Using switch to extend number of network ports in host processor

Hi Harrison,

The Logical MAC are primarily validated for Layer 2 traffic.

Applying ingress tag for Layer 3 traffic can still work, although this behavior has not been extensively tested.

Could you confirm if I understand that the main goal is to expand network ports for the host? Is tagging ingress traffic a strict requirement?

This will help me recommend other features for your use case.

Thanks,

Karl

Hi Karl,

Thanks for getting back to me.

That is correct, the main goal is to expand the network ports for the host. Tagging is not a strict requirement, only that the network traffic from the ports should be completely isolated from each other on both ingress and egress.

I have looked into using VLAN for this but we currently have customers who use VLANs with our device and and from my understanding the switch can only tag untagged packets.

Thanks 

I think you can use the forward mask (SES_ApplyForwardMask) for the network ports.

You can configure the API so that incoming frames on the extended network ports to only forwarded to the host port.

Could you share more details about your system?

This will help me better understand your requirements

Also, are you using extended table to route L3 traffic?

I think you can use the forward mask (SES_ApplyForwardMask) for the network ports.

You can configure the API so that incoming frames on the extended network ports to only forwarded to the host port.

Could you share more details about your system?

This will help me better understand your requirements

Also, are you using extended table to route L3 traffic?

From my understanding using the forward mask would not add a tag to the end of the packet meaning the host processor couldn't differentiate which network port on the switch the packet originated from.

Currently I am only using Logical mac groups and static table entries, on ingress a tail is added to both L2 and L3 traffic so that host processor can differentiate which port the packet originated from. On egress, the source mac address is changed to ensure the packet is egressed from a specific network port on the switch.

Could you please provide an example of how to setup the extended table to route L3 traffic such that it adds a tag to the packet on ingress and is there a method to egress packets on specific network ports only? 

To elaborate on our system and requirements,

SYSTEM
We have a dual core arm processor running Linux which is directly connected to the ADIN6310 switch in a MAC-MAC connection over SGMII. The host port on the switch is port 1 and we have 3 external network ports on our custom board, 2 SFPs and 1 LAN port.

Requirements 
All network traffic is ingressing or egressing from the host processor. 
The network traffic ingressing from each of the 3 network ports must be isolated from each other and once it reaches the host processor, there must be a way to determine which network port the packet originated from (both L2 & L3) so the host processor can process these network streams independently.
When transmitting packets from the host processor, we must be able to egress the packets out of specific network ports (both L2 & L3 traffic) on the switch.

Hi Karl,

Just following up on this thread to see if you have found any more information for my use case.

Thank you

Hi Harrison,

I've attached an example configuration.

This approach is similar to what you mentioned about adding a static table entry and passing 0s in the address mask.

In the example, traffic ingressing on Port 1 or Port 2 is set to be forwarded to the link port (Port 0) with a tail tag.

int main(int argc, char* argv[])
{
	printf("\n--START--\n");
	int32_t rv = SES_PORT_ERROR;
	rv = SES_Initialize();
	rv = SES_FirmwareCheck();

	Sleep(5000); // Wait for PHYs to link up

	memcpy(&logicalMacConfig1.logicalMac, logicalMac1, 6);
	logicalMacConfig1.portMap = 0x01 | 0x02; // Port 0 and Port 1
	logicalMacConfig1.deviceLinkPort = 0; // Link Port - Port 0
	logicalMacConfig1.rxLogicalMacCallback = NULL;
	rv = SES_AddLogicalMac(&logicalMacConfig1, &logicalMacReturns1);
	printf("Added Logical MAC 1 : %d\n", rv);

	memcpy(&logicalMacConfig2.logicalMac, logicalMac2, 6);
	logicalMacConfig2.portMap = 0x01 | 0x04; // Port 0 and Port 2
	logicalMacConfig2.deviceLinkPort = 0; // Link Port - Port 0
	logicalMacConfig2.rxLogicalMacCallback = NULL;
	rv = SES_AddLogicalMac(&logicalMacConfig2, &logicalMacReturns2);
	printf("Added Logical MAC 2 : %d\n", rv);

	uint8_t macAddr[6] = { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 }; //sample mac address
    uint8_t macMask[6] = { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 }; //wildcard mask

	rv = SES_AddStaticTableEntryEx(0, 			// lookup priority
        &macAddr, 								// mac address
        &macMask,                     			// mac mask
        SES_DYNTBL_NO_VLAN,          			// vlan id
        SES_DYNTBL_VLAN_EXACT_MATCH, 			// vlan mask
        0,                           			// source entry
        0,                           			// override
        0,                           			// source override
        0,                           			// hsr prp supervisory
		SES_dynamicTblLookupBasic, 				// lookup type
        0x01, 									// port map
        0,										// send to ae
		0,										// danp
        0,										// cut through -> store and forward operation
        0,										// synt timestamp
        0,										// local timestamp
        dynamicTblSequenceGeneration,  			// sequencemgmt
        0,										// rx sequence set
        logicalMacReturns1.txPrpXformIndex,     // transmit filter
        SES_IGNORE_FIELD,            // receive filter
        NULL);                       // index*
	printf("Added Static Table Entry : %d\n", rv);

	printf("\n--END--\n");

	while(1) {

	}
	return(0);
}

Let me know if you'd like more details on the setup.

Regards,

Karl