[dpdk-dev] [PATCH 0/5] net/mlx5: fixes

[dpdk-dev] [PATCH 0/5] net/mlx5: fixes

[Nelio Laranjeiro]

- Multiple fixes for tunnels in flow actions.
- fix some mistakes in flow director code.

Nelio Laranjeiro (5):
  net/mlx5: fix parser inner value
  net/mlx5: fix flows when VXLAN tunnel is 0
  net/mlx5: fix RSS tunneled flow without outer
  net/mlx5: fix fdir flow removal
  net/mlx5: fix flow director TCPv6 filter

 doc/guides/nics/mlx5.rst     |  2 ++
 drivers/net/mlx5/mlx5_flow.c | 29 +++++++++++++++++++----------
 2 files changed, 21 insertions(+), 10 deletions(-)

-- 
2.11.0

[dpdk-dev] [PATCH 1/5] net/mlx5: fix parser inner value

[Nelio Laranjeiro]

In validate() it sets the value to 1 whereas the expected one is
IBV_FLOW_SPEC_INNER.

Fixes: c2c6ddd51da4 ("net/mlx5: limit flow API rules to one tunnel")
Cc: stable@dpdk.org

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---
 drivers/net/mlx5/mlx5_flow.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 3eb84e2a4..68cc8bff5 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -866,7 +866,7 @@ priv_flow_convert_items_validate(struct priv *priv,
 						   " VXLAN encapsulations");
 				return -rte_errno;
 			}
-			parser->inner = 1;
+			parser->inner = IBV_FLOW_SPEC_INNER;
 		}
 		if (parser->drop) {
 			parser->drop_q.offset += cur_item->dst_sz;
-- 
2.11.0

Re: [dpdk-dev] [PATCH 1/5] net/mlx5: fix parser inner value

[Yongseok Koh]

> On Oct 31, 2017, at 8:51 AM, Nelio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:
> 
> In validate() it sets the value to 1 whereas the expected one is
> IBV_FLOW_SPEC_INNER.
> 
> Fixes: c2c6ddd51da4 ("net/mlx5: limit flow API rules to one tunnel")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
> ---
Acked-by: Yongseok Koh <yskoh@mellanox.com>
 
Thanks

[dpdk-dev] [PATCH 2/5] net/mlx5: fix flows when VXLAN tunnel is 0

[Nelio Laranjeiro]

Fix a strange behavior from the NIC, when the flow starts with a VXLAN
layer with a VNI equals to zero all the traffic will match within this
rule.

Fixes: 2e709b6aa0f5 ("net/mlx5: support VXLAN flow item")
Cc: stable@dpdk.org

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---
 doc/guides/nics/mlx5.rst     |  2 ++
 drivers/net/mlx5/mlx5_flow.c | 12 +++++++++++-
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst
index cdb880a4c..19da9f1ed 100644
--- a/doc/guides/nics/mlx5.rst
+++ b/doc/guides/nics/mlx5.rst
@@ -129,6 +129,8 @@ Limitations
   is set to multi-packet send or Enhanced multi-packet send. Otherwise it must have
   less than 50 segments.
 - Count action for RTE flow is only supported in Mellanox OFED 4.2.
+- Flows with a VXLAN Network Identifier equal (or ends to be equal)
+  to 0 are not supported.
 
 Configuration
 -------------
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 68cc8bff5..a7864871c 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1158,7 +1158,7 @@ priv_flow_convert(struct priv *priv,
 					 cur_item->mask),
 					parser);
 		if (ret) {
-			rte_flow_error_set(error, ENOTSUP,
+			rte_flow_error_set(error, ret,
 					   RTE_FLOW_ERROR_TYPE_ITEM,
 					   items, "item not supported");
 			goto exit_free;
@@ -1605,6 +1605,16 @@ mlx5_flow_create_vxlan(const struct rte_flow_item *item,
 		/* Remove unwanted bits from values. */
 		vxlan.val.tunnel_id &= vxlan.mask.tunnel_id;
 	}
+	/*
+	 * Tunnel id 0 is equivalent as not adding a VXLAN layer, if only this
+	 * layer is defined in the Verbs specification it is interpreted as
+	 * wildcard and all packets will match this rule, if it follows a full
+	 * stack layer (ex: eth / ipv4 / udp), all packets matching the layers
+	 * before will also match this rule.
+	 * To avoid such situation, VNI 0 is currently refused.
+	 */
+	if (!vxlan.val.tunnel_id)
+		return EINVAL;
 	mlx5_flow_create_copy(parser, &vxlan, size);
 	return 0;
 }
-- 
2.11.0

Re: [dpdk-dev] [PATCH 2/5] net/mlx5: fix flows when VXLAN tunnel is 0

[Yongseok Koh]

> On Oct 31, 2017, at 8:51 AM, Nelio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:
> 
> Fix a strange behavior from the NIC, when the flow starts with a VXLAN
> layer with a VNI equals to zero all the traffic will match within this
> rule.
> 
> Fixes: 2e709b6aa0f5 ("net/mlx5: support VXLAN flow item")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
> ---
Acked-by: Yongseok Koh <yskoh@mellanox.com>
 
Thanks

[dpdk-dev] [PATCH 3/5] net/mlx5: fix RSS tunneled flow without outer

[Nelio Laranjeiro]

Allow to use the RSS when the outer part of the packet is not defined.  The
RSS is still performed on the outer.
This modification only serves the internal engine to create the necessary
Verbs flows according to the user request.

Fixes: 39797c6db80b ("net/mlx5: fix RSS action for tunneled packets")
Cc: shahafs@mellanox.com

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---
 drivers/net/mlx5/mlx5_flow.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index a7864871c..fcc9a260e 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1010,15 +1010,13 @@ priv_flow_convert_finalise(struct priv *priv, struct mlx5_flow_parse *parser)
 			if (hash_rxq_init[i].ip_version == MLX5_IPV4) {
 				size = sizeof(struct ibv_flow_spec_ipv4_ext);
 				specs.ipv4 = (struct ibv_flow_spec_ipv4_ext){
-					.type = IBV_FLOW_SPEC_IPV4_EXT |
-						parser->inner,
+					.type = IBV_FLOW_SPEC_IPV4_EXT,
 					.size = size,
 				};
 			} else {
 				size = sizeof(struct ibv_flow_spec_ipv6);
 				specs.ipv6 = (struct ibv_flow_spec_ipv6){
-					.type = IBV_FLOW_SPEC_IPV6 |
-						parser->inner,
+					.type = IBV_FLOW_SPEC_IPV6,
 					.size = size,
 				};
 			}
@@ -1038,8 +1036,7 @@ priv_flow_convert_finalise(struct priv *priv, struct mlx5_flow_parse *parser)
 				.type = ((i == HASH_RXQ_UDPV4 ||
 					  i == HASH_RXQ_UDPV6) ?
 					 IBV_FLOW_SPEC_UDP :
-					 IBV_FLOW_SPEC_TCP) |
-					parser->inner,
+					 IBV_FLOW_SPEC_TCP),
 				.size = size,
 			};
 			if (parser->queue[i].ibv_attr) {
-- 
2.11.0

Re: [dpdk-dev] [PATCH 3/5] net/mlx5: fix RSS tunneled flow without outer

[Yongseok Koh]

> On Oct 31, 2017, at 8:51 AM, Nelio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:
> 
> Allow to use the RSS when the outer part of the packet is not defined.  The
> RSS is still performed on the outer.
> This modification only serves the internal engine to create the necessary
> Verbs flows according to the user request.
> 
> Fixes: 39797c6db80b ("net/mlx5: fix RSS action for tunneled packets")
> Cc: shahafs@mellanox.com
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
> ---
Acked-by: Yongseok Koh <yskoh@mellanox.com>
 
Thanks

[dpdk-dev] [PATCH 4/5] net/mlx5: fix fdir flow removal

[Nelio Laranjeiro]

Initialise struct mlx5_fdir to avoid passing non initialised items/actions
arrays elements to the generic flow interpreter.

Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---
 drivers/net/mlx5/mlx5_flow.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index fcc9a260e..d8b0d7508 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -2847,7 +2847,9 @@ static int
 priv_fdir_filter_delete(struct priv *priv,
 			const struct rte_eth_fdir_filter *fdir_filter)
 {
-	struct mlx5_fdir attributes;
+	struct mlx5_fdir attributes = {
+		.attr.group = 0,
+	};
 	struct mlx5_flow_parse parser = {
 		.create = 1,
 		.layer = HASH_RXQ_ETH,
-- 
2.11.0

Re: [dpdk-dev] [PATCH 4/5] net/mlx5: fix fdir flow removal

[Yongseok Koh]

> On Oct 31, 2017, at 8:51 AM, Nelio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:
> 
> Initialise struct mlx5_fdir to avoid passing non initialised items/actions
> arrays elements to the generic flow interpreter.
> 
> Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
> ---
Acked-by: Yongseok Koh <yskoh@mellanox.com>
 
Thanks

[dpdk-dev] [PATCH 5/5] net/mlx5: fix flow director TCPv6 filter

[Nelio Laranjeiro]

Flow director to generic flow conversion wrongly handles TCPv6 layer.

Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---
 drivers/net/mlx5/mlx5_flow.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index d8b0d7508..8ff2a02c8 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -2754,7 +2754,7 @@ priv_fdir_filter_convert(struct priv *priv,
 			.spec = &attributes->l3,
 		};
 		attributes->items[2] = (struct rte_flow_item){
-			.type = RTE_FLOW_ITEM_TYPE_UDP,
+			.type = RTE_FLOW_ITEM_TYPE_TCP,
 			.spec = &attributes->l4,
 		};
 		break;
-- 
2.11.0

Re: [dpdk-dev] [PATCH 5/5] net/mlx5: fix flow director TCPv6 filter

[Yongseok Koh]

> On Oct 31, 2017, at 8:51 AM, Nelio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:
> 
> Flow director to generic flow conversion wrongly handles TCPv6 layer.
> 
> Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
> ---
Acked-by: Yongseok Koh <yskoh@mellanox.com>
 
Thanks

Re: [dpdk-dev] [PATCH 0/5] net/mlx5: fixes

[Ferruh Yigit]

On 10/31/2017 8:51 AM, Nelio Laranjeiro wrote:
> - Multiple fixes for tunnels in flow actions.
> - fix some mistakes in flow director code.
> 
> Nelio Laranjeiro (5):
>   net/mlx5: fix parser inner value
>   net/mlx5: fix flows when VXLAN tunnel is 0
>   net/mlx5: fix RSS tunneled flow without outer
>   net/mlx5: fix fdir flow removal
>   net/mlx5: fix flow director TCPv6 filter

Series applied to dpdk-next-net/master, thanks.


Except patch 3/5, it squashed into relevant commit [1].

[1]: "net/mlx5: fix RSS action for tunneled packets"