[dpdk-dev] Issues with ixgbe and rte_flow

[Lu, Wenzhuo]

Hi Le Scouarnec,

> -----Original Message-----
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Le Scouarnec Nicolas
> Sent: Tuesday, March 7, 2017 7:12 PM
> To: dev at dpdk.org
> Cc: Yigit, Ferruh
> Subject: [dpdk-dev] Issues with ixgbe and rte_flow
> 
> Dear all,
> 
> I have been using the new API rte_flow to program filtering on an X540 (ixgbe)
> NIC. My goal is to send packets from different VLANs to different queues
> (filtering which should be supported by flow director as far as I understand). I
> enclosed the setup code at the bottom of this email. However, the code in ixgbe
> rejects the request I formulated. I looked at the rte_flow related code in ixgbe
> and I found some weird checkings in function ixgbe_parse_fdir_filter_normal.
> The code verifies that vlan_spec->tpid == ETHER_TYPE_VLAN, which should
> never happen on real packets, as it is eth_spec->type which has this value in
> typical setups. The same comment applies to the mask, only eth_mask->tpid
> should be required to be 0xffff. Overall, unless I misunderstood the rte_flow API,
> the checks should be done on eth_spec/eth_mask rather than
> vlan_spec/vlan_mask.
> 
> I also have a side comment which might be more related to the general rte_flow
> API than to the specific implementation in ixgbe. The rte_flow_error returned is
> not very useful for it does return the error of the last tried filter-type (L2 tunnel
> in ixgbe), and not the error of the filter-type that my setup should use (flow
> director). As NICs can have several filter-types, It would be be useful to the user
> if rte_flow_validate/create could return the errors for all filter types tried
> although that would require to change the rte_flow API and returning an array
> of rte_flow_error and not a single struct.
> 
> Excerpt of the code that reject my filtering rule:
> 
>               if (vlan_spec->tpid != rte_cpu_to_be_16(ETHER_TYPE_VLAN)) {
> 			memset(rule, 0, sizeof(struct ixgbe_fdir_rule));
> 			rte_flow_error_set(error, EINVAL,
> 				RTE_FLOW_ERROR_TYPE_ITEM,
> 				item, "Not supported by fdir filter");
> 			return -rte_errno;
> 		}
> 
> 
> 		rule->ixgbe_fdir.formatted.vlan_id = vlan_spec->tci;
> 
> 
> 		if (vlan_mask->tpid != (uint16_t)~0U) {
> 			memset(rule, 0, sizeof(struct ixgbe_fdir_rule));
> 			rte_flow_error_set(error, EINVAL,
> 				RTE_FLOW_ERROR_TYPE_ITEM,
> 				item, "Not supported by fdir filter");
> 			return -rte_errno;
> 		}
> 
> For reference, here is the setup code I use
> 
> 	struct rte_flow_attr attr;
> 	attr.group = 0;
> 	attr.priority = 0;
> 	attr.egress = 0;
> 	attr.ingress = 1;
> 
> 
> 	struct rte_flow_item_eth eth_mask;
> 	struct rte_flow_item_eth eth_spec;
> 	struct rte_flow_item_vlan vlan_mask;
> 	struct rte_flow_item_vlan vlan_spec;
> 	struct rte_flow_action_queue queue;
> 
> 
> 	struct rte_flow_item items[3];
> 	items[0].type = RTE_FLOW_ITEM_TYPE_ETH;
> 	items[1].type = RTE_FLOW_ITEM_TYPE_VLAN;
> 	items[0].last = NULL;
> 	items[0].spec = &eth_spec;
> 	items[0].mask = &eth_mask;
> 	items[1].last = NULL;
> 	items[1].spec = &vlan_spec;
> 	items[1].mask = &vlan_mask;
> 	items[2].type = RTE_FLOW_ITEM_TYPE_END;
> 
> 
> 	eth_mask.src = ETH(00,00,00,00,00,00);
> 	eth_mask.dst = ETH(ff,ff,ff,ff,ff,ff);
> 	eth_mask.type = 0xffff;
> 
> 
> 	eth_spec.src = ETH(00,00,00,00,00,00);
> 	eth_spec.dst = *dst;
> 	eth_spec.type = rte_cpu_to_be_16(ETHER_TYPE_VLAN);
> 
> 
> 	vlan_spec.tci = vlan_be;
> 	vlan_spec.tpid = 0;
> 
> 
> 	vlan_mask.tci = rte_cpu_to_be_16(0x0fff);
> 	vlan_mask.tpid =  0;
To my opinion, this setting is not right. As we know, vlan tag is inserted between MAC source address and Ether type.
So if we have a MAC+VLAN+IPv4 packet, the vlan_spec.tpid should be 0x8100, the eth_spec.type should be 0x0800.
+ Adrien, the author. He can correct me if I'm wrong.