DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] net/mlx5: improve validation of item order
@ 2019-09-11  5:25 Xiaoyu Min
  2019-09-11  8:46 ` [dpdk-dev] [PATCH v2] " Xiaoyu Min
  0 siblings, 1 reply; 4+ messages in thread
From: Xiaoyu Min @ 2019-09-11  5:25 UTC (permalink / raw)
  To: Matan Azrad, Shahaf Shuler, Viacheslav Ovsiienko; +Cc: dev

The Item order validation between L2 and L3 is missing, which leading to
the following flow rule is accepted:

  testpmd> flow create 0 ingress pattern ipv4 / eth / end actions drop /
           end

Only the outer L3 layer should check whether the L2 layer is present,
because the L3 layer could directly follow the tunnel layer
without L2 layer.

Meanwhile inner L2 layer should check whether there is inner L3 layer
before it.

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
---
 drivers/net/mlx5/mlx5_flow.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index eb360525da..45bd9c8025 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1224,6 +1224,11 @@ mlx5_flow_validate_item_eth(const struct rte_flow_item *item,
 		return rte_flow_error_set(error, ENOTSUP,
 					  RTE_FLOW_ERROR_TYPE_ITEM, item,
 					  "multiple L2 layers not supported");
+	if (tunnel && (item_flags & MLX5_FLOW_LAYER_INNER_L3))
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ITEM, item,
+					  "inner L2 layer should not "
+					  "follow inner L3 layers");
 	if (!mask)
 		mask = &rte_flow_item_eth_mask;
 	ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask,
@@ -1270,6 +1275,8 @@ mlx5_flow_validate_item_vlan(const struct rte_flow_item *item,
 	const uint64_t vlanm = tunnel ? MLX5_FLOW_LAYER_INNER_VLAN :
 					MLX5_FLOW_LAYER_OUTER_VLAN;
 
+	const uint64_t l2m = tunnel ? MLX5_FLOW_LAYER_INNER_L2 :
+				      MLX5_FLOW_LAYER_OUTER_L2;
 	if (item_flags & vlanm)
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ITEM, item,
@@ -1278,6 +1285,10 @@ mlx5_flow_validate_item_vlan(const struct rte_flow_item *item,
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ITEM, item,
 					  "L2 layer cannot follow L3/L4 layer");
+	else if ((item_flags & l2m) == 0)
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ITEM, item,
+					  "no L2 layer before VLAN");
 	if (!mask)
 		mask = &rte_flow_item_vlan_mask;
 	ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask,
@@ -1390,6 +1401,10 @@ mlx5_flow_validate_item_ipv4(const struct rte_flow_item *item,
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ITEM, item,
 					  "L3 cannot follow an NVGRE layer.");
+	else if (!tunnel && !(item_flags & MLX5_FLOW_LAYER_OUTER_L2))
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ITEM, item,
+					  "no L2 layer before IPV4");
 	if (!mask)
 		mask = &rte_flow_item_ipv4_mask;
 	else if (mask->hdr.next_proto_id != 0 &&
@@ -1481,6 +1496,10 @@ mlx5_flow_validate_item_ipv6(const struct rte_flow_item *item,
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ITEM, item,
 					  "L3 cannot follow an NVGRE layer.");
+	else if (!tunnel && !(item_flags & MLX5_FLOW_LAYER_OUTER_L2))
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ITEM, item,
+					  "no L2 layer before IPV6");
 	if (!mask)
 		mask = &rte_flow_item_ipv6_mask;
 	ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask,
-- 
2.23.0


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2019-10-08 11:53 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-09-11  5:25 [dpdk-dev] [PATCH] net/mlx5: improve validation of item order Xiaoyu Min
2019-09-11  8:46 ` [dpdk-dev] [PATCH v2] " Xiaoyu Min
2019-10-08  9:43   ` Slava Ovsiienko
2019-10-08 11:53   ` Raslan Darawsheh

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).