[PATCH net 0/2] mlx4 driver encapsulation/steering fixes

[PATCH net 0/2] mlx4 driver encapsulation/steering fixes

[Or Gerlitz]

Hi Dave,

The 1st patch fixes a bug in the TX path that supports offloading the 
TX checksum of (VXLAN) encapsulated TCP packets. It turns out that the 
bug is revealed only when the receiver runs in non-offloaded mode, so
we somehow missed it so far... please queue it for -stable >= 3.14 

The 2nd patch makes sure not to leak steering entry on error flow, 
please queue it to 3.17-stable 

thanks,

Or.

Or Gerlitz (2):
  net/mlx4_en: Don't attempt to TX offload the outer UDP checksum for VXLAN
  mlx4: Avoid leaking steering rules on flow creation error flow

 drivers/infiniband/hw/mlx4/main.c          |   10 ++++++++--
 drivers/net/ethernet/mellanox/mlx4/en_tx.c |    7 +++++--
 drivers/net/ethernet/mellanox/mlx4/mcg.c   |    4 ++++
 3 files changed, 17 insertions(+), 4 deletions(-)

[PATCH net 1/2] net/mlx4_en: Don't attempt to TX offload the outer UDP checksum for VXLAN

[Or Gerlitz]

For VXLAN/NVGRE encapsulation, the current HW doesn't support offloading
both the outer UDP TX checksum and the inner TCP/UDP TX checksum. 

The driver doesn't advertize SKB_GSO_UDP_TUNNEL_CSUM, however we are wrongly
telling the HW to offload the outer UDP checksum for encapsulated packets,
fix that.

Fixes: 837052d0ccc5 ('net/mlx4_en: Add netdev support for TCP/IP
		     offloads of vxlan tunneling')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/en_tx.c |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_tx.c b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
index 34c1378..454d9fe 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
@@ -836,8 +836,11 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
 	 * whether LSO is used */
 	tx_desc->ctrl.srcrb_flags = priv->ctrl_flags;
 	if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
-		tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM |
-							 MLX4_WQE_CTRL_TCP_UDP_CSUM);
+		if (!skb->encapsulation)
+			tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM |
+								 MLX4_WQE_CTRL_TCP_UDP_CSUM);
+		else
+			tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM);
 		ring->tx_csum++;
 	}
 
-- 
1.7.1

[PATCH net 2/2] mlx4: Avoid leaking steering rules on flow creation error flow

[Or Gerlitz]

If mlx4_ib_create_flow() attempts to create > 1 rules with the
firmware, and one of these registrations fail, we leaked the
already created flow rules.

One example of the leak is when the registration of the VXLAN ghost
steering rule fails, we didn't unregister the original rule requested
by the user, introduced in commit d2fce8a9060d "mlx4: Set
user-space raw Ethernet QPs to properly handle VXLAN traffic".

While here, add dump of the VXLAN portion of steering rules
so it can actually be seen when flow creation fails.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
---
 drivers/infiniband/hw/mlx4/main.c        |   10 ++++++++--
 drivers/net/ethernet/mellanox/mlx4/mcg.c |    4 ++++
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index bda5994..8b72cf3 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -1173,18 +1173,24 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp,
 		err = __mlx4_ib_create_flow(qp, flow_attr, domain, type[i],
 					    &mflow->reg_id[i]);
 		if (err)
-			goto err_free;
+			goto err_create_flow;
 		i++;
 	}
 
 	if (i < ARRAY_SIZE(type) && flow_attr->type == IB_FLOW_ATTR_NORMAL) {
 		err = mlx4_ib_tunnel_steer_add(qp, flow_attr, &mflow->reg_id[i]);
 		if (err)
-			goto err_free;
+			goto err_create_flow;
+		i++;
 	}
 
 	return &mflow->ibflow;
 
+err_create_flow:
+	while (i) {
+		(void)__mlx4_ib_destroy_flow(to_mdev(qp->device)->dev, mflow->reg_id[i]);
+		i--;
+	}
 err_free:
 	kfree(mflow);
 	return ERR_PTR(err);
diff --git a/drivers/net/ethernet/mellanox/mlx4/mcg.c b/drivers/net/ethernet/mellanox/mlx4/mcg.c
index ca0f98c..8728431 100644
--- a/drivers/net/ethernet/mellanox/mlx4/mcg.c
+++ b/drivers/net/ethernet/mellanox/mlx4/mcg.c
@@ -955,6 +955,10 @@ static void mlx4_err_rule(struct mlx4_dev *dev, char *str,
 					cur->ib.dst_gid_msk);
 			break;
 
+		case MLX4_NET_TRANS_RULE_ID_VXLAN:
+			len += snprintf(buf + len, BUF_SIZE - len,
+					"VNID = %d ", be32_to_cpu(cur->vxlan.vni));
+			break;
 		case MLX4_NET_TRANS_RULE_ID_IPV6:
 			break;
 
-- 
1.7.1

Re: [PATCH net 0/2] mlx4 driver encapsulation/steering fixes

[David Miller]

From: Or Gerlitz <ogerlitz@mellanox.com>
Date: Thu, 30 Oct 2014 15:59:26 +0200

> The 1st patch fixes a bug in the TX path that supports offloading the 
> TX checksum of (VXLAN) encapsulated TCP packets. It turns out that the 
> bug is revealed only when the receiver runs in non-offloaded mode, so
> we somehow missed it so far... please queue it for -stable >= 3.14 
> 
> The 2nd patch makes sure not to leak steering entry on error flow, 
> please queue it to 3.17-stable 

Applied and queue up for -stable.