV, priv, "Unmaping buffer at dma:0x%llx\n", (u64) dma);
-			pci_unmap_single(mdev->pdev, dma, skb_frags[nr].size,
-					 PCI_DMA_FROMDEVICE);
-			put_page(skb_frags[nr].page);
-		}
+		mlx4_en_free_rx_desc(priv, ring, index);
 		++ring->cons;
 	}
 }
@@ -454,7 +473,7 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 		mlx4_en_update_rx_prod_db(ring);

 		/* Configure SRQ representing the ring */
-		ring->srq.max    = ring->size;
+		ring->srq.max    = ring->actual_size;
 		ring->srq.max_gs = max_gs;
 		ring->srq.wqe_shift = ilog2(ring->stride);

-- 
1.6.0

Re: [PATCH 2/2] mlx4_en: Fix partial rings feature

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Sun, 24 May 2009 16:17:11 +0300

> In case of allocation failure, the actual ring size is rounded down to
> nearest power of 2. The remaining descriptors are freed.
> The CQ and SRQ are allocated with the actual size and the mask is updated.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Applied to net-next-2.6

[net-2.6 PATCH] mlx4_en: Fix a kernel panic when waking tx queue

[Yevgeny Petrilin]

This patch fixes a kernel panic when waking the TX queue
from irq context.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_tx.c |   11 ++++++-----
 1 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index ac6fc49..bbb082b 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -326,7 +326,8 @@ void mlx4_en_set_prio_map(struct mlx4_en_priv *priv, u16 *prio_map, u32 ring_num
 	}
 }

-static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
+static void mlx4_en_process_tx_cq(struct net_device *dev,
+				  struct mlx4_en_cq *cq, int poll)
 {
 	struct mlx4_en_priv *priv = netdev_priv(dev);
 	struct mlx4_cq *mcq = &cq->mcq;
@@ -383,7 +384,7 @@ static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
 	ring->cons += txbbs_skipped;

 	/* Wakeup Tx queue if this ring stopped it */
-	if (unlikely(ring->blocked)) {
+	if (unlikely(ring->blocked) && poll) {
 		if ((u32) (ring->prod - ring->cons) <=
 		     ring->size - HEADROOM - MAX_DESC_TXBBS) {

@@ -411,7 +412,7 @@ void mlx4_en_tx_irq(struct mlx4_cq *mcq)

 	if (!spin_trylock(&ring->comp_lock))
 		return;
-	mlx4_en_process_tx_cq(cq->dev, cq);
+	mlx4_en_process_tx_cq(cq->dev, cq, 0);
 	mod_timer(&cq->timer, jiffies + 1);
 	spin_unlock(&ring->comp_lock);
 }
@@ -430,7 +431,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)
 		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
 		return;
 	}
-	mlx4_en_process_tx_cq(cq->dev, cq);
+	mlx4_en_process_tx_cq(cq->dev, cq, 1);
 	inflight = (u32) (ring->prod - ring->cons - ring->last_nr_txbb);

 	/* If there are still packets in flight and the timer has not already
@@ -483,7 +484,7 @@ static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)
 	/* Poll the CQ every mlx4_en_TX_MODER_POLL packets */
 	if ((++ring->poll_cnt & (MLX4_EN_TX_POLL_MODER - 1)) == 0)
 		if (spin_trylock(&ring->comp_lock)) {
-			mlx4_en_process_tx_cq(priv->dev, cq);
+			mlx4_en_process_tx_cq(priv->dev, cq, 1);
 			spin_unlock(&ring->comp_lock);
 		}
 }
-- 
1.6.0

Re: [net-2.6 PATCH] mlx4_en: Fix a kernel panic when waking tx queue

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Mon, 25 May 2009 11:32:42 +0300

> This patch fixes a kernel panic when waking the TX queue
> from irq context.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

Why does it fix a panic?  What does the panic look like?
What causes it?

Commit message is way too terse, please fix this and resubmit.

Re: [net-2.6 PATCH] mlx4_en: Fix a kernel panic when waking tx queue

[Yevgeny Petrilin]

David Miller wrote:
> From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
> Date: Mon, 25 May 2009 11:32:42 +0300
> 
>> This patch fixes a kernel panic when waking the TX queue
>> from irq context.
>>
>> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
> 
> Why does it fix a panic?  What does the panic look like?
> What causes it?
> 
> Commit message is way too terse, please fix this and resubmit.
> 
Hello Dave,

I did some further investigation on this case.
The patch that I sent was only handling the symptoms instead of fixing the problem.
The panic was caused due to a race when handling TX completions and waking the queue if it was stopped.
When the transmit queue gets full, the driver enables interrupts for TX completions in order to handle
them as fast as possible.
The result was that the driver approached the transmit queue from several different contexts and this
was the bug.
I prepared a new patch fixing the race, it was running for several hours now, the bug didn't happen.

Thanks,
Yevgeny

[PATCH 2/8] mlx4_en: Moved all module parameters handling to en_main.c

[Yevgeny Petrilin]

en_params.c file now only handles Ethtool functionality

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_main.c   |   54 +++++++++++++++++++++++++++++++++++++++
 drivers/net/mlx4/en_params.c |   58 ------------------------------------------
 drivers/net/mlx4/mlx4_en.h   |    2 -
 3 files changed, 54 insertions(+), 60 deletions(-)

diff --git a/drivers/net/mlx4/en_main.c b/drivers/net/mlx4/en_main.c
index 510633f..23955d8 100644
--- a/drivers/net/mlx4/en_main.c
+++ b/drivers/net/mlx4/en_main.c
@@ -51,6 +51,60 @@ static const char mlx4_en_version[] =
 	DRV_NAME ": Mellanox ConnectX HCA Ethernet driver v"
 	DRV_VERSION " (" DRV_RELDATE ")\n";
 
+#define MLX4_EN_PARM_INT(X, def_val, desc) \
+	static unsigned int X = def_val;\
+	module_param(X , uint, 0444); \
+	MODULE_PARM_DESC(X, desc);
+
+
+/*
+ * Device scope module parameters
+ */
+
+
+/* Use a XOR rathern than Toeplitz hash function for RSS */
+MLX4_EN_PARM_INT(rss_xor, 0, "Use XOR hash function for RSS");
+
+/* RSS hash type mask - default to <saddr, daddr, sport, dport> */
+MLX4_EN_PARM_INT(rss_mask, 0xf, "RSS hash type bitmask");
+
+/* Number of LRO sessions per Rx ring (rounded up to a power of two) */
+MLX4_EN_PARM_INT(num_lro, MLX4_EN_MAX_LRO_DESCRIPTORS,
+		 "Number of LRO sessions per ring or disabled (0)");
+
+/* Priority pausing */
+MLX4_EN_PARM_INT(pfctx, 0, "Priority based Flow Control policy on TX[7:0]."
+			   " Per priority bit mask");
+MLX4_EN_PARM_INT(pfcrx, 0, "Priority based Flow Control policy on RX[7:0]."
+			   " Per priority bit mask");
+
+static int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
+{
+	struct mlx4_en_profile *params = &mdev->profile;
+	int i;
+
+	params->rss_xor = (rss_xor != 0);
+	params->rss_mask = rss_mask & 0x1f;
+	params->num_lro = min_t(int, num_lro , MLX4_EN_MAX_LRO_DESCRIPTORS);
+	for (i = 1; i <= MLX4_MAX_PORTS; i++) {
+		params->prof[i].rx_pause = 1;
+		params->prof[i].rx_ppp = pfcrx;
+		params->prof[i].tx_pause = 1;
+		params->prof[i].tx_ppp = pfctx;
+		params->prof[i].tx_ring_size = MLX4_EN_DEF_TX_RING_SIZE;
+		params->prof[i].rx_ring_size = MLX4_EN_DEF_RX_RING_SIZE;
+	}
+	if (pfcrx || pfctx) {
+		params->prof[1].tx_ring_num = MLX4_EN_TX_RING_NUM;
+		params->prof[2].tx_ring_num = MLX4_EN_TX_RING_NUM;
+	} else {
+		params->prof[1].tx_ring_num = 1;
+		params->prof[2].tx_ring_num = 1;
+	}
+
+	return 0;
+}
+
 static void mlx4_en_event(struct mlx4_dev *dev, void *endev_ptr,
 			  enum mlx4_dev_event event, int port)
 {
diff --git a/drivers/net/mlx4/en_params.c b/drivers/net/mlx4/en_params.c
index 3290fec..91d8116 100644
--- a/drivers/net/mlx4/en_params.c
+++ b/drivers/net/mlx4/en_params.c
@@ -38,64 +38,6 @@
 #include "mlx4_en.h"
 #include "en_port.h"
 
-#define MLX4_EN_PARM_INT(X, def_val, desc) \
-	static unsigned int X = def_val;\
-	module_param(X , uint, 0444); \
-	MODULE_PARM_DESC(X, desc);
-
-
-/*
- * Device scope module parameters
- */
-
-
-/* Use a XOR rathern than Toeplitz hash function for RSS */
-MLX4_EN_PARM_INT(rss_xor, 0, "Use XOR hash function for RSS");
-
-/* RSS hash type mask - default to <saddr, daddr, sport, dport> */
-MLX4_EN_PARM_INT(rss_mask, 0xf, "RSS hash type bitmask");
-
-/* Number of LRO sessions per Rx ring (rounded up to a power of two) */
-MLX4_EN_PARM_INT(num_lro, MLX4_EN_MAX_LRO_DESCRIPTORS,
-		 "Number of LRO sessions per ring or disabled (0)");
-
-/* Priority pausing */
-MLX4_EN_PARM_INT(pfctx, 0, "Priority based Flow Control policy on TX[7:0]."
-			   " Per priority bit mask");
-MLX4_EN_PARM_INT(pfcrx, 0, "Priority based Flow Control policy on RX[7:0]."
-			   " Per priority bit mask");
-
-int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
-{
-	struct mlx4_en_profile *params = &mdev->profile;
-	int i;
-
-	params->rss_xor = (rss_xor != 0);
-	params->rss_mask = rss_mask & 0x1f;
-	params->num_lro = min_t(int, num_lro , MLX4_EN_MAX_LRO_DESCRIPTORS);
-	for (i = 1; i <= MLX4_MAX_PORTS; i++) {
-		params->prof[i].rx_pause = 1;
-		params->prof[i].rx_ppp = pfcrx;
-		params->prof[i].tx_pause = 1;
-		params->prof[i].tx_ppp = pfctx;
-		params->prof[i].tx_ring_size = MLX4_EN_DEF_TX_RING_SIZE;
-		params->prof[i].rx_ring_size = MLX4_EN_DEF_RX_RING_SIZE;
-	}
-	if (pfcrx || pfctx) {
-		params->prof[1].tx_ring_num = MLX4_EN_TX_RING_NUM;
-		params->prof[2].tx_ring_num = MLX4_EN_TX_RING_NUM;
-	} else {
-		params->prof[1].tx_ring_num = 1;
-		params->prof[2].tx_ring_num = 1;
-	}
-
-	return 0;
-}
-
-
-/*
- * Ethtool support
- */
 
 static void mlx4_en_update_lro_stats(struct mlx4_en_priv *priv)
 {
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index c92b382..174ae86 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -516,8 +516,6 @@ void mlx4_en_stop_port(struct net_device *dev);
 void mlx4_en_free_resources(struct mlx4_en_priv *priv);
 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv);
 
-int mlx4_en_get_profile(struct mlx4_en_dev *mdev);
-
 int mlx4_en_create_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 		      int entries, int ring, enum cq_type mode);
 void mlx4_en_destroy_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
-- 
1.6.0

[PATCH 3/8] mlx4_en renamed en_params.c to en_ethtool.c

[Yevgeny Petrilin]

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/Makefile     |    2 +-
 drivers/net/mlx4/en_ethtool.c |  427 +++++++++++++++++++++++++++++++++++++++++
 drivers/net/mlx4/en_params.c  |  427 -----------------------------------------
 3 files changed, 428 insertions(+), 428 deletions(-)
 create mode 100644 drivers/net/mlx4/en_ethtool.c
 delete mode 100644 drivers/net/mlx4/en_params.c

diff --git a/drivers/net/mlx4/Makefile b/drivers/net/mlx4/Makefile
index 21040a0..1fd068e 100644
--- a/drivers/net/mlx4/Makefile
+++ b/drivers/net/mlx4/Makefile
@@ -5,5 +5,5 @@ mlx4_core-y :=	alloc.o catas.o cmd.o cq.o eq.o fw.o icm.o intf.o main.o mcg.o \
 
 obj-$(CONFIG_MLX4_EN)               += mlx4_en.o
 
-mlx4_en-y := 	en_main.o en_tx.o en_rx.o en_params.o en_port.o en_cq.o \
+mlx4_en-y := 	en_main.o en_tx.o en_rx.o en_ethtool.o en_port.o en_cq.o \
 		en_resources.o en_netdev.o
diff --git a/drivers/net/mlx4/en_ethtool.c b/drivers/net/mlx4/en_ethtool.c
new file mode 100644
index 0000000..91d8116
--- /dev/null
+++ b/drivers/net/mlx4/en_ethtool.c
@@ -0,0 +1,427 @@
+/*
+ * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/ethtool.h>
+#include <linux/netdevice.h>
+
+#include "mlx4_en.h"
+#include "en_port.h"
+
+
+static void mlx4_en_update_lro_stats(struct mlx4_en_priv *priv)
+{
+	int i;
+
+	priv->port_stats.lro_aggregated = 0;
+	priv->port_stats.lro_flushed = 0;
+	priv->port_stats.lro_no_desc = 0;
+
+	for (i = 0; i < priv->rx_ring_num; i++) {
+		priv->port_stats.lro_aggregated += priv->rx_ring[i].lro.stats.aggregated;
+		priv->port_stats.lro_flushed += priv->rx_ring[i].lro.stats.flushed;
+		priv->port_stats.lro_no_desc += priv->rx_ring[i].lro.stats.no_desc;
+	}
+}
+
+static void
+mlx4_en_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *drvinfo)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	struct mlx4_en_dev *mdev = priv->mdev;
+
+	sprintf(drvinfo->driver, DRV_NAME " (%s)", mdev->dev->board_id);
+	strncpy(drvinfo->version, DRV_VERSION " (" DRV_RELDATE ")", 32);
+	sprintf(drvinfo->fw_version, "%d.%d.%d",
+		(u16) (mdev->dev->caps.fw_ver >> 32),
+		(u16) ((mdev->dev->caps.fw_ver >> 16) & 0xffff),
+		(u16) (mdev->dev->caps.fw_ver & 0xffff));
+	strncpy(drvinfo->bus_info, pci_name(mdev->dev->pdev), 32);
+	drvinfo->n_stats = 0;
+	drvinfo->regdump_len = 0;
+	drvinfo->eedump_len = 0;
+}
+
+static u32 mlx4_en_get_tso(struct net_device *dev)
+{
+	return (dev->features & NETIF_F_TSO) != 0;
+}
+
+static int mlx4_en_set_tso(struct net_device *dev, u32 data)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+
+	if (data) {
+		if (!priv->mdev->LSO_support)
+			return -EPERM;
+		dev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
+	} else
+		dev->features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
+	return 0;
+}
+
+static u32 mlx4_en_get_rx_csum(struct net_device *dev)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	return priv->rx_csum;
+}
+
+static int mlx4_en_set_rx_csum(struct net_device *dev, u32 data)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	priv->rx_csum = (data != 0);
+	return 0;
+}
+
+static const char main_strings[][ETH_GSTRING_LEN] = {
+	"rx_packets", "tx_packets", "rx_bytes", "tx_bytes", "rx_errors",
+	"tx_errors", "rx_dropped", "tx_dropped", "multicast", "collisions",
+	"rx_length_errors", "rx_over_errors", "rx_crc_errors",
+	"rx_frame_errors", "rx_fifo_errors", "rx_missed_errors",
+	"tx_aborted_errors", "tx_carrier_errors", "tx_fifo_errors",
+	"tx_heartbeat_errors", "tx_window_errors",
+
+	/* port statistics */
+	"lro_aggregated", "lro_flushed", "lro_no_desc", "tso_packets",
+	"queue_stopped", "wake_queue", "tx_timeout", "rx_alloc_failed",
+	"rx_csum_good", "rx_csum_none", "tx_chksum_offload",
+
+	/* packet statistics */
+	"broadcast", "rx_prio_0", "rx_prio_1", "rx_prio_2", "rx_prio_3",
+	"rx_prio_4", "rx_prio_5", "rx_prio_6", "rx_prio_7", "tx_prio_0",
+	"tx_prio_1", "tx_prio_2", "tx_prio_3", "tx_prio_4", "tx_prio_5",
+	"tx_prio_6", "tx_prio_7",
+};
+#define NUM_MAIN_STATS	21
+#define NUM_ALL_STATS	(NUM_MAIN_STATS + NUM_PORT_STATS + NUM_PKT_STATS + NUM_PERF_STATS)
+
+static u32 mlx4_en_get_msglevel(struct net_device *dev)
+{
+	return ((struct mlx4_en_priv *) netdev_priv(dev))->msg_enable;
+}
+
+static void mlx4_en_set_msglevel(struct net_device *dev, u32 val)
+{
+	((struct mlx4_en_priv *) netdev_priv(dev))->msg_enable = val;
+}
+
+static void mlx4_en_get_wol(struct net_device *netdev,
+			    struct ethtool_wolinfo *wol)
+{
+	wol->supported = 0;
+	wol->wolopts = 0;
+
+	return;
+}
+
+static int mlx4_en_get_sset_count(struct net_device *dev, int sset)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+
+	if (sset != ETH_SS_STATS)
+		return -EOPNOTSUPP;
+
+	return NUM_ALL_STATS + (priv->tx_ring_num + priv->rx_ring_num) * 2;
+}
+
+static void mlx4_en_get_ethtool_stats(struct net_device *dev,
+		struct ethtool_stats *stats, uint64_t *data)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	int index = 0;
+	int i;
+
+	spin_lock_bh(&priv->stats_lock);
+
+	mlx4_en_update_lro_stats(priv);
+
+	for (i = 0; i < NUM_MAIN_STATS; i++)
+		data[index++] = ((unsigned long *) &priv->stats)[i];
+	for (i = 0; i < NUM_PORT_STATS; i++)
+		data[index++] = ((unsigned long *) &priv->port_stats)[i];
+	for (i = 0; i < priv->tx_ring_num; i++) {
+		data[index++] = priv->tx_ring[i].packets;
+		data[index++] = priv->tx_ring[i].bytes;
+	}
+	for (i = 0; i < priv->rx_ring_num; i++) {
+		data[index++] = priv->rx_ring[i].packets;
+		data[index++] = priv->rx_ring[i].bytes;
+	}
+	for (i = 0; i < NUM_PKT_STATS; i++)
+		data[index++] = ((unsigned long *) &priv->pkstats)[i];
+	spin_unlock_bh(&priv->stats_lock);
+
+}
+
+static void mlx4_en_get_strings(struct net_device *dev,
+				uint32_t stringset, uint8_t *data)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	int index = 0;
+	int i;
+
+	if (stringset != ETH_SS_STATS)
+		return;
+
+	/* Add main counters */
+	for (i = 0; i < NUM_MAIN_STATS; i++)
+		strcpy(data + (index++) * ETH_GSTRING_LEN, main_strings[i]);
+	for (i = 0; i < NUM_PORT_STATS; i++)
+		strcpy(data + (index++) * ETH_GSTRING_LEN,
+			main_strings[i + NUM_MAIN_STATS]);
+	for (i = 0; i < priv->tx_ring_num; i++) {
+		sprintf(data + (index++) * ETH_GSTRING_LEN,
+			"tx%d_packets", i);
+		sprintf(data + (index++) * ETH_GSTRING_LEN,
+			"tx%d_bytes", i);
+	}
+	for (i = 0; i < priv->rx_ring_num; i++) {
+		sprintf(data + (index++) * ETH_GSTRING_LEN,
+			"rx%d_packets", i);
+		sprintf(data + (index++) * ETH_GSTRING_LEN,
+			"rx%d_bytes", i);
+	}
+	for (i = 0; i < NUM_PKT_STATS; i++)
+		strcpy(data + (index++) * ETH_GSTRING_LEN,
+			main_strings[i + NUM_MAIN_STATS + NUM_PORT_STATS]);
+}
+
+static int mlx4_en_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
+{
+	cmd->autoneg = AUTONEG_DISABLE;
+	cmd->supported = SUPPORTED_10000baseT_Full;
+	cmd->advertising = SUPPORTED_10000baseT_Full;
+	if (netif_carrier_ok(dev)) {
+		cmd->speed = SPEED_10000;
+		cmd->duplex = DUPLEX_FULL;
+	} else {
+		cmd->speed = -1;
+		cmd->duplex = -1;
+	}
+	return 0;
+}
+
+static int mlx4_en_set_settings(struct net_device *dev, struct ethtool_cmd *cmd)
+{
+	if ((cmd->autoneg == AUTONEG_ENABLE) ||
+	    (cmd->speed != SPEED_10000) || (cmd->duplex != DUPLEX_FULL))
+		return -EINVAL;
+
+	/* Nothing to change */
+	return 0;
+}
+
+static int mlx4_en_get_coalesce(struct net_device *dev,
+			      struct ethtool_coalesce *coal)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+
+	coal->tx_coalesce_usecs = 0;
+	coal->tx_max_coalesced_frames = 0;
+	coal->rx_coalesce_usecs = priv->rx_usecs;
+	coal->rx_max_coalesced_frames = priv->rx_frames;
+
+	coal->pkt_rate_low = priv->pkt_rate_low;
+	coal->rx_coalesce_usecs_low = priv->rx_usecs_low;
+	coal->pkt_rate_high = priv->pkt_rate_high;
+	coal->rx_coalesce_usecs_high = priv->rx_usecs_high;
+	coal->rate_sample_interval = priv->sample_interval;
+	coal->use_adaptive_rx_coalesce = priv->adaptive_rx_coal;
+	return 0;
+}
+
+static int mlx4_en_set_coalesce(struct net_device *dev,
+			      struct ethtool_coalesce *coal)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	int err, i;
+
+	priv->rx_frames = (coal->rx_max_coalesced_frames ==
+			   MLX4_EN_AUTO_CONF) ?
+				MLX4_EN_RX_COAL_TARGET /
+				priv->dev->mtu + 1 :
+				coal->rx_max_coalesced_frames;
+	priv->rx_usecs = (coal->rx_coalesce_usecs ==
+			  MLX4_EN_AUTO_CONF) ?
+				MLX4_EN_RX_COAL_TIME :
+				coal->rx_coalesce_usecs;
+
+	/* Set adaptive coalescing params */
+	priv->pkt_rate_low = coal->pkt_rate_low;
+	priv->rx_usecs_low = coal->rx_coalesce_usecs_low;
+	priv->pkt_rate_high = coal->pkt_rate_high;
+	priv->rx_usecs_high = coal->rx_coalesce_usecs_high;
+	priv->sample_interval = coal->rate_sample_interval;
+	priv->adaptive_rx_coal = coal->use_adaptive_rx_coalesce;
+	priv->last_moder_time = MLX4_EN_AUTO_CONF;
+	if (priv->adaptive_rx_coal)
+		return 0;
+
+	for (i = 0; i < priv->rx_ring_num; i++) {
+		priv->rx_cq[i].moder_cnt = priv->rx_frames;
+		priv->rx_cq[i].moder_time = priv->rx_usecs;
+		err = mlx4_en_set_cq_moder(priv, &priv->rx_cq[i]);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static int mlx4_en_set_pauseparam(struct net_device *dev,
+				struct ethtool_pauseparam *pause)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	struct mlx4_en_dev *mdev = priv->mdev;
+	int err;
+
+	priv->prof->tx_pause = pause->tx_pause != 0;
+	priv->prof->rx_pause = pause->rx_pause != 0;
+	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
+				    priv->rx_skb_size + ETH_FCS_LEN,
+				    priv->prof->tx_pause,
+				    priv->prof->tx_ppp,
+				    priv->prof->rx_pause,
+				    priv->prof->rx_ppp);
+	if (err)
+		en_err(priv, "Failed setting pause params\n");
+
+	return err;
+}
+
+static void mlx4_en_get_pauseparam(struct net_device *dev,
+				 struct ethtool_pauseparam *pause)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+
+	pause->tx_pause = priv->prof->tx_pause;
+	pause->rx_pause = priv->prof->rx_pause;
+}
+
+static int mlx4_en_set_ringparam(struct net_device *dev,
+				 struct ethtool_ringparam *param)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	struct mlx4_en_dev *mdev = priv->mdev;
+	u32 rx_size, tx_size;
+	int port_up = 0;
+	int err = 0;
+
+	if (param->rx_jumbo_pending || param->rx_mini_pending)
+		return -EINVAL;
+
+	rx_size = roundup_pow_of_two(param->rx_pending);
+	rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
+	rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
+	tx_size = roundup_pow_of_two(param->tx_pending);
+	tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
+	tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
+
+	if (rx_size == priv->prof->rx_ring_size &&
+	    tx_size == priv->prof->tx_ring_size)
+		return 0;
+
+	mutex_lock(&mdev->state_lock);
+	if (priv->port_up) {
+		port_up = 1;
+		mlx4_en_stop_port(dev);
+	}
+
+	mlx4_en_free_resources(priv);
+
+	priv->prof->tx_ring_size = tx_size;
+	priv->prof->rx_ring_size = rx_size;
+
+	err = mlx4_en_alloc_resources(priv);
+	if (err) {
+		en_err(priv, "Failed reallocating port resources\n");
+		goto out;
+	}
+	if (port_up) {
+		err = mlx4_en_start_port(dev);
+		if (err)
+			en_err(priv, "Failed starting port\n");
+	}
+
+out:
+	mutex_unlock(&mdev->state_lock);
+	return err;
+}
+
+static void mlx4_en_get_ringparam(struct net_device *dev,
+				  struct ethtool_ringparam *param)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	struct mlx4_en_dev *mdev = priv->mdev;
+
+	memset(param, 0, sizeof(*param));
+	param->rx_max_pending = MLX4_EN_MAX_RX_SIZE;
+	param->tx_max_pending = MLX4_EN_MAX_TX_SIZE;
+	param->rx_pending = mdev->profile.prof[priv->port].rx_ring_size;
+	param->tx_pending = mdev->profile.prof[priv->port].tx_ring_size;
+}
+
+const struct ethtool_ops mlx4_en_ethtool_ops = {
+	.get_drvinfo = mlx4_en_get_drvinfo,
+	.get_settings = mlx4_en_get_settings,
+	.set_settings = mlx4_en_set_settings,
+#ifdef NETIF_F_TSO
+	.get_tso = mlx4_en_get_tso,
+	.set_tso = mlx4_en_set_tso,
+#endif
+	.get_sg = ethtool_op_get_sg,
+	.set_sg = ethtool_op_set_sg,
+	.get_link = ethtool_op_get_link,
+	.get_rx_csum = mlx4_en_get_rx_csum,
+	.set_rx_csum = mlx4_en_set_rx_csum,
+	.get_tx_csum = ethtool_op_get_tx_csum,
+	.set_tx_csum = ethtool_op_set_tx_ipv6_csum,
+	.get_strings = mlx4_en_get_strings,
+	.get_sset_count = mlx4_en_get_sset_count,
+	.get_ethtool_stats = mlx4_en_get_ethtool_stats,
+	.get_wol = mlx4_en_get_wol,
+	.get_msglevel = mlx4_en_get_msglevel,
+	.set_msglevel = mlx4_en_set_msglevel,
+	.get_coalesce = mlx4_en_get_coalesce,
+	.set_coalesce = mlx4_en_set_coalesce,
+	.get_pauseparam = mlx4_en_get_pauseparam,
+	.set_pauseparam = mlx4_en_set_pauseparam,
+	.get_ringparam = mlx4_en_get_ringparam,
+	.set_ringparam = mlx4_en_set_ringparam,
+	.get_flags = ethtool_op_get_flags,
+	.set_flags = ethtool_op_set_flags,
+};
+
+
+
+
+
diff --git a/drivers/net/mlx4/en_params.c b/drivers/net/mlx4/en_params.c
deleted file mode 100644
index 91d8116..0000000
--- a/drivers/net/mlx4/en_params.c
+++ /dev/null
@@ -1,427 +0,0 @@
-/*
- * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- *
- */
-
-#include <linux/kernel.h>
-#include <linux/ethtool.h>
-#include <linux/netdevice.h>
-
-#include "mlx4_en.h"
-#include "en_port.h"
-
-
-static void mlx4_en_update_lro_stats(struct mlx4_en_priv *priv)
-{
-	int i;
-
-	priv->port_stats.lro_aggregated = 0;
-	priv->port_stats.lro_flushed = 0;
-	priv->port_stats.lro_no_desc = 0;
-
-	for (i = 0; i < priv->rx_ring_num; i++) {
-		priv->port_stats.lro_aggregated += priv->rx_ring[i].lro.stats.aggregated;
-		priv->port_stats.lro_flushed += priv->rx_ring[i].lro.stats.flushed;
-		priv->port_stats.lro_no_desc += priv->rx_ring[i].lro.stats.no_desc;
-	}
-}
-
-static void
-mlx4_en_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *drvinfo)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
-
-	sprintf(drvinfo->driver, DRV_NAME " (%s)", mdev->dev->board_id);
-	strncpy(drvinfo->version, DRV_VERSION " (" DRV_RELDATE ")", 32);
-	sprintf(drvinfo->fw_version, "%d.%d.%d",
-		(u16) (mdev->dev->caps.fw_ver >> 32),
-		(u16) ((mdev->dev->caps.fw_ver >> 16) & 0xffff),
-		(u16) (mdev->dev->caps.fw_ver & 0xffff));
-	strncpy(drvinfo->bus_info, pci_name(mdev->dev->pdev), 32);
-	drvinfo->n_stats = 0;
-	drvinfo->regdump_len = 0;
-	drvinfo->eedump_len = 0;
-}
-
-static u32 mlx4_en_get_tso(struct net_device *dev)
-{
-	return (dev->features & NETIF_F_TSO) != 0;
-}
-
-static int mlx4_en_set_tso(struct net_device *dev, u32 data)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-
-	if (data) {
-		if (!priv->mdev->LSO_support)
-			return -EPERM;
-		dev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
-	} else
-		dev->features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
-	return 0;
-}
-
-static u32 mlx4_en_get_rx_csum(struct net_device *dev)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	return priv->rx_csum;
-}
-
-static int mlx4_en_set_rx_csum(struct net_device *dev, u32 data)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	priv->rx_csum = (data != 0);
-	return 0;
-}
-
-static const char main_strings[][ETH_GSTRING_LEN] = {
-	"rx_packets", "tx_packets", "rx_bytes", "tx_bytes", "rx_errors",
-	"tx_errors", "rx_dropped", "tx_dropped", "multicast", "collisions",
-	"rx_length_errors", "rx_over_errors", "rx_crc_errors",
-	"rx_frame_errors", "rx_fifo_errors", "rx_missed_errors",
-	"tx_aborted_errors", "tx_carrier_errors", "tx_fifo_errors",
-	"tx_heartbeat_errors", "tx_window_errors",
-
-	/* port statistics */
-	"lro_aggregated", "lro_flushed", "lro_no_desc", "tso_packets",
-	"queue_stopped", "wake_queue", "tx_timeout", "rx_alloc_failed",
-	"rx_csum_good", "rx_csum_none", "tx_chksum_offload",
-
-	/* packet statistics */
-	"broadcast", "rx_prio_0", "rx_prio_1", "rx_prio_2", "rx_prio_3",
-	"rx_prio_4", "rx_prio_5", "rx_prio_6", "rx_prio_7", "tx_prio_0",
-	"tx_prio_1", "tx_prio_2", "tx_prio_3", "tx_prio_4", "tx_prio_5",
-	"tx_prio_6", "tx_prio_7",
-};
-#define NUM_MAIN_STATS	21
-#define NUM_ALL_STATS	(NUM_MAIN_STATS + NUM_PORT_STATS + NUM_PKT_STATS + NUM_PERF_STATS)
-
-static u32 mlx4_en_get_msglevel(struct net_device *dev)
-{
-	return ((struct mlx4_en_priv *) netdev_priv(dev))->msg_enable;
-}
-
-static void mlx4_en_set_msglevel(struct net_device *dev, u32 val)
-{
-	((struct mlx4_en_priv *) netdev_priv(dev))->msg_enable = val;
-}
-
-static void mlx4_en_get_wol(struct net_device *netdev,
-			    struct ethtool_wolinfo *wol)
-{
-	wol->supported = 0;
-	wol->wolopts = 0;
-
-	return;
-}
-
-static int mlx4_en_get_sset_count(struct net_device *dev, int sset)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-
-	if (sset != ETH_SS_STATS)
-		return -EOPNOTSUPP;
-
-	return NUM_ALL_STATS + (priv->tx_ring_num + priv->rx_ring_num) * 2;
-}
-
-static void mlx4_en_get_ethtool_stats(struct net_device *dev,
-		struct ethtool_stats *stats, uint64_t *data)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	int index = 0;
-	int i;
-
-	spin_lock_bh(&priv->stats_lock);
-
-	mlx4_en_update_lro_stats(priv);
-
-	for (i = 0; i < NUM_MAIN_STATS; i++)
-		data[index++] = ((unsigned long *) &priv->stats)[i];
-	for (i = 0; i < NUM_PORT_STATS; i++)
-		data[index++] = ((unsigned long *) &priv->port_stats)[i];
-	for (i = 0; i < priv->tx_ring_num; i++) {
-		data[index++] = priv->tx_ring[i].packets;
-		data[index++] = priv->tx_ring[i].bytes;
-	}
-	for (i = 0; i < priv->rx_ring_num; i++) {
-		data[index++] = priv->rx_ring[i].packets;
-		data[index++] = priv->rx_ring[i].bytes;
-	}
-	for (i = 0; i < NUM_PKT_STATS; i++)
-		data[index++] = ((unsigned long *) &priv->pkstats)[i];
-	spin_unlock_bh(&priv->stats_lock);
-
-}
-
-static void mlx4_en_get_strings(struct net_device *dev,
-				uint32_t stringset, uint8_t *data)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	int index = 0;
-	int i;
-
-	if (stringset != ETH_SS_STATS)
-		return;
-
-	/* Add main counters */
-	for (i = 0; i < NUM_MAIN_STATS; i++)
-		strcpy(data + (index++) * ETH_GSTRING_LEN, main_strings[i]);
-	for (i = 0; i < NUM_PORT_STATS; i++)
-		strcpy(data + (index++) * ETH_GSTRING_LEN,
-			main_strings[i + NUM_MAIN_STATS]);
-	for (i = 0; i < priv->tx_ring_num; i++) {
-		sprintf(data + (index++) * ETH_GSTRING_LEN,
-			"tx%d_packets", i);
-		sprintf(data + (index++) * ETH_GSTRING_LEN,
-			"tx%d_bytes", i);
-	}
-	for (i = 0; i < priv->rx_ring_num; i++) {
-		sprintf(data + (index++) * ETH_GSTRING_LEN,
-			"rx%d_packets", i);
-		sprintf(data + (index++) * ETH_GSTRING_LEN,
-			"rx%d_bytes", i);
-	}
-	for (i = 0; i < NUM_PKT_STATS; i++)
-		strcpy(data + (index++) * ETH_GSTRING_LEN,
-			main_strings[i + NUM_MAIN_STATS + NUM_PORT_STATS]);
-}
-
-static int mlx4_en_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
-{
-	cmd->autoneg = AUTONEG_DISABLE;
-	cmd->supported = SUPPORTED_10000baseT_Full;
-	cmd->advertising = SUPPORTED_10000baseT_Full;
-	if (netif_carrier_ok(dev)) {
-		cmd->speed = SPEED_10000;
-		cmd->duplex = DUPLEX_FULL;
-	} else {
-		cmd->speed = -1;
-		cmd->duplex = -1;
-	}
-	return 0;
-}
-
-static int mlx4_en_set_settings(struct net_device *dev, struct ethtool_cmd *cmd)
-{
-	if ((cmd->autoneg == AUTONEG_ENABLE) ||
-	    (cmd->speed != SPEED_10000) || (cmd->duplex != DUPLEX_FULL))
-		return -EINVAL;
-
-	/* Nothing to change */
-	return 0;
-}
-
-static int mlx4_en_get_coalesce(struct net_device *dev,
-			      struct ethtool_coalesce *coal)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-
-	coal->tx_coalesce_usecs = 0;
-	coal->tx_max_coalesced_frames = 0;
-	coal->rx_coalesce_usecs = priv->rx_usecs;
-	coal->rx_max_coalesced_frames = priv->rx_frames;
-
-	coal->pkt_rate_low = priv->pkt_rate_low;
-	coal->rx_coalesce_usecs_low = priv->rx_usecs_low;
-	coal->pkt_rate_high = priv->pkt_rate_high;
-	coal->rx_coalesce_usecs_high = priv->rx_usecs_high;
-	coal->rate_sample_interval = priv->sample_interval;
-	coal->use_adaptive_rx_coalesce = priv->adaptive_rx_coal;
-	return 0;
-}
-
-static int mlx4_en_set_coalesce(struct net_device *dev,
-			      struct ethtool_coalesce *coal)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	int err, i;
-
-	priv->rx_frames = (coal->rx_max_coalesced_frames ==
-			   MLX4_EN_AUTO_CONF) ?
-				MLX4_EN_RX_COAL_TARGET /
-				priv->dev->mtu + 1 :
-				coal->rx_max_coalesced_frames;
-	priv->rx_usecs = (coal->rx_coalesce_usecs ==
-			  MLX4_EN_AUTO_CONF) ?
-				MLX4_EN_RX_COAL_TIME :
-				coal->rx_coalesce_usecs;
-
-	/* Set adaptive coalescing params */
-	priv->pkt_rate_low = coal->pkt_rate_low;
-	priv->rx_usecs_low = coal->rx_coalesce_usecs_low;
-	priv->pkt_rate_high = coal->pkt_rate_high;
-	priv->rx_usecs_high = coal->rx_coalesce_usecs_high;
-	priv->sample_interval = coal->rate_sample_interval;
-	priv->adaptive_rx_coal = coal->use_adaptive_rx_coalesce;
-	priv->last_moder_time = MLX4_EN_AUTO_CONF;
-	if (priv->adaptive_rx_coal)
-		return 0;
-
-	for (i = 0; i < priv->rx_ring_num; i++) {
-		priv->rx_cq[i].moder_cnt = priv->rx_frames;
-		priv->rx_cq[i].moder_time = priv->rx_usecs;
-		err = mlx4_en_set_cq_moder(priv, &priv->rx_cq[i]);
-		if (err)
-			return err;
-	}
-	return 0;
-}
-
-static int mlx4_en_set_pauseparam(struct net_device *dev,
-				struct ethtool_pauseparam *pause)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
-	int err;
-
-	priv->prof->tx_pause = pause->tx_pause != 0;
-	priv->prof->rx_pause = pause->rx_pause != 0;
-	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
-				    priv->rx_skb_size + ETH_FCS_LEN,
-				    priv->prof->tx_pause,
-				    priv->prof->tx_ppp,
-				    priv->prof->rx_pause,
-				    priv->prof->rx_ppp);
-	if (err)
-		en_err(priv, "Failed setting pause params\n");
-
-	return err;
-}
-
-static void mlx4_en_get_pauseparam(struct net_device *dev,
-				 struct ethtool_pauseparam *pause)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-
-	pause->tx_pause = priv->prof->tx_pause;
-	pause->rx_pause = priv->prof->rx_pause;
-}
-
-static int mlx4_en_set_ringparam(struct net_device *dev,
-				 struct ethtool_ringparam *param)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
-	u32 rx_size, tx_size;
-	int port_up = 0;
-	int err = 0;
-
-	if (param->rx_jumbo_pending || param->rx_mini_pending)
-		return -EINVAL;
-
-	rx_size = roundup_pow_of_two(param->rx_pending);
-	rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
-	rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
-	tx_size = roundup_pow_of_two(param->tx_pending);
-	tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
-	tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
-
-	if (rx_size == priv->prof->rx_ring_size &&
-	    tx_size == priv->prof->tx_ring_size)
-		return 0;
-
-	mutex_lock(&mdev->state_lock);
-	if (priv->port_up) {
-		port_up = 1;
-		mlx4_en_stop_port(dev);
-	}
-
-	mlx4_en_free_resources(priv);
-
-	priv->prof->tx_ring_size = tx_size;
-	priv->prof->rx_ring_size = rx_size;
-
-	err = mlx4_en_alloc_resources(priv);
-	if (err) {
-		en_err(priv, "Failed reallocating port resources\n");
-		goto out;
-	}
-	if (port_up) {
-		err = mlx4_en_start_port(dev);
-		if (err)
-			en_err(priv, "Failed starting port\n");
-	}
-
-out:
-	mutex_unlock(&mdev->state_lock);
-	return err;
-}
-
-static void mlx4_en_get_ringparam(struct net_device *dev,
-				  struct ethtool_ringparam *param)
-{
-	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
-
-	memset(param, 0, sizeof(*param));
-	param->rx_max_pending = MLX4_EN_MAX_RX_SIZE;
-	param->tx_max_pending = MLX4_EN_MAX_TX_SIZE;
-	param->rx_pending = mdev->profile.prof[priv->port].rx_ring_size;
-	param->tx_pending = mdev->profile.prof[priv->port].tx_ring_size;
-}
-
-const struct ethtool_ops mlx4_en_ethtool_ops = {
-	.get_drvinfo = mlx4_en_get_drvinfo,
-	.get_settings = mlx4_en_get_settings,
-	.set_settings = mlx4_en_set_settings,
-#ifdef NETIF_F_TSO
-	.get_tso = mlx4_en_get_tso,
-	.set_tso = mlx4_en_set_tso,
-#endif
-	.get_sg = ethtool_op_get_sg,
-	.set_sg = ethtool_op_set_sg,
-	.get_link = ethtool_op_get_link,
-	.get_rx_csum = mlx4_en_get_rx_csum,
-	.set_rx_csum = mlx4_en_set_rx_csum,
-	.get_tx_csum = ethtool_op_get_tx_csum,
-	.set_tx_csum = ethtool_op_set_tx_ipv6_csum,
-	.get_strings = mlx4_en_get_strings,
-	.get_sset_count = mlx4_en_get_sset_count,
-	.get_ethtool_stats = mlx4_en_get_ethtool_stats,
-	.get_wol = mlx4_en_get_wol,
-	.get_msglevel = mlx4_en_get_msglevel,
-	.set_msglevel = mlx4_en_set_msglevel,
-	.get_coalesce = mlx4_en_get_coalesce,
-	.set_coalesce = mlx4_en_set_coalesce,
-	.get_pauseparam = mlx4_en_get_pauseparam,
-	.set_pauseparam = mlx4_en_set_pauseparam,
-	.get_ringparam = mlx4_en_get_ringparam,
-	.set_ringparam = mlx4_en_set_ringparam,
-	.get_flags = ethtool_op_get_flags,
-	.set_flags = ethtool_op_set_flags,
-};
-
-
-
-
-
-- 
1.6.0

[PATCH 4/8] mlx4_en: Work with part of the ports.

[Yevgeny Petrilin]

If the initialization of one of the ports failed,
there is no need to fail the other one as well.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_main.c |   19 +------------------
 1 files changed, 1 insertions(+), 18 deletions(-)

diff --git a/drivers/net/mlx4/en_main.c b/drivers/net/mlx4/en_main.c
index 23955d8..b510000 100644
--- a/drivers/net/mlx4/en_main.c
+++ b/drivers/net/mlx4/en_main.c
@@ -248,28 +248,11 @@ static void *mlx4_en_add(struct mlx4_dev *dev)
 	/* Create a netdev for each port */
 	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
 		mlx4_info(mdev, "Activating port:%d\n", i);
-		if (mlx4_en_init_netdev(mdev, i, &mdev->profile.prof[i])) {
+		if (mlx4_en_init_netdev(mdev, i, &mdev->profile.prof[i]))
 			mdev->pndev[i] = NULL;
-			goto err_free_netdev;
-		}
 	}
 	return mdev;
 
-
-err_free_netdev:
-	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
-		if (mdev->pndev[i])
-			mlx4_en_destroy_netdev(mdev->pndev[i]);
-	}
-
-	mutex_lock(&mdev->state_lock);
-	mdev->device_up = false;
-	mutex_unlock(&mdev->state_lock);
-	flush_workqueue(mdev->workqueue);
-
-	/* Stop event queue before we drop down to release shared SW state */
-	destroy_workqueue(mdev->workqueue);
-
 err_mr:
 	mlx4_mr_free(dev, &mdev->mr);
 err_uar:
-- 
1.6.0

[PATCH 5/8] mlx4_en: Coalescing target is equal for all mtu's

[Yevgeny Petrilin]

The interrupt moderation should not depend on number of incoming
bytes, but on number of incoming packets.
The previous scheme caused very high interrupts rate for small
messages when big MTU was configured.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_ethtool.c |    3 +--
 drivers/net/mlx4/en_netdev.c  |    2 +-
 drivers/net/mlx4/mlx4_en.h    |    4 ++--
 3 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/drivers/net/mlx4/en_ethtool.c b/drivers/net/mlx4/en_ethtool.c
index 91d8116..091f990 100644
--- a/drivers/net/mlx4/en_ethtool.c
+++ b/drivers/net/mlx4/en_ethtool.c
@@ -268,8 +268,7 @@ static int mlx4_en_set_coalesce(struct net_device *dev,
 
 	priv->rx_frames = (coal->rx_max_coalesced_frames ==
 			   MLX4_EN_AUTO_CONF) ?
-				MLX4_EN_RX_COAL_TARGET /
-				priv->dev->mtu + 1 :
+				MLX4_EN_RX_COAL_TARGET :
 				coal->rx_max_coalesced_frames;
 	priv->rx_usecs = (coal->rx_coalesce_usecs ==
 			  MLX4_EN_AUTO_CONF) ?
diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index fea65e7..16a634f 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -372,7 +372,7 @@ static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
 	 *   satisfy our coelsing target.
 	 * - moder_time is set to a fixed value.
 	 */
-	priv->rx_frames = MLX4_EN_RX_COAL_TARGET / priv->dev->mtu + 1;
+	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
 	en_dbg(INTR, priv, "Default coalesing params for mtu:%d - "
 			   "rx_frames:%d rx_usecs:%d\n",
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index 174ae86..fcbfcfc 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -143,8 +143,8 @@ enum {
 #define MLX4_EN_DEF_TX_RING_SIZE	1024
 #define MLX4_EN_DEF_RX_RING_SIZE  	1024
 
-/* Target number of bytes to coalesce with interrupt moderation */
-#define MLX4_EN_RX_COAL_TARGET	0x20000
+/* Target number of packets to coalesce with interrupt moderation */
+#define MLX4_EN_RX_COAL_TARGET	44
 #define MLX4_EN_RX_COAL_TIME	0x10
 
 #define MLX4_EN_TX_COAL_PKTS	5
-- 
1.6.0

[PATCH 6/8] mlx4_en: multiqueue support

[Yevgeny Petrilin]

By default the driver opens 8 TX queues (defined by MLX4_EN_NUM_TX_RINGS).
If the driver is configured to support Per Priority Flow Control, we open
8 additional TX rings.
dev->real_num_tx_queues is always set to be MLX4_EN_NUM_TX_RINGS.
The mlx4_en_select_queue() function uses standard hashing (skb_tx_hash)
in case that PPFC is not supported or the skb contain a vlan tag,
otherwise the queue is selected according to vlan priority.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_main.c   |    9 +----
 drivers/net/mlx4/en_netdev.c |    7 ++--
 drivers/net/mlx4/en_tx.c     |   74 ++++++++++-------------------------------
 drivers/net/mlx4/mlx4_en.h   |    9 +++--
 4 files changed, 28 insertions(+), 71 deletions(-)

diff --git a/drivers/net/mlx4/en_main.c b/drivers/net/mlx4/en_main.c
index b510000..9ed4a15 100644
--- a/drivers/net/mlx4/en_main.c
+++ b/drivers/net/mlx4/en_main.c
@@ -93,13 +93,8 @@ static int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
 		params->prof[i].tx_ppp = pfctx;
 		params->prof[i].tx_ring_size = MLX4_EN_DEF_TX_RING_SIZE;
 		params->prof[i].rx_ring_size = MLX4_EN_DEF_RX_RING_SIZE;
-	}
-	if (pfcrx || pfctx) {
-		params->prof[1].tx_ring_num = MLX4_EN_TX_RING_NUM;
-		params->prof[2].tx_ring_num = MLX4_EN_TX_RING_NUM;
-	} else {
-		params->prof[1].tx_ring_num = 1;
-		params->prof[2].tx_ring_num = 1;
+		params->prof[i].tx_ring_num = MLX4_EN_NUM_TX_RINGS +
+			(!!pfcrx) * MLX4_EN_NUM_PPP_RINGS;
 	}
 
 	return 0;
diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 16a634f..37e4d30 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -934,6 +934,7 @@ static const struct net_device_ops mlx4_netdev_ops = {
 	.ndo_open		= mlx4_en_open,
 	.ndo_stop		= mlx4_en_close,
 	.ndo_start_xmit		= mlx4_en_xmit,
+	.ndo_select_queue	= mlx4_en_select_queue,
 	.ndo_get_stats		= mlx4_en_get_stats,
 	.ndo_set_multicast_list	= mlx4_en_set_multicast,
 	.ndo_set_mac_address	= mlx4_en_set_mac,
@@ -956,7 +957,7 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	int i;
 	int err;
 
-	dev = alloc_etherdev(sizeof(struct mlx4_en_priv));
+	dev = alloc_etherdev_mq(sizeof(struct mlx4_en_priv), prof->tx_ring_num);
 	if (dev == NULL) {
 		mlx4_err(mdev, "Net device allocation failed\n");
 		return -ENOMEM;
@@ -1018,14 +1019,12 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	}
 	priv->allocated = 1;
 
-	/* Populate Tx priority mappings */
-	mlx4_en_set_prio_map(priv, priv->tx_prio_map, prof->tx_ring_num);
-
 	/*
 	 * Initialize netdev entry points
 	 */
 	dev->netdev_ops = &mlx4_netdev_ops;
 	dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
+	dev->real_num_tx_queues = MLX4_EN_NUM_TX_RINGS;
 
 	SET_ETHTOOL_OPS(dev, &mlx4_en_ethtool_ops);
 
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index e32e869..5dc7466 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -297,34 +297,6 @@ int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring)
 	return cnt;
 }
 
-void mlx4_en_set_prio_map(struct mlx4_en_priv *priv, u16 *prio_map, u32 ring_num)
-{
-	int block = 8 / ring_num;
-	int extra = 8 - (block * ring_num);
-	int num = 0;
-	u16 ring = 1;
-	int prio;
-
-	if (ring_num == 1) {
-		for (prio = 0; prio < 8; prio++)
-			prio_map[prio] = 0;
-		return;
-	}
-
-	for (prio = 0; prio < 8; prio++) {
-		if (extra && (num == block + 1)) {
-			ring++;
-			num = 0;
-			extra--;
-		} else if (!extra && (num == block)) {
-			ring++;
-			num = 0;
-		}
-		prio_map[prio] = ring;
-		en_dbg(DRV, priv, " prio:%d --> ring:%d\n", prio, ring);
-		num++;
-	}
-}
 
 static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
 {
@@ -386,18 +358,8 @@ static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
 	if (unlikely(ring->blocked)) {
 		if ((u32) (ring->prod - ring->cons) <=
 		     ring->size - HEADROOM - MAX_DESC_TXBBS) {
-
-			/* TODO: support multiqueue netdevs. Currently, we block
-			 * when *any* ring is full. Note that:
-			 * - 2 Tx rings can unblock at the same time and call
-			 *   netif_wake_queue(), which is OK since this
-			 *   operation is idempotent.
-			 * - We might wake the queue just after another ring
-			 *   stopped it. This is no big deal because the next
-			 *   transmission on that ring would stop the queue.
-			 */
 			ring->blocked = 0;
-			netif_wake_queue(dev);
+			netif_tx_wake_queue(netdev_get_tx_queue(dev, cq->ring));
 			priv->port_stats.wake_queue++;
 		}
 	}
@@ -616,21 +578,20 @@ static void build_inline_wqe(struct mlx4_en_tx_desc *tx_desc, struct sk_buff *sk
 	tx_desc->ctrl.fence_size = (real_size / 16) & 0x3f;
 }
 
-static int get_vlan_info(struct mlx4_en_priv *priv, struct sk_buff *skb,
-			 u16 *vlan_tag)
+u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb)
 {
-	int tx_ind;
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	u16 vlan_tag = 0;
 
-	/* Obtain VLAN information if present */
-	if (priv->vlgrp && vlan_tx_tag_present(skb)) {
-		*vlan_tag = vlan_tx_tag_get(skb);
-		/* Set the Tx ring to use according to vlan priority */
-		tx_ind = priv->tx_prio_map[*vlan_tag >> 13];
-	} else {
-		*vlan_tag = 0;
-		tx_ind = 0;
+	/* If we support per priority flow control and the packet contains
+	 * a vlan tag, send the packet to the TX ring assigned to that priority
+	 */
+	if (priv->prof->rx_ppp && priv->vlgrp && vlan_tx_tag_present(skb)) {
+		vlan_tag = vlan_tx_tag_get(skb);
+		return MLX4_EN_NUM_TX_RINGS + (vlan_tag >> 13);
 	}
-	return tx_ind;
+
+	return skb_tx_hash(dev, skb);
 }
 
 int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
@@ -650,7 +611,7 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
 	dma_addr_t dma;
 	u32 index;
 	__be32 op_own;
-	u16 vlan_tag;
+	u16 vlan_tag = 0;
 	int i;
 	int lso_header_size;
 	void *fragptr;
@@ -673,15 +634,16 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
 		return NETDEV_TX_OK;
 	}
 
-	tx_ind = get_vlan_info(priv, skb, &vlan_tag);
+	tx_ind = skb->queue_mapping;
 	ring = &priv->tx_ring[tx_ind];
+	if (priv->vlgrp && vlan_tx_tag_present(skb))
+		vlan_tag = vlan_tx_tag_get(skb);
 
 	/* Check available TXBBs And 2K spare for prefetch */
 	if (unlikely(((int)(ring->prod - ring->cons)) >
 		     ring->size - HEADROOM - MAX_DESC_TXBBS)) {
-		/* every full Tx ring stops queue.
-		 * TODO: implement multi-queue support (per-queue stop) */
-		netif_stop_queue(dev);
+		/* every full Tx ring stops queue */
+		netif_tx_stop_queue(netdev_get_tx_queue(dev, tx_ind));
 		ring->blocked = 1;
 		priv->port_stats.queue_stopped++;
 
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index fcbfcfc..4de8db0 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -139,8 +139,10 @@ enum {
 #define MLX4_EN_MIN_RX_SIZE	(MLX4_EN_ALLOC_SIZE / SMP_CACHE_BYTES)
 #define MLX4_EN_MIN_TX_SIZE	(4096 / TXBB_SIZE)
 
-#define MLX4_EN_TX_RING_NUM		9
-#define MLX4_EN_DEF_TX_RING_SIZE	1024
+#define MLX4_EN_SMALL_PKT_SIZE		64
+#define MLX4_EN_NUM_TX_RINGS		8
+#define MLX4_EN_NUM_PPP_RINGS		8
+#define MLX4_EN_DEF_TX_RING_SIZE	512
 #define MLX4_EN_DEF_RX_RING_SIZE  	1024
 
 /* Target number of packets to coalesce with interrupt moderation */
@@ -478,7 +480,6 @@ struct mlx4_en_priv {
 	int base_qpn;
 
 	struct mlx4_en_rss_map rss_map;
-	u16 tx_prio_map[8];
 	u32 flags;
 #define MLX4_EN_FLAG_PROMISC	0x1
 	u32 tx_ring_num;
@@ -526,6 +527,7 @@ int mlx4_en_arm_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
 
 void mlx4_en_poll_tx_cq(unsigned long data);
 void mlx4_en_tx_irq(struct mlx4_cq *mcq);
+u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb);
 int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev);
 
 int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv, struct mlx4_en_tx_ring *ring,
@@ -560,7 +562,6 @@ void mlx4_en_calc_rx_buf(struct net_device *dev);
 void mlx4_en_set_default_rss_map(struct mlx4_en_priv *priv,
 				 struct mlx4_en_rss_map *rss_map,
 				 int num_entries, int num_rings);
-void mlx4_en_set_prio_map(struct mlx4_en_priv *priv, u16 *prio_map, u32 ring_num);
 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);
-- 
1.6.0

[PATCH 7/8] mlx4_en: Added vlan_features support

[Yevgeny Petrilin]

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_netdev.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 37e4d30..0a7e78a 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -1038,7 +1038,9 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	 * Set driver features
 	 */
 	dev->features |= NETIF_F_SG;
+	dev->vlan_features |= NETIF_F_SG;
 	dev->features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+	dev->vlan_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
 	dev->features |= NETIF_F_HIGHDMA;
 	dev->features |= NETIF_F_HW_VLAN_TX |
 			 NETIF_F_HW_VLAN_RX |
@@ -1048,6 +1050,8 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	if (mdev->LSO_support) {
 		dev->features |= NETIF_F_TSO;
 		dev->features |= NETIF_F_TSO6;
+		dev->vlan_features |= NETIF_F_TSO;
+		dev->vlan_features |= NETIF_F_TSO6;
 	}
 
 	mdev->pndev[port] = dev;
-- 
1.6.0

[PATCH 8/8] mlx4_en: Updated driver version

[Yevgeny Petrilin]

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/mlx4_en.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index 4de8db0..f5909c7 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -49,8 +49,8 @@
 #include "en_port.h"
 
 #define DRV_NAME	"mlx4_en"
-#define DRV_VERSION	"1.4.0"
-#define DRV_RELDATE	"Sep 2008"
+#define DRV_VERSION	"1.4.1.1
+#define DRV_RELDATE	"June 2009"
 
 
 #define MLX4_EN_MSG_LEVEL	(NETIF_MSG_LINK | NETIF_MSG_IFDOWN)
-- 
1.6.0

Re: [PATCH 8/8] mlx4_en: Updated driver version

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Tue, 02 Jun 2009 12:29:42 +0300

> +#define DRV_VERSION	"1.4.1.1

You checked in this commit to your tree WITHOUT EVEN BUILD TESTING IT!

drivers/net/mlx4/en_main.c:48: error: missing terminating " character
drivers/net/mlx4/en_main.c:48: error: expected ‘,’ or ‘;’ before string constant
drivers/net/mlx4/en_main.c:52: error: missing terminating " character
drivers/net/mlx4/en_main.c:52: error: expected ‘,’ or ‘;’ before string constant

I fixed this up in my tree but I'm really pissed that you submitted
patches to me that YOU DID NOT EVEN BUILD TEST.

[net-2.6 PATCH V2] mlx4_en: Fix a kernel panic when waking tx queue

[Yevgeny Petrilin]

When the transmit queue gets full we enable interrupts for TX completions
There was a race that we handled the TX queue both from the interrupt context
and from the transmit function. Using "spin_trylock_irq()" ensures this
doesn't happen.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_tx.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index ac6fc49..e5c98a9 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -426,7 +426,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)

 	INC_PERF_COUNTER(priv->pstats.tx_poll);

-	if (!spin_trylock(&ring->comp_lock)) {
+	if (!spin_trylock_irq(&ring->comp_lock)) {
 		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
 		return;
 	}
@@ -439,7 +439,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)
 	if (inflight && priv->port_up)
 		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);

-	spin_unlock(&ring->comp_lock);
+	spin_unlock_irq(&ring->comp_lock);
 }

 static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
@@ -482,9 +482,9 @@ static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)

 	/* Poll the CQ every mlx4_en_TX_MODER_POLL packets */
 	if ((++ring->poll_cnt & (MLX4_EN_TX_POLL_MODER - 1)) == 0)
-		if (spin_trylock(&ring->comp_lock)) {
+		if (spin_trylock_irq(&ring->comp_lock)) {
 			mlx4_en_process_tx_cq(priv->dev, cq);
-			spin_unlock(&ring->comp_lock);
+			spin_unlock_irq(&ring->comp_lock);
 		}
 }

-- 
1.6.0

Re: [net-2.6 PATCH V2] mlx4_en: Fix a kernel panic when waking tx queue

[Eric Dumazet]

Yevgeny Petrilin a écrit :
> When the transmit queue gets full we enable interrupts for TX completions
> There was a race that we handled the TX queue both from the interrupt context
> and from the transmit function. Using "spin_trylock_irq()" ensures this
> doesn't happen.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
> ---
>  drivers/net/mlx4/en_tx.c |    8 ++++----
>  1 files changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
> index ac6fc49..e5c98a9 100644
> --- a/drivers/net/mlx4/en_tx.c
> +++ b/drivers/net/mlx4/en_tx.c
> @@ -426,7 +426,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)
> 
>  	INC_PERF_COUNTER(priv->pstats.tx_poll);
> 
> -	if (!spin_trylock(&ring->comp_lock)) {
> +	if (!spin_trylock_irq(&ring->comp_lock)) {
>  		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
>  		return;
>  	}
> @@ -439,7 +439,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)
>  	if (inflight && priv->port_up)
>  		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
> 
> -	spin_unlock(&ring->comp_lock);
> +	spin_unlock_irq(&ring->comp_lock);
>  }
> 
>  static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
> @@ -482,9 +482,9 @@ static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)
> 
>  	/* Poll the CQ every mlx4_en_TX_MODER_POLL packets */
>  	if ((++ring->poll_cnt & (MLX4_EN_TX_POLL_MODER - 1)) == 0)
> -		if (spin_trylock(&ring->comp_lock)) {
> +		if (spin_trylock_irq(&ring->comp_lock)) {
>  			mlx4_en_process_tx_cq(priv->dev, cq);
> -			spin_unlock(&ring->comp_lock);
> +			spin_unlock_irq(&ring->comp_lock);
>  		}
>  }
> 

Just curious.

Blocking hard IRQ while doing TX completion is quite nasty,
as freeing one hundred of skb take long.

Could you try something in process context instead ?

Is this timer driven tx completion thing really good for performance ?


static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)
{
        struct mlx4_en_cq *cq = &priv->tx_cq[tx_ind];
        struct mlx4_en_tx_ring *ring = &priv->tx_ring[tx_ind];

        /* If we don't have a pending timer, set one up to catch our recent
           post in case the interface becomes idle */
        if (!timer_pending(&cq->timer))
                mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);


        /* Poll the CQ every mlx4_en_TX_MODER_POLL packets */
        if ((++ring->poll_cnt & (MLX4_EN_TX_POLL_MODER - 1)) == 0)
                if (spin_trylock(&ring->comp_lock)) {
                        mlx4_en_process_tx_cq(priv->dev, cq);
                        spin_unlock(&ring->comp_lock);
                }
}


One interesting thing is calling mlx4_en_process_tx_cq() once every 16 packets,
this means the producer cpu is also the cpu doing the TX completion, I like this.

Re: [net-2.6 PATCH V2] mlx4_en: Fix a kernel panic when waking tx queue

[Yevgeny Petrilin]

Hi Eric,

Eric Dumazet wrote:
>
> Just curious.
>
> Blocking hard IRQ while doing TX completion is quite nasty,
> as freeing one hundred of skb take long.
>
> Could you try something in process context instead ?

I am planning some changes for handling TX completions, but I think it is too
late to
add them to 2.6.30.
In the code that will be submitted to 2.6.31 we will not block irqs.

>
> Is this timer driven tx completion thing really good for performance ?
>
When we are CPU bounded on the transmit side, the driver shown better
performance with
the timer driven completions rather then with interrupts.

Re: [net-2.6 PATCH V2] mlx4_en: Fix a kernel panic when waking tx queue

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Tue, 26 May 2009 09:57:21 +0300

> When the transmit queue gets full we enable interrupts for TX completions
> There was a race that we handled the TX queue both from the interrupt context
> and from the transmit function. Using "spin_trylock_irq()" ensures this
> doesn't happen.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

I'll apply this, thanks.

[PATCH 1/8] mlx4_en: Giving interface name in debug messages

[Yevgeny Petrilin]

For each debug message, the message will show interface name in case
that the net device was registered, and PCI bus ID with port number
if we were not registered yet. Messages that are not port/netdev specific
stayed in the old format

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
---
 drivers/net/mlx4/en_netdev.c |  162 ++++++++++++++++++++----------------------
 drivers/net/mlx4/en_params.c |    6 +-
 drivers/net/mlx4/en_rx.c     |   78 ++++++++++-----------
 drivers/net/mlx4/en_tx.c     |   39 +++++-----
 drivers/net/mlx4/mlx4_en.h   |   30 ++++++--
 5 files changed, 159 insertions(+), 156 deletions(-)

diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 0cd185a..fea65e7 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -51,14 +51,14 @@ static void mlx4_en_vlan_rx_register(struct net_device *dev, struct vlan_group *
 	struct mlx4_en_dev *mdev = priv->mdev;
 	int err;

-	mlx4_dbg(HW, priv, "Registering VLAN group:%p\n", grp);
+	en_dbg(HW, priv, "Registering VLAN group:%p\n", grp);
 	priv->vlgrp = grp;

 	mutex_lock(&mdev->state_lock);
 	if (mdev->device_up && priv->port_up) {
 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv->port, grp);
 		if (err)
-			mlx4_err(mdev, "Failed configuring VLAN filter\n");
+			en_err(priv, "Failed configuring VLAN filter\n");
 	}
 	mutex_unlock(&mdev->state_lock);
 }
@@ -72,15 +72,15 @@ static void mlx4_en_vlan_rx_add_vid(struct net_device *dev, unsigned short vid)
 	if (!priv->vlgrp)
 		return;

-	mlx4_dbg(HW, priv, "adding VLAN:%d (vlgrp entry:%p)\n",
-		 vid, vlan_group_get_device(priv->vlgrp, vid));
+	en_dbg(HW, priv, "adding VLAN:%d (vlgrp entry:%p)\n",
+	       vid, vlan_group_get_device(priv->vlgrp, vid));

 	/* Add VID to port VLAN filter */
 	mutex_lock(&mdev->state_lock);
 	if (mdev->device_up && priv->port_up) {
 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv->port, priv->vlgrp);
 		if (err)
-			mlx4_err(mdev, "Failed configuring VLAN filter\n");
+			en_err(priv, "Failed configuring VLAN filter\n");
 	}
 	mutex_unlock(&mdev->state_lock);
 }
@@ -94,9 +94,8 @@ static void mlx4_en_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
 	if (!priv->vlgrp)
 		return;

-	mlx4_dbg(HW, priv, "Killing VID:%d (vlgrp:%p vlgrp "
-		 "entry:%p)\n", vid, priv->vlgrp,
-		 vlan_group_get_device(priv->vlgrp, vid));
+	en_dbg(HW, priv, "Killing VID:%d (vlgrp:%p vlgrp entry:%p)\n",
+	       vid, priv->vlgrp, vlan_group_get_device(priv->vlgrp, vid));
 	vlan_group_set_device(priv->vlgrp, vid, NULL);

 	/* Remove VID from port VLAN filter */
@@ -104,7 +103,7 @@ static void mlx4_en_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
 	if (mdev->device_up && priv->port_up) {
 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv->port, priv->vlgrp);
 		if (err)
-			mlx4_err(mdev, "Failed configuring VLAN filter\n");
+			en_err(priv, "Failed configuring VLAN filter\n");
 	}
 	mutex_unlock(&mdev->state_lock);
 }
@@ -150,9 +149,10 @@ static void mlx4_en_do_set_mac(struct work_struct *work)
 		err = mlx4_register_mac(mdev->dev, priv->port,
 					priv->mac, &priv->mac_index);
 		if (err)
-			mlx4_err(mdev, "Failed changing HW MAC address\n");
+			en_err(priv, "Failed changing HW MAC address\n");
 	} else
-		mlx4_dbg(HW, priv, "Port is down, exiting...\n");
+		en_dbg(HW, priv, "Port is down while "
+				 "registering mac, exiting...\n");

 	mutex_unlock(&mdev->state_lock);
 }
@@ -174,7 +174,6 @@ static void mlx4_en_clear_list(struct net_device *dev)
 static void mlx4_en_cache_mclist(struct net_device *dev)
 {
 	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
 	struct dev_mc_list *mclist;
 	struct dev_mc_list *tmp;
 	struct dev_mc_list *plist = NULL;
@@ -182,7 +181,7 @@ static void mlx4_en_cache_mclist(struct net_device *dev)
 	for (mclist = dev->mc_list; mclist; mclist = mclist->next) {
 		tmp = kmalloc(sizeof(struct dev_mc_list), GFP_ATOMIC);
 		if (!tmp) {
-			mlx4_err(mdev, "failed to allocate multicast list\n");
+			en_err(priv, "failed to allocate multicast list\n");
 			mlx4_en_clear_list(dev);
 			return;
 		}
@@ -219,13 +218,13 @@ static void mlx4_en_do_set_multicast(struct work_struct *work)

 	mutex_lock(&mdev->state_lock);
 	if (!mdev->device_up) {
-		mlx4_dbg(HW, priv, "Card is not up, ignoring "
-				   "multicast change.\n");
+		en_dbg(HW, priv, "Card is not up, "
+				 "ignoring multicast change.\n");
 		goto out;
 	}
 	if (!priv->port_up) {
-		mlx4_dbg(HW, priv, "Port is down, ignoring "
-				   "multicast change.\n");
+		en_dbg(HW, priv, "Port is down, "
+				 "ignoring  multicast change.\n");
 		goto out;
 	}

@@ -236,29 +235,27 @@ static void mlx4_en_do_set_multicast(struct work_struct *work)
 	if (dev->flags & IFF_PROMISC) {
 		if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
 			if (netif_msg_rx_status(priv))
-				mlx4_warn(mdev, "Port:%d entering promiscuous mode\n",
-					  priv->port);
+				en_warn(priv, "Entering promiscuous mode\n");
 			priv->flags |= MLX4_EN_FLAG_PROMISC;

 			/* Enable promiscouos mode */
 			err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port,
 						     priv->base_qpn, 1);
 			if (err)
-				mlx4_err(mdev, "Failed enabling "
-					 "promiscous mode\n");
+				en_err(priv, "Failed enabling "
+					     "promiscous mode\n");

 			/* Disable port multicast filter (unconditionally) */
 			err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 						  0, MLX4_MCAST_DISABLE);
 			if (err)
-				mlx4_err(mdev, "Failed disabling "
-					 "multicast filter\n");
+				en_err(priv, "Failed disabling "
+					     "multicast filter\n");

 			/* Disable port VLAN filter */
 			err = mlx4_SET_VLAN_FLTR(mdev->dev, priv->port, NULL);
 			if (err)
-				mlx4_err(mdev, "Failed disabling "
-					 "VLAN filter\n");
+				en_err(priv, "Failed disabling VLAN filter\n");
 		}
 		goto out;
 	}
@@ -269,20 +266,19 @@ static void mlx4_en_do_set_multicast(struct work_struct *work)

 	if (priv->flags & MLX4_EN_FLAG_PROMISC) {
 		if (netif_msg_rx_status(priv))
-			mlx4_warn(mdev, "Port:%d leaving promiscuous mode\n",
-				  priv->port);
+			en_warn(priv, "Leaving promiscuous mode\n");
 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;

 		/* Disable promiscouos mode */
 		err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port,
 					     priv->base_qpn, 0);
 		if (err)
-			mlx4_err(mdev, "Failed disabling promiscous mode\n");
+			en_err(priv, "Failed disabling promiscous mode\n");

 		/* Enable port VLAN filter */
 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv->port, priv->vlgrp);
 		if (err)
-			mlx4_err(mdev, "Failed enabling VLAN filter\n");
+			en_err(priv, "Failed enabling VLAN filter\n");
 	}

 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
@@ -290,12 +286,12 @@ static void mlx4_en_do_set_multicast(struct work_struct *work)
 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 					  0, MLX4_MCAST_DISABLE);
 		if (err)
-			mlx4_err(mdev, "Failed disabling multicast filter\n");
+			en_err(priv, "Failed disabling multicast filter\n");
 	} else {
 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 					  0, MLX4_MCAST_DISABLE);
 		if (err)
-			mlx4_err(mdev, "Failed disabling multicast filter\n");
+			en_err(priv, "Failed disabling multicast filter\n");

 		/* Flush mcast filter and init it with broadcast address */
 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
@@ -314,7 +310,7 @@ static void mlx4_en_do_set_multicast(struct work_struct *work)
 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 					  0, MLX4_MCAST_ENABLE);
 		if (err)
-			mlx4_err(mdev, "Failed enabling multicast filter\n");
+			en_err(priv, "Failed enabling multicast filter\n");

 		mlx4_en_clear_list(dev);
 	}
@@ -346,10 +342,10 @@ static void mlx4_en_tx_timeout(struct net_device *dev)
 	struct mlx4_en_dev *mdev = priv->mdev;

 	if (netif_msg_timer(priv))
-		mlx4_warn(mdev, "Tx timeout called on port:%d\n", priv->port);
+		en_warn(priv, "Tx timeout called on port:%d\n", priv->port);

 	priv->port_stats.tx_timeout++;
-	mlx4_dbg(DRV, priv, "Scheduling watchdog\n");
+	en_dbg(DRV, priv, "Scheduling watchdog\n");
 	queue_work(mdev->workqueue, &priv->watchdog_task);
 }

@@ -378,8 +374,8 @@ static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
 	 */
 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET / priv->dev->mtu + 1;
 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
-	mlx4_dbg(INTR, priv, "Default coalesing params for mtu:%d - "
-			     "rx_frames:%d rx_usecs:%d\n",
+	en_dbg(INTR, priv, "Default coalesing params for mtu:%d - "
+			   "rx_frames:%d rx_usecs:%d\n",
 		 priv->dev->mtu, priv->rx_frames, priv->rx_usecs);

 	/* Setup cq moderation params */
@@ -412,7 +408,6 @@ static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
 {
 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
-	struct mlx4_en_dev *mdev = priv->mdev;
 	struct mlx4_en_cq *cq;
 	unsigned long packets;
 	unsigned long rate;
@@ -472,11 +467,11 @@ static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
 		moder_time = priv->rx_usecs;
 	}

-	mlx4_dbg(INTR, priv, "tx rate:%lu rx_rate:%lu\n",
-		 tx_pkt_diff * HZ / period, rx_pkt_diff * HZ / period);
+	en_dbg(INTR, priv, "tx rate:%lu rx_rate:%lu\n",
+	       tx_pkt_diff * HZ / period, rx_pkt_diff * HZ / period);

-	mlx4_dbg(INTR, priv, "Rx moder_time changed from:%d to %d period:%lu "
-		 "[jiff] packets:%lu avg_pkt_size:%lu rate:%lu [p/s])\n",
+	en_dbg(INTR, priv, "Rx moder_time changed from:%d to %d period:%lu "
+	       "[jiff] packets:%lu avg_pkt_size:%lu rate:%lu [p/s])\n",
 		 priv->last_moder_time, moder_time, period, packets,
 		 avg_pkt_size, rate);

@@ -487,8 +482,7 @@ static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
 			cq->moder_time = moder_time;
 			err = mlx4_en_set_cq_moder(priv, cq);
 			if (err) {
-				mlx4_err(mdev, "Failed modifying moderation for cq:%d "
-					 "on port:%d\n", i, priv->port);
+				en_err(priv, "Failed modifying moderation for cq:%d\n", i);
 				break;
 			}
 		}
@@ -511,8 +505,7 @@ static void mlx4_en_do_get_stats(struct work_struct *work)

 	err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
 	if (err)
-		mlx4_dbg(HW, priv, "Could not update stats for "
-				   "port:%d\n", priv->port);
+		en_dbg(HW, priv, "Could not update stats \n");

 	mutex_lock(&mdev->state_lock);
 	if (mdev->device_up) {
@@ -536,12 +529,10 @@ static void mlx4_en_linkstate(struct work_struct *work)
 	 * report to system log */
 	if (priv->last_link_state != linkstate) {
 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
-			if (netif_msg_link(priv))
-				mlx4_info(mdev, "Port %d - link down\n", priv->port);
+			en_dbg(LINK, priv, "Link Down\n");
 			netif_carrier_off(priv->dev);
 		} else {
-			if (netif_msg_link(priv))
-				mlx4_info(mdev, "Port %d - link up\n", priv->port);
+			en_dbg(LINK, priv, "Link Up\n");
 			netif_carrier_on(priv->dev);
 		}
 	}
@@ -563,19 +554,19 @@ int mlx4_en_start_port(struct net_device *dev)
 	int j;

 	if (priv->port_up) {
-		mlx4_dbg(DRV, priv, "start port called while port already up\n");
+		en_dbg(DRV, priv, "start port called while port already up\n");
 		return 0;
 	}

 	/* Calculate Rx buf size */
 	dev->mtu = min(dev->mtu, priv->max_mtu);
 	mlx4_en_calc_rx_buf(dev);
-	mlx4_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
+	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);

 	/* Configure rx cq's and rings */
 	err = mlx4_en_activate_rx_rings(priv);
 	if (err) {
-		mlx4_err(mdev, "Failed to activate RX rings\n");
+		en_err(priv, "Failed to activate RX rings\n");
 		return err;
 	}
 	for (i = 0; i < priv->rx_ring_num; i++) {
@@ -583,14 +574,14 @@ int mlx4_en_start_port(struct net_device *dev)

 		err = mlx4_en_activate_cq(priv, cq);
 		if (err) {
-			mlx4_err(mdev, "Failed activating Rx CQ\n");
+			en_err(priv, "Failed activating Rx CQ\n");
 			goto cq_err;
 		}
 		for (j = 0; j < cq->size; j++)
 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
 		err = mlx4_en_set_cq_moder(priv, cq);
 		if (err) {
-			mlx4_err(mdev, "Failed setting cq moderation parameters");
+			en_err(priv, "Failed setting cq moderation parameters");
 			mlx4_en_deactivate_cq(priv, cq);
 			goto cq_err;
 		}
@@ -601,7 +592,7 @@ int mlx4_en_start_port(struct net_device *dev)

 	err = mlx4_en_config_rss_steer(priv);
 	if (err) {
-		mlx4_err(mdev, "Failed configuring rss steering\n");
+		en_err(priv, "Failed configuring rss steering\n");
 		goto cq_err;
 	}

@@ -611,16 +602,16 @@ int mlx4_en_start_port(struct net_device *dev)
 		cq = &priv->tx_cq[i];
 		err = mlx4_en_activate_cq(priv, cq);
 		if (err) {
-			mlx4_err(mdev, "Failed allocating Tx CQ\n");
+			en_err(priv, "Failed allocating Tx CQ\n");
 			goto tx_err;
 		}
 		err = mlx4_en_set_cq_moder(priv, cq);
 		if (err) {
-			mlx4_err(mdev, "Failed setting cq moderation parameters");
+			en_err(priv, "Failed setting cq moderation parameters");
 			mlx4_en_deactivate_cq(priv, cq);
 			goto tx_err;
 		}
-		mlx4_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
+		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
 		cq->buf->wqe_index = cpu_to_be16(0xffff);

 		/* Configure ring */
@@ -628,7 +619,7 @@ int mlx4_en_start_port(struct net_device *dev)
 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
 					       priv->rx_ring[0].srq.srqn);
 		if (err) {
-			mlx4_err(mdev, "Failed allocating Tx ring\n");
+			en_err(priv, "Failed allocating Tx ring\n");
 			mlx4_en_deactivate_cq(priv, cq);
 			goto tx_err;
 		}
@@ -646,30 +637,30 @@ int mlx4_en_start_port(struct net_device *dev)
 				    priv->prof->rx_pause,
 				    priv->prof->rx_ppp);
 	if (err) {
-		mlx4_err(mdev, "Failed setting port general configurations"
-			       " for port %d, with error %d\n", priv->port, err);
+		en_err(priv, "Failed setting port general configurations "
+			     "for port %d, with error %d\n", priv->port, err);
 		goto tx_err;
 	}
 	/* Set default qp number */
 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
 	if (err) {
-		mlx4_err(mdev, "Failed setting default qp numbers\n");
+		en_err(priv, "Failed setting default qp numbers\n");
 		goto tx_err;
 	}
 	/* Set port mac number */
-	mlx4_dbg(DRV, priv, "Setting mac for port %d\n", priv->port);
+	en_dbg(DRV, priv, "Setting mac for port %d\n", priv->port);
 	err = mlx4_register_mac(mdev->dev, priv->port,
 				priv->mac, &priv->mac_index);
 	if (err) {
-		mlx4_err(mdev, "Failed setting port mac\n");
+		en_err(priv, "Failed setting port mac\n");
 		goto tx_err;
 	}

 	/* Init port */
-	mlx4_dbg(HW, priv, "Initializing port\n");
+	en_dbg(HW, priv, "Initializing port\n");
 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
 	if (err) {
-		mlx4_err(mdev, "Failed Initializing port\n");
+		en_err(priv, "Failed Initializing port\n");
 		goto mac_err;
 	}

@@ -706,8 +697,7 @@ void mlx4_en_stop_port(struct net_device *dev)
 	int i;

 	if (!priv->port_up) {
-		mlx4_dbg(DRV, priv, "stop port (%d) called while port already down\n",
-			 priv->port);
+		en_dbg(DRV, priv, "stop port called while port already down\n");
 		return;
 	}
 	netif_stop_queue(dev);
@@ -752,13 +742,13 @@ static void mlx4_en_restart(struct work_struct *work)
 	struct mlx4_en_dev *mdev = priv->mdev;
 	struct net_device *dev = priv->dev;

-	mlx4_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
+	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);

 	mutex_lock(&mdev->state_lock);
 	if (priv->port_up) {
 		mlx4_en_stop_port(dev);
 		if (mlx4_en_start_port(dev))
-			mlx4_err(mdev, "Failed restarting port %d\n", priv->port);
+			en_err(priv, "Failed restarting port %d\n", priv->port);
 	}
 	mutex_unlock(&mdev->state_lock);
 }
@@ -774,14 +764,14 @@ static int mlx4_en_open(struct net_device *dev)
 	mutex_lock(&mdev->state_lock);

 	if (!mdev->device_up) {
-		mlx4_err(mdev, "Cannot open - device down/disabled\n");
+		en_err(priv, "Cannot open - device down/disabled\n");
 		err = -EBUSY;
 		goto out;
 	}

 	/* Reset HW statistics and performance counters */
 	if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
-		mlx4_dbg(HW, priv, "Failed dumping statistics\n");
+		en_dbg(HW, priv, "Failed dumping statistics\n");

 	memset(&priv->stats, 0, sizeof(priv->stats));
 	memset(&priv->pstats, 0, sizeof(priv->pstats));
@@ -798,7 +788,7 @@ static int mlx4_en_open(struct net_device *dev)
 	mlx4_en_set_default_moderation(priv);
 	err = mlx4_en_start_port(dev);
 	if (err)
-		mlx4_err(mdev, "Failed starting port:%d\n", priv->port);
+		en_err(priv, "Failed starting port:%d\n", priv->port);

 out:
 	mutex_unlock(&mdev->state_lock);
@@ -811,8 +801,7 @@ static int mlx4_en_close(struct net_device *dev)
 	struct mlx4_en_priv *priv = netdev_priv(dev);
 	struct mlx4_en_dev *mdev = priv->mdev;

-	if (netif_msg_ifdown(priv))
-		mlx4_info(mdev, "Close called for port:%d\n", priv->port);
+	en_dbg(IFDOWN, priv, "Close port called\n");

 	mutex_lock(&mdev->state_lock);

@@ -844,7 +833,6 @@ void mlx4_en_free_resources(struct mlx4_en_priv *priv)

 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
 {
-	struct mlx4_en_dev *mdev = priv->mdev;
 	struct mlx4_en_port_profile *prof = priv->prof;
 	int i;

@@ -873,7 +861,7 @@ int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
 	return 0;

 err:
-	mlx4_err(mdev, "Failed to allocate NIC resources\n");
+	en_err(priv, "Failed to allocate NIC resources\n");
 	return -ENOMEM;
 }

@@ -883,7 +871,7 @@ void mlx4_en_destroy_netdev(struct net_device *dev)
 	struct mlx4_en_priv *priv = netdev_priv(dev);
 	struct mlx4_en_dev *mdev = priv->mdev;

-	mlx4_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
+	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);

 	/* Unregister device - this will close the port if it was up */
 	if (priv->registered)
@@ -912,11 +900,11 @@ static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
 	struct mlx4_en_dev *mdev = priv->mdev;
 	int err = 0;

-	mlx4_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
+	en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
 		 dev->mtu, new_mtu);

 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
-		mlx4_err(mdev, "Bad MTU size:%d.\n", new_mtu);
+		en_err(priv, "Bad MTU size:%d.\n", new_mtu);
 		return -EPERM;
 	}
 	dev->mtu = new_mtu;
@@ -926,13 +914,13 @@ static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
 		if (!mdev->device_up) {
 			/* NIC is probably restarting - let watchdog task reset
 			 * the port */
-			mlx4_dbg(DRV, priv, "Change MTU called with card down!?\n");
+			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
 		} else {
 			mlx4_en_stop_port(dev);
 			mlx4_en_set_default_moderation(priv);
 			err = mlx4_en_start_port(dev);
 			if (err) {
-				mlx4_err(mdev, "Failed restarting port:%d\n",
+				en_err(priv, "Failed restarting port:%d\n",
 					 priv->port);
 				queue_work(mdev->workqueue, &priv->watchdog_task);
 			}
@@ -1006,7 +994,7 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
 	priv->mac = mdev->dev->caps.def_mac[priv->port];
 	if (ILLEGAL_MAC(priv->mac)) {
-		mlx4_err(mdev, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
+		en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
 			 priv->port, priv->mac);
 		err = -EINVAL;
 		goto out;
@@ -1025,7 +1013,7 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
 	if (err) {
-		mlx4_err(mdev, "Failed to allocate page for rx qps\n");
+		en_err(priv, "Failed to allocate page for rx qps\n");
 		goto out;
 	}
 	priv->allocated = 1;
@@ -1068,9 +1056,13 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	netif_carrier_off(dev);
 	err = register_netdev(dev);
 	if (err) {
-		mlx4_err(mdev, "Netdev registration failed\n");
+		en_err(priv, "Netdev registration failed for port %d\n", port);
 		goto out;
 	}
+
+	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
+	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
+
 	priv->registered = 1;
 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
 	return 0;
diff --git a/drivers/net/mlx4/en_params.c b/drivers/net/mlx4/en_params.c
index c1bd040..3290fec 100644
--- a/drivers/net/mlx4/en_params.c
+++ b/drivers/net/mlx4/en_params.c
@@ -371,7 +371,7 @@ static int mlx4_en_set_pauseparam(struct net_device *dev,
 				    priv->prof->rx_pause,
 				    priv->prof->rx_ppp);
 	if (err)
-		mlx4_err(mdev, "Failed setting pause params to\n");
+		en_err(priv, "Failed setting pause params\n");

 	return err;
 }
@@ -421,13 +421,13 @@ static int mlx4_en_set_ringparam(struct net_device *dev,

 	err = mlx4_en_alloc_resources(priv);
 	if (err) {
-		mlx4_err(mdev, "Failed reallocating port resources\n");
+		en_err(priv, "Failed reallocating port resources\n");
 		goto out;
 	}
 	if (port_up) {
 		err = mlx4_en_start_port(dev);
 		if (err)
-			mlx4_err(mdev, "Failed starting port\n");
+			en_err(priv, "Failed starting port\n");
 	}

 out:
diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 6bfab6e..5a14899 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -114,8 +114,8 @@ static int mlx4_en_init_allocator(struct mlx4_en_priv *priv,
 			goto out;

 		page_alloc->offset = priv->frag_info[i].frag_align;
-		mlx4_dbg(DRV, priv, "Initialized allocator:%d with page:%p\n",
-			 i, page_alloc->page);
+		en_dbg(DRV, priv, "Initialized allocator:%d with page:%p\n",
+		       i, page_alloc->page);
 	}
 	return 0;

@@ -136,8 +136,8 @@ static void mlx4_en_destroy_allocator(struct mlx4_en_priv *priv,

 	for (i = 0; i < priv->num_frags; i++) {
 		page_alloc = &ring->page_alloc[i];
-		mlx4_dbg(DRV, priv, "Freeing allocator:%d count:%d\n",
-			 i, page_count(page_alloc->page));
+		en_dbg(DRV, priv, "Freeing allocator:%d count:%d\n",
+		       i, page_count(page_alloc->page));

 		put_page(page_alloc->page);
 		page_alloc->page = NULL;
@@ -214,10 +214,10 @@ static void mlx4_en_free_rx_desc(struct mlx4_en_priv *priv,

 	skb_frags = ring->rx_info + (index << priv->log_rx_info);
 	for (nr = 0; nr < priv->num_frags; nr++) {
-		mlx4_dbg(DRV, priv, "Freeing fragment:%d\n", nr);
+		en_dbg(DRV, priv, "Freeing fragment:%d\n", nr);
 		dma = be64_to_cpu(rx_desc->data[nr].addr);

-		mlx4_dbg(DRV, priv, "Unmaping buffer at dma:0x%llx\n", (u64) dma);
+		en_dbg(DRV, priv, "Unmaping buffer at dma:0x%llx\n", (u64) dma);
 		pci_unmap_single(mdev->pdev, dma, skb_frags[nr].size,
 				 PCI_DMA_FROMDEVICE);
 		put_page(skb_frags[nr].page);
@@ -226,7 +226,6 @@ static void mlx4_en_free_rx_desc(struct mlx4_en_priv *priv,

 static int mlx4_en_fill_rx_buffers(struct mlx4_en_priv *priv)
 {
-	struct mlx4_en_dev *mdev = priv->mdev;
 	struct mlx4_en_rx_ring *ring;
 	int ring_ind;
 	int buf_ind;
@@ -239,14 +238,14 @@ static int mlx4_en_fill_rx_buffers(struct mlx4_en_priv *priv)
 			if (mlx4_en_prepare_rx_desc(priv, ring,
 						    ring->actual_size)) {
 				if (ring->actual_size < MLX4_EN_MIN_RX_SIZE) {
-					mlx4_err(mdev, "Failed to allocate "
-						       "enough rx buffers\n");
+					en_err(priv, "Failed to allocate "
+						     "enough rx buffers\n");
 					return -ENOMEM;
 				} else {
 					new_size = rounddown_pow_of_two(ring->actual_size);
-					mlx4_warn(mdev, "Only %d buffers allocated "
-							"reducing ring size to %d",
-						  ring->actual_size, new_size);
+					en_warn(priv, "Only %d buffers allocated "
+						      "reducing ring size to %d",
+						ring->actual_size, new_size);
 					goto reduce_rings;
 				}
 			}
@@ -282,8 +281,7 @@ static int mlx4_en_fill_rx_buf(struct net_device *dev,
 					      ring->size_mask);
 		if (err) {
 			if (netif_msg_rx_err(priv))
-				mlx4_warn(priv->mdev,
-					  "Failed preparing rx descriptor\n");
+				en_warn(priv, "Failed preparing rx descriptor\n");
 			priv->port_stats.rx_alloc_failed++;
 			break;
 		}
@@ -301,14 +299,14 @@ static void mlx4_en_free_rx_buf(struct mlx4_en_priv *priv,
 {
 	int index;

-	mlx4_dbg(DRV, priv, "Freeing Rx buf - cons:%d prod:%d\n",
-			ring->cons, ring->prod);
+	en_dbg(DRV, priv, "Freeing Rx buf - cons:%d prod:%d\n",
+	       ring->cons, ring->prod);

 	/* Unmap and free Rx buffers */
 	BUG_ON((u32) (ring->prod - ring->cons) > ring->actual_size);
 	while (ring->cons != ring->prod) {
 		index = ring->cons & ring->size_mask;
-		mlx4_dbg(DRV, priv, "Processing descriptor:%d\n", index);
+		en_dbg(DRV, priv, "Processing descriptor:%d\n", index);
 		mlx4_en_free_rx_desc(priv, ring, index);
 		++ring->cons;
 	}
@@ -373,10 +371,10 @@ int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,
 					sizeof(struct skb_frag_struct));
 	ring->rx_info = vmalloc(tmp);
 	if (!ring->rx_info) {
-		mlx4_err(mdev, "Failed allocating rx_info ring\n");
+		en_err(priv, "Failed allocating rx_info ring\n");
 		return -ENOMEM;
 	}
-	mlx4_dbg(DRV, priv, "Allocated rx_info ring at addr:%p size:%d\n",
+	en_dbg(DRV, priv, "Allocated rx_info ring at addr:%p size:%d\n",
 		 ring->rx_info, tmp);

 	err = mlx4_alloc_hwq_res(mdev->dev, &ring->wqres,
@@ -386,7 +384,7 @@ int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,

 	err = mlx4_en_map_buffer(&ring->wqres.buf);
 	if (err) {
-		mlx4_err(mdev, "Failed to map RX buffer\n");
+		en_err(priv, "Failed to map RX buffer\n");
 		goto err_hwq;
 	}
 	ring->buf = ring->wqres.buf.direct.buf;
@@ -404,7 +402,7 @@ int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,
 				    sizeof(struct net_lro_desc),
 				    GFP_KERNEL);
 	if (!ring->lro.lro_arr) {
-		mlx4_err(mdev, "Failed to allocate lro array\n");
+		en_err(priv, "Failed to allocate lro array\n");
 		goto err_map;
 	}
 	ring->lro.get_frag_header = mlx4_en_get_frag_header;
@@ -455,7 +453,7 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 		/* Initialize page allocators */
 		err = mlx4_en_init_allocator(priv, ring);
 		if (err) {
-			mlx4_err(mdev, "Failed initializing ring allocator\n");
+			en_err(priv, "Failed initializing ring allocator\n");
 			ring_ind--;
 			goto err_allocator;
 		}
@@ -486,7 +484,7 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 		err = mlx4_srq_alloc(mdev->dev, mdev->priv_pdn, &ring->wqres.mtt,
 				     ring->wqres.db.dma, &ring->srq);
 		if (err){
-			mlx4_err(mdev, "Failed to allocate srq\n");
+			en_err(priv, "Failed to allocate srq\n");
 			ring_ind--;
 			goto err_srq;
 		}
@@ -601,7 +599,7 @@ static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,

 	skb = dev_alloc_skb(SMALL_PACKET_SIZE + NET_IP_ALIGN);
 	if (!skb) {
-		mlx4_dbg(RX_ERR, priv, "Failed allocating skb\n");
+		en_dbg(RX_ERR, priv, "Failed allocating skb\n");
 		return NULL;
 	}
 	skb->dev = priv->dev;
@@ -680,7 +678,6 @@ static void mlx4_en_copy_desc(struct mlx4_en_priv *priv,
 int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int budget)
 {
 	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
 	struct mlx4_cqe *cqe;
 	struct mlx4_en_rx_ring *ring = &priv->rx_ring[cq->ring];
 	struct skb_frag_struct *skb_frags;
@@ -717,14 +714,14 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
 		/* Drop packet on bad receive or bad checksum */
 		if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
 						MLX4_CQE_OPCODE_ERROR)) {
-			mlx4_err(mdev, "CQE completed in error - vendor "
+			en_err(priv, "CQE completed in error - vendor "
 				  "syndrom:%d syndrom:%d\n",
 				  ((struct mlx4_err_cqe *) cqe)->vendor_err_syndrome,
 				  ((struct mlx4_err_cqe *) cqe)->syndrome);
 			goto next;
 		}
 		if (unlikely(cqe->badfcs_enc & MLX4_CQE_BAD_FCS)) {
-			mlx4_dbg(RX_ERR, priv, "Accepted frame with bad FCS\n");
+			en_dbg(RX_ERR, priv, "Accepted frame with bad FCS\n");
 			goto next;
 		}

@@ -874,7 +871,7 @@ static int mlx4_en_last_alloc_offset(struct mlx4_en_priv *priv, u16 stride, u16
 	u16 res = MLX4_EN_ALLOC_SIZE % stride;
 	u16 offset = MLX4_EN_ALLOC_SIZE - stride - res + align;

-	mlx4_dbg(DRV, priv, "Calculated last offset for stride:%d align:%d "
+	en_dbg(DRV, priv, "Calculated last offset for stride:%d align:%d "
 			    "res:%d offset:%d\n", stride, align, res, offset);
 	return offset;
 }
@@ -919,10 +916,10 @@ void mlx4_en_calc_rx_buf(struct net_device *dev)
 	priv->rx_skb_size = eff_mtu;
 	priv->log_rx_info = ROUNDUP_LOG2(i * sizeof(struct skb_frag_struct));

-	mlx4_dbg(DRV, priv, "Rx buffer scatter-list (effective-mtu:%d "
+	en_dbg(DRV, priv, "Rx buffer scatter-list (effective-mtu:%d "
 		  "num_frags:%d):\n", eff_mtu, priv->num_frags);
 	for (i = 0; i < priv->num_frags; i++) {
-		mlx4_dbg(DRV, priv, "  frag:%d - size:%d prefix:%d align:%d "
+		en_dbg(DRV, priv, "  frag:%d - size:%d prefix:%d align:%d "
 				"stride:%d last_offset:%d\n", i,
 				priv->frag_info[i].frag_size,
 				priv->frag_info[i].frag_prefix_size,
@@ -942,12 +939,12 @@ void mlx4_en_set_default_rss_map(struct mlx4_en_priv *priv,
 	int i;

 	rss_map->size = roundup_pow_of_two(num_entries);
-	mlx4_dbg(DRV, priv, "Setting default RSS map of %d entires\n",
-		 rss_map->size);
+	en_dbg(DRV, priv, "Setting default RSS map of %d entires\n",
+	       rss_map->size);

 	for (i = 0; i < rss_map->size; i++) {
 		rss_map->map[i] = i % num_rings;
-		mlx4_dbg(DRV, priv, "Entry %d ---> ring %d\n", i, rss_map->map[i]);
+		en_dbg(DRV, priv, "Entry %d ---> ring %d\n", i, rss_map->map[i]);
 	}
 }

@@ -962,13 +959,13 @@ static int mlx4_en_config_rss_qp(struct mlx4_en_priv *priv,

 	context = kmalloc(sizeof *context , GFP_KERNEL);
 	if (!context) {
-		mlx4_err(mdev, "Failed to allocate qp context\n");
+		en_err(priv, "Failed to allocate qp context\n");
 		return -ENOMEM;
 	}

 	err = mlx4_qp_alloc(mdev->dev, qpn, qp);
 	if (err) {
-		mlx4_err(mdev, "Failed to allocate qp #%d\n", qpn);
+		en_err(priv, "Failed to allocate qp #%x\n", qpn);
 		goto out;
 	}
 	qp->event = mlx4_en_sqp_event;
@@ -1000,12 +997,11 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
 	int err = 0;
 	int good_qps = 0;

-	mlx4_dbg(DRV, priv, "Configuring rss steering for port %u\n", priv->port);
+	en_dbg(DRV, priv, "Configuring rss steering\n");
 	err = mlx4_qp_reserve_range(mdev->dev, rss_map->size,
 				    rss_map->size, &rss_map->base_qpn);
 	if (err) {
-		mlx4_err(mdev, "Failed reserving %d qps for port %u\n",
-			 rss_map->size, priv->port);
+		en_err(priv, "Failed reserving %d qps\n", rss_map->size);
 		return err;
 	}

@@ -1025,13 +1021,13 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
 	/* Configure RSS indirection qp */
 	err = mlx4_qp_reserve_range(mdev->dev, 1, 1, &priv->base_qpn);
 	if (err) {
-		mlx4_err(mdev, "Failed to reserve range for RSS "
-			       "indirection qp\n");
+		en_err(priv, "Failed to reserve range for RSS "
+			     "indirection qp\n");
 		goto rss_err;
 	}
 	err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, &rss_map->indir_qp);
 	if (err) {
-		mlx4_err(mdev, "Failed to allocate RSS indirection QP\n");
+		en_err(priv, "Failed to allocate RSS indirection QP\n");
 		goto reserve_err;
 	}
 	rss_map->indir_qp.event = mlx4_en_sqp_event;
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index aeeb334..e32e869 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -68,15 +68,15 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
 	tmp = size * sizeof(struct mlx4_en_tx_info);
 	ring->tx_info = vmalloc(tmp);
 	if (!ring->tx_info) {
-		mlx4_err(mdev, "Failed allocating tx_info ring\n");
+		en_err(priv, "Failed allocating tx_info ring\n");
 		return -ENOMEM;
 	}
-	mlx4_dbg(DRV, priv, "Allocated tx_info ring at addr:%p size:%d\n",
+	en_dbg(DRV, priv, "Allocated tx_info ring at addr:%p size:%d\n",
 		 ring->tx_info, tmp);

 	ring->bounce_buf = kmalloc(MAX_DESC_SIZE, GFP_KERNEL);
 	if (!ring->bounce_buf) {
-		mlx4_err(mdev, "Failed allocating bounce buffer\n");
+		en_err(priv, "Failed allocating bounce buffer\n");
 		err = -ENOMEM;
 		goto err_tx;
 	}
@@ -85,31 +85,31 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
 	err = mlx4_alloc_hwq_res(mdev->dev, &ring->wqres, ring->buf_size,
 				 2 * PAGE_SIZE);
 	if (err) {
-		mlx4_err(mdev, "Failed allocating hwq resources\n");
+		en_err(priv, "Failed allocating hwq resources\n");
 		goto err_bounce;
 	}

 	err = mlx4_en_map_buffer(&ring->wqres.buf);
 	if (err) {
-		mlx4_err(mdev, "Failed to map TX buffer\n");
+		en_err(priv, "Failed to map TX buffer\n");
 		goto err_hwq_res;
 	}

 	ring->buf = ring->wqres.buf.direct.buf;

-	mlx4_dbg(DRV, priv, "Allocated TX ring (addr:%p) - buf:%p size:%d "
-		 "buf_size:%d dma:%llx\n", ring, ring->buf, ring->size,
-		 ring->buf_size, (unsigned long long) ring->wqres.buf.direct.map);
+	en_dbg(DRV, priv, "Allocated TX ring (addr:%p) - buf:%p size:%d "
+	       "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) {
-		mlx4_err(mdev, "Failed reserving qp for tx ring.\n");
+		en_err(priv, "Failed reserving qp for tx ring.\n");
 		goto err_map;
 	}

 	err = mlx4_qp_alloc(mdev->dev, ring->qpn, &ring->qp);
 	if (err) {
-		mlx4_err(mdev, "Failed allocating qp %d\n", ring->qpn);
+		en_err(priv, "Failed allocating qp %d\n", ring->qpn);
 		goto err_reserve;
 	}
 	ring->qp.event = mlx4_en_sqp_event;
@@ -135,7 +135,7 @@ void mlx4_en_destroy_tx_ring(struct mlx4_en_priv *priv,
 			     struct mlx4_en_tx_ring *ring)
 {
 	struct mlx4_en_dev *mdev = priv->mdev;
-	mlx4_dbg(DRV, priv, "Destroying tx ring, qpn: %d\n", ring->qpn);
+	en_dbg(DRV, priv, "Destroying tx ring, qpn: %d\n", ring->qpn);

 	mlx4_qp_remove(mdev->dev, &ring->qp);
 	mlx4_qp_free(mdev->dev, &ring->qp);
@@ -274,12 +274,12 @@ int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring)

 	/* Skip last polled descriptor */
 	ring->cons += ring->last_nr_txbb;
-	mlx4_dbg(DRV, priv, "Freeing Tx buf - cons:0x%x prod:0x%x\n",
+	en_dbg(DRV, priv, "Freeing Tx buf - cons:0x%x prod:0x%x\n",
 		 ring->cons, ring->prod);

 	if ((u32) (ring->prod - ring->cons) > ring->size) {
 		if (netif_msg_tx_err(priv))
-			mlx4_warn(priv->mdev, "Tx consumer passed producer!\n");
+			en_warn(priv, "Tx consumer passed producer!\n");
 		return 0;
 	}

@@ -292,7 +292,7 @@ int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring)
 	}

 	if (cnt)
-		mlx4_dbg(DRV, priv, "Freed %d uncompleted tx descriptors\n", cnt);
+		en_dbg(DRV, priv, "Freed %d uncompleted tx descriptors\n", cnt);

 	return cnt;
 }
@@ -321,7 +321,7 @@ void mlx4_en_set_prio_map(struct mlx4_en_priv *priv, u16 *prio_map, u32 ring_num
 			num = 0;
 		}
 		prio_map[prio] = ring;
-		mlx4_dbg(DRV, priv, " prio:%d --> ring:%d\n", prio, ring);
+		en_dbg(DRV, priv, " prio:%d --> ring:%d\n", prio, ring);
 		num++;
 	}
 }
@@ -539,7 +539,6 @@ static int get_real_size(struct sk_buff *skb, struct net_device *dev,
 			 int *lso_header_size)
 {
 	struct mlx4_en_priv *priv = netdev_priv(dev);
-	struct mlx4_en_dev *mdev = priv->mdev;
 	int real_size;

 	if (skb_is_gso(skb)) {
@@ -553,14 +552,14 @@ static int get_real_size(struct sk_buff *skb, struct net_device *dev,
 				real_size += DS_SIZE;
 			else {
 				if (netif_msg_tx_err(priv))
-					mlx4_warn(mdev, "Non-linear headers\n");
+					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))
-				mlx4_warn(mdev, "LSO header size too big\n");
+				en_warn(priv, "LSO header size too big\n");
 			dev_kfree_skb_any(skb);
 			return 0;
 		}
@@ -669,7 +668,7 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
 	nr_txbb = desc_size / TXBB_SIZE;
 	if (unlikely(nr_txbb > MAX_DESC_TXBBS)) {
 		if (netif_msg_tx_err(priv))
-			mlx4_warn(mdev, "Oversized header or SG list\n");
+			en_warn(priv, "Oversized header or SG list\n");
 		dev_kfree_skb_any(skb);
 		return NETDEV_TX_OK;
 	}
@@ -695,7 +694,7 @@ int mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
 	/* Now that we know what Tx ring to use */
 	if (unlikely(!priv->port_up)) {
 		if (netif_msg_tx_err(priv))
-			mlx4_warn(mdev, "xmit: port down!\n");
+			en_warn(priv, "xmit: port down!\n");
 		dev_kfree_skb_any(skb);
 		return NETDEV_TX_OK;
 	}
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index ef840ab..c92b382 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -55,20 +55,36 @@

 #define MLX4_EN_MSG_LEVEL	(NETIF_MSG_LINK | NETIF_MSG_IFDOWN)

-#define mlx4_dbg(mlevel, priv, format, arg...)	\
-	if (NETIF_MSG_##mlevel & priv->msg_enable) \
-	printk(KERN_DEBUG "%s %s: " format , DRV_NAME ,\
-		(dev_name(&priv->mdev->pdev->dev)) , ## arg)
+#define en_print(level, priv, format, arg...)			\
+	{							\
+	if ((priv)->registered)					\
+		printk(level "%s: %s: " format, DRV_NAME,	\
+			(priv->dev)->name, ## arg);		\
+	else							\
+		printk(level "%s: %s: Port %d: " format,	\
+			DRV_NAME, dev_name(&priv->mdev->pdev->dev), \
+			(priv)->port, ## arg);			\
+	}
+
+#define en_dbg(mlevel, priv, format, arg...)			\
+	{							\
+	if (NETIF_MSG_##mlevel & priv->msg_enable)		\
+		en_print(KERN_DEBUG, priv, format, ## arg)	\
+	}
+#define en_warn(priv, format, arg...)				\
+	en_print(KERN_WARNING, priv, format, ## arg)
+#define en_err(priv, format, arg...)				\
+	en_print(KERN_ERR, priv, format, ## arg)

 #define mlx4_err(mdev, format, arg...) \
 	printk(KERN_ERR "%s %s: " format , DRV_NAME ,\
-		(dev_name(&mdev->pdev->dev)) , ## arg)
+		dev_name(&mdev->pdev->dev) , ## arg)
 #define mlx4_info(mdev, format, arg...) \
 	printk(KERN_INFO "%s %s: " format , DRV_NAME ,\
-		(dev_name(&mdev->pdev->dev)) , ## arg)
+		dev_name(&mdev->pdev->dev) , ## arg)
 #define mlx4_warn(mdev, format, arg...) \
 	printk(KERN_WARNING "%s %s: " format , DRV_NAME ,\
-		(dev_name(&mdev->pdev->dev)) , ## arg)
+		dev_name(&mdev->pdev->dev) , ## arg)

 /*
  * Device constants
-- 
1.6.0

Re: [PATCH 1/8] mlx4_en: Giving interface name in debug messages

[David Miller]

From: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Date: Tue, 02 Jun 2009 09:27:13 +0300

> For each debug message, the message will show interface name in case
> that the net device was registered, and PCI bus ID with port number
> if we were not registered yet. Messages that are not port/netdev specific
> stayed in the old format
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>

So where are the 7 other patches of this 8 patch set?