DRA821U: Queries regarding the prioritization at switch

Part Number: DRA821U

Tool/software:

We have few questions regarding the Shaping of Traffic generated by NAD that is going to external network

Could you please confirm if our understanding is correct:

  • Could you please confirm if the above understanding is correct?
  • HQ queues configured on NAD recognizes the Traffic generated by NAD based on the PCP field set by NAD in the VLAN tag?
  • Does the TI driver automatically map HW quesues with PCP values or is there any specific configuration needed to acive this ( I mean, is it still possible to map PCP2 to a high priority queue)
  • Can the Traffic generated by NAD leaving BAM be routed to BAM kernel space or handled directly by switch ? This is needed because we need to conclude on the specific configuration needed to handle the Traffic handling on switch side 
  • Hi,

    Could you please confirm if the above understanding is correct?

    The only thing I would modify here is that the tc rate-limiting here is built into the egress queues and does not exist as a separate block in the hardware.

    HQ queues configured on NAD recognizes the Traffic generated by NAD based on the PCP field set by NAD in the VLAN tag?

    The ingress queue on socnet1 (there is only 1 Rx queue) will assign a switch priority to the packet based on the PCP value of VLAN tag. By default it is 1-1 mapped. While egressing, the switch priority will be used as egress priority. So in your case, the priority will be maintained for the packet going through the switch.

    Does the TI driver automatically map HW quesues with PCP values or is there any specific configuration needed to acive this ( I mean, is it still possible to map PCP2 to a high priority queue)

    It is possible, but it will require a patch. It is not configurable at runtime.

    Can the Traffic generated by NAD leaving BAM be routed to BAM kernel space or handled directly by switch ? This is needed because we need to conclude on the specific configuration needed to handle the Traffic handling on switch side 

    I didn't get what you are trying to do here. Can you explain a bit more here.

    Regards,
    Tanmay

  • Can the Traffic generated by NAD leaving BAM be routed to BAM kernel space or handled directly by switch ? This is needed because we need to conclude on the specific configuration needed to handle the Traffic handling on switch side 

    I didn't get what you are trying to do here. Can you explain a bit more here.

    Hi Tanmay,

    What i meant was If there is a EGGRESS TRAFFIC on BAM coming directly from NAD 

    For example. VLAN77 internet traffic :  WAN--> IPA/ Modem -->Eth0 (NAD) --> Port2( socnet1) --> port3 (socnet0) --> BAM

    Would the HTB qdesc applied on BMA kernel be applicable for such traffic? or they are directly prioritized by Tx queues based on PCP?

  • Hi Tanmay,

    As discussed over the call today,

    • It is confirmed that by default Tx queue size is 8 on external ports and no need to increase
    • On host port, we have an option to increase Tx queue size from 1 to 8 ( patch is available)
    • By default PCP is mapped 1:1 to Tx queue --> meaning highest PCP goes to highest queue 
    • However, this can be configured on host port to make sure certain PCP is given the highest priority
    • On external ports, a patch is needed to map PCPs to queues based on required priority

    The following is the expectation 

    • Confirm the feasibility of providing patch with above priorities.
    • Please let us know how to configure the PCP with priorities on host port 
    • Is it possible to give untagged packets highest prio? what is the default priority for untagged packets 
  • Hi Tanmay,

    After further discussion with BMW, the following priorities 

    Could you please provide a patch for the same:

    Priority PCP ( >=3) Applicable port Traffic type
    untagged All PTP/ EAPOL
    7 4 All Frame preemption
    6 3 All AVnu class A traffic, noise cancelling, warnings
    5 2 All Telephony, AbVnu class C, computer vision
    4 5 All future use
    3 7 All some/ip, npdu
    2 6 All future use 
    1 0 All Best effort (web, streaming, internet etc..)
    0 (Lowest) 1 All Lower effort , DLT, Debugging

    Also, as discussed kindly update us with the handling of untagged packets