net/mlx5: fix hairpin split with set VLAN VID action
PATCH 1/1 net/mlx5: fix hairpin split with set VLAN VID action
Before this patch any flow rule which works on hairpin queues and which has OF_SET_VLAN_VID action was split into 2 flow rules:
- one subflow for Rx,
- one subflow for Tx.
OF_SET_VLAN_VID action was always placed in the Tx subflow.
Assuming a flow rule which matches VLAN traffic and has both OF_SET_VLAN_VID action, and MODIFY_FIELD action on VLAN VID, but no OF_PUSH_VLAN action, the following happened:
- MODIFY_FIELD action was placed in Rx subflow,
- OF_SET_VLAN_VID action was placed in Tx subflow,
- OF_SET_VLAN_VID action is internally compiled to a header modify command.
This caused the following issues:
1. Since OF_SET_VLAN_VID was placed in Tx subflow, 2 header modify actions were allocated. One for Rx and one for Tx.
2. If OF_SET_VLAN_VID action was placed before MODIFY_FIELD on VLAN VID, the flow rule executed header modifications in reverse order. MODIFY_FIELD actions were executed first in the Rx subflow and OF_SET_VLAN_VID was executed second in Tx subflow.
This patch fixes this behavior by not splitting hairpin flow rules if OF_SET_VLAN_VID action is used without OF_PUSH_VLAN. On top of that, if flow rule is split, the OF_SET_VLAN_VID action is not moved to Tx subflow (for flow rules mentioned above).
--- drivers/net/mlx5/mlx5_flow.c | 24 ++++++++++++++++++++++-- 1 file changed, 22 insertions(+), 2 deletions(-)
drivers/net/mlx5/mlx5_flow.c +22 -2
@@ -4591,6 +4591,7 @@ flow_check_hairpin_split(struct rte_eth_dev *dev,int queue_action =0 ;int action_n =0 ;int split =0 ; +int push_vlan =0 ;const struct rte_flow_action_queue *queue;const struct rte_flow_action_rss *rss;const struct rte_flow_action_raw_encap *raw_encap;
@@ -4599,6 +4600,8 @@ flow_check_hairpin_split(struct rte_eth_dev *dev,if (!attr->ingress)return 0 ;for (; actions->type !=RTE_FLOW_ACTION_TYPE_END ; actions++) { +if (actions->type ==RTE_FLOW_ACTION_TYPE_OF_PUSH_VLAN ) + push_vlan =1 ;switch (actions->type) {case RTE_FLOW_ACTION_TYPE_QUEUE : queue = actions->conf;
@@ -4623,11 +4626,15 @@ flow_check_hairpin_split(struct rte_eth_dev *dev,case RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP :case RTE_FLOW_ACTION_TYPE_NVGRE_ENCAP :case RTE_FLOW_ACTION_TYPE_OF_PUSH_VLAN : -case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_VID :case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_PCP : split++; action_n++;break ; +case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_VID : +if (push_vlan) + split++; + action_n++; +break ;case RTE_FLOW_ACTION_TYPE_RAW_ENCAP : raw_encap = actions->conf;if (raw_encap->size >MLX5_ENCAPSULATION_DECISION_SIZE )
@@ -5088,19 +5095,32 @@ flow_hairpin_split(struct rte_eth_dev *dev,struct mlx5_rte_flow_item_tag *tag_item;struct rte_flow_item *item;char *addr; +int push_vlan =0 ;int encap =0 ;for (; actions->type !=RTE_FLOW_ACTION_TYPE_END ; actions++) { +if (actions->type ==RTE_FLOW_ACTION_TYPE_OF_PUSH_VLAN ) + push_vlan =1 ;switch (actions->type) {case RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP :case RTE_FLOW_ACTION_TYPE_NVGRE_ENCAP :case RTE_FLOW_ACTION_TYPE_OF_PUSH_VLAN : -case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_VID :case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_PCP :rte_memcpy (actions_tx, actions,sizeof (struct rte_flow_action )); actions_tx++;break ; +case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_VID : +if (push_vlan) { +rte_memcpy (actions_tx, actions, +sizeof (struct rte_flow_action )); + actions_tx++; + }else { +rte_memcpy (actions_rx, actions, +sizeof (struct rte_flow_action )); + actions_rx++; + } +break ;case RTE_FLOW_ACTION_TYPE_COUNT :if (encap) { rte_memcpy(actions_tx, actions,
Signature · 1 line
2.25.1
Hi,
Quoted 41 lines
-----Original Message-----
From: Dariusz Sosnowski <[email protected]>
Sent: Monday, November 7, 2022 12:29 PM
To: Matan Azrad <[email protected]>; Slava Ovsiienko <[email protected]>
Cc: [email protected]; Raslan Darawsheh <[email protected]>; [email protected]
Subject: [PATCH] net/mlx5: fix hairpin split with set VLAN VID actionBefore this patch any flow rule which works on hairpin queues and which has
OF_SET_VLAN_VIDaction was split into 2 flow rules:- one subflow for Rx,
- one subflow for Tx.
OF_SET_VLAN_VIDaction was always placed in the Tx subflow.Assuming a flow rule which matches VLAN traffic and has both
OF_SET_VLAN_VIDaction, andMODIFY_FIELDaction on VLAN VID, but noOF_PUSH_VLANaction, the following happened:-
MODIFY_FIELDaction was placed in Rx subflow,
-OF_SET_VLAN_VIDaction was placed in Tx subflow,
-OF_SET_VLAN_VIDaction is internally compiled to a header modify command.This caused the following issues:
1. Since
OF_SET_VLAN_VIDwas placed in Tx subflow, 2 header modify actions were allocated. One for Rx and one for Tx.
2. IfOF_SET_VLAN_VIDaction was placed beforeMODIFY_FIELDon VLAN VID,
the flow rule executed header modifications in reverse order.MODIFY_FIELDactions were executed first in the Rx subflow andOF_SET_VLAN_VIDwas executed second in Tx subflow.This patch fixes this behavior by not splitting hairpin flow rules if
OF_SET_VLAN_VIDaction is used withoutOF_PUSH_VLAN.
On top of that, if flow rule is split, theOF_SET_VLAN_VIDaction is not moved to Tx subflow (for flow rules mentioned above).
Quoted 5 lines
Fixes 210008309b45 ("net/mlx5: fix VLAN push action on hairpin queue")Signed-off-by Dariusz Sosnowski <[email protected]>
Patch applied to next-net-mlx,
Kindest regards,
Raslan Darawsheh