[PATCH] net/mlx5: allow jump to current group

[PATCH] net/mlx5: allow jump to current group

[Michael Baum]

In current implementation, jump action is allowed only if target
group is different than the current flow group.

Change jump action validation such that it allows jump to current group.

Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
---
 drivers/net/mlx5/mlx5_flow_dv.c | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index 9753af2cb1..0d5bf92320 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -5480,13 +5480,6 @@ flow_dv_validate_action_jump(struct rte_eth_dev *dev,
 				       &grp_info, error);
 	if (ret)
 		return ret;
-	if (attributes->group == target_group &&
-	    !(action_flags & (MLX5_FLOW_ACTION_TUNNEL_SET |
-			      MLX5_FLOW_ACTION_TUNNEL_MATCH)))
-		return rte_flow_error_set(error, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_ACTION, NULL,
-					  "target group must be other than"
-					  " the current flow group");
 	if (table == 0)
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ACTION_CONF,
-- 
2.25.1

RE: [PATCH] net/mlx5: allow jump to current group

[Raslan Darawsheh]

Hi Michael,

> -----Original Message-----
> From: Michael Baum <michaelba@nvidia.com>
> Sent: Monday, November 13, 2023 12:44 PM
> To: dev@dpdk.org
> Cc: Matan Azrad <matan@nvidia.com>; Raslan Darawsheh
> <rasland@nvidia.com>; Slava Ovsiienko <viacheslavo@nvidia.com>; Ori Kam
> <orika@nvidia.com>; Suanming Mou <suanmingm@nvidia.com>
> Subject: [PATCH] net/mlx5: allow jump to current group
I think this is a fix 
> 
> In current implementation, jump action is allowed only if target
> group is different than the current flow group.
> 
> Change jump action validation such that it allows jump to current group.
also don't forget to add Fixes tag
You should have a Cc: stable@dpdk.org

Kindest regards
Raslan Darawsheh

[PATCH v2] net/mlx5: fix jump action validation

[Michael Baum]

Currently PMD doesn't allow to jump to the same group in order to
avoid dead loop. But this also prevent experienced user to create
flow with less Hops in order to have better performance.

For example, rules in [1] should have better performance then [2].

Furthermore, this protection will not really prevent dead loop, i.e
[3]. So just remove this protection and user should take the
responsibility to avoid dead loop.

This patch enables jumping to the same group.

[1]:
flow create 0 group 1 priority 1 pattern eth / ipv4 / udp / gtp / end
actions raw_decap / raw_encap / jump group 1 / end
flow create 0 group 1 priority 0 pattern eth / ipv4 src is 1.0.0.1 / tcp
/ end actions queues index 1 / end

[2]:
flow create 0 group 1 priority 0 pattern eth / ipv4 / udp / gtp / end
actions raw_decap / raw_encap / jump group 2 / end
flow create 0 group 2 priority 0 pattern eth / ipv4 src is 1.0.0.1 / tcp
/ end actions queues index 1 / end

[3]:
flow create 0 group 1 pattern eth / end actions jump group 2 / end
flow create 0 group 2 pattern eth / end actions jump group 1 / end

Fixes: f78f747f41d0 ("net/mlx5: allow jump to group lower than current")
Cc: dekelp@mellanox.com
Cc: stable@dpdk.org

Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
---

V2: change commit message to fix template.

 drivers/net/mlx5/mlx5_flow_dv.c | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index 115d730317..5d0cb4278b 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -5484,13 +5484,6 @@ flow_dv_validate_action_jump(struct rte_eth_dev *dev,
 				       &grp_info, error);
 	if (ret)
 		return ret;
-	if (attributes->group == target_group &&
-	    !(action_flags & (MLX5_FLOW_ACTION_TUNNEL_SET |
-			      MLX5_FLOW_ACTION_TUNNEL_MATCH)))
-		return rte_flow_error_set(error, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_ACTION, NULL,
-					  "target group must be other than"
-					  " the current flow group");
 	if (table == 0)
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ACTION_CONF,
-- 
2.25.1

RE: [PATCH v2] net/mlx5: fix jump action validation

[Raslan Darawsheh]

Hi,

> -----Original Message-----
> From: Michael Baum <michaelba@nvidia.com>
> Sent: Monday, November 27, 2023 2:43 PM
> To: dev@dpdk.org
> Cc: Matan Azrad <matan@nvidia.com>; Raslan Darawsheh
> <rasland@nvidia.com>; Slava Ovsiienko <viacheslavo@nvidia.com>; Ori Kam
> <orika@nvidia.com>; Suanming Mou <suanmingm@nvidia.com>;
> dekelp@mellanox.com; stable@dpdk.org
> Subject: [PATCH v2] net/mlx5: fix jump action validation
> 
> Currently PMD doesn't allow to jump to the same group in order to
> avoid dead loop. But this also prevent experienced user to create
> flow with less Hops in order to have better performance.
> 
> For example, rules in [1] should have better performance then [2].
> 
> Furthermore, this protection will not really prevent dead loop, i.e
> [3]. So just remove this protection and user should take the
> responsibility to avoid dead loop.
> 
> This patch enables jumping to the same group.
> 
> [1]:
> flow create 0 group 1 priority 1 pattern eth / ipv4 / udp / gtp / end
> actions raw_decap / raw_encap / jump group 1 / end
> flow create 0 group 1 priority 0 pattern eth / ipv4 src is 1.0.0.1 / tcp
> / end actions queues index 1 / end
> 
> [2]:
> flow create 0 group 1 priority 0 pattern eth / ipv4 / udp / gtp / end
> actions raw_decap / raw_encap / jump group 2 / end
> flow create 0 group 2 priority 0 pattern eth / ipv4 src is 1.0.0.1 / tcp
> / end actions queues index 1 / end
> 
> [3]:
> flow create 0 group 1 pattern eth / end actions jump group 2 / end
> flow create 0 group 2 pattern eth / end actions jump group 1 / end
> 
> Fixes: f78f747f41d0 ("net/mlx5: allow jump to group lower than current")
> Cc: dekelp@mellanox.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Michael Baum <michaelba@nvidia.com>
> Acked-by: Matan Azrad <matan@nvidia.com>
> ---
> 
> V2: change commit message to fix template. 
 
Patch applied to next-net-mlx, 
Kindest regards
Raslan Darawsheh