Skip to content

Latest commit

 

History

History
171 lines (130 loc) · 7.85 KB

priority-translation.rst

File metadata and controls

171 lines (130 loc) · 7.85 KB
Raw

VLAN and Priority Translation

The translation process is necessary to reserve Linux socket priority in the network during hop-to-top data transmission. The socket priority is only visible within a Linux system. When a packet goes outbound from a network device, the priority number is translated and is recorded in the packet header. The header can be either a PCP field in VLAN or the ToS field in IPv4. In this article, the VLAN PCP is used.

A VLAN Network Example

To enable VLAN PCP tagging, a virtual VLAN device is created atop of a physical device with an assigned IP address. Packets going out from the VLAN device are encapsulated with a VLAN header. Let's start with a network of Alice and Bob.

.. mermaid::

   flowchart TB
       subgraph peer2 ["Bob"]
           direction TB

           peer2-vlan2["vlan2
           192.168.2.2"]
           peer2-eth0["eth0
           10.8.0.2"]

           peer2-vlan2 --- peer2-eth0
       end

       subgraph peer1 ["Alice"]
           direction TB

           peer1-vlan2["vlan2
           192.168.2.1"]
           peer1-eno1["eno1
           10.8.0.1"]

           peer1-vlan2 --- peer1-eno1
       end

       phy("Local Network
       10.8.0.0/24")

       phy --- peer2-eth0
       phy --- peer1-eno1

In the following sections, we will go through these steps to enable VLAN tagging.

  1. Identify the backend physical device.
  2. Create a VLAN device.
  3. Configure the priority mapping.

Identify the Backend Physical Device

Alice and Bob are connected through a local Ethernet network. Run ip address on Alice and Bob sides and you'll see this.

On Alice's computer,

1: eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether ff:ff:ff:ff:ff:ff brd ff:ff:ff:ff:ff:ff
    inet 10.8.0.1/24 brd 10.8.0.255 scope global dynamic noprefixroute eno1
       valid_lft 5300sec preferred_lft 5300sec

On Bob's computer,

1: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether ff:ff:ff:ff:ff:ff brd ff:ff:ff:ff:ff:ff
    inet 10.8.0.2/24 brd 10.8.0.255 scope global dynamic noprefixroute eno1
       valid_lft 5300sec preferred_lft 5300sec

Both of them physically connected on a local network, prefixed with IP 10.8.0.x. Now, we'll start a virtual local-area network (VLAN), prefixed with 192.168.2.x. It works as though Alice and Bob are connected through a virtual hub. To do so, we'll create a virtual network device named vlan2 on both sides. It's named vlan2 because they share the VLAN ID 2.

Create VLAN Devices

On the Alice's side, run sudo nmtui → Edit a connection → Add → Select VLAN and click Create. Fill in the form like this. Bob will do so similarly but with a different IP.

┌───────────────────────────┤ Edit Connection ├───────────────────────────┐
│                                                                         │
│         Profile name VLAN2___________________________________           │
│               Device vlan2___________________________________           │
│                                                                         │
│ ╤ VLAN                                                        <Hide>    │
│ │             Parent eno1____________________________________           │
│ │            VLAN id 2_______                                           │
│ │                                                                       │
│ │ Cloned MAC address ________________________________________           │
│ │                MTU __________ bytes                                   │
│ └                                                                       │
│                                                                         │
│ ╤ IPv4 CONFIGURATION <Manual>                                 <Hide>    │
│ │          Addresses 192.168.2.1/24___________ <Remove>                 │
│ │                    <Add...>                                           │
│ │            Gateway _________________________                          │
│ │        DNS servers <Add...>                                           │
│ │     Search domains <Add...>                                           │
│ │                                                                       │
│ │            Routing (No custom routes) <Edit...>                       │
│ │ [X] Never use this network for default route                          │
│ │ [ ] Ignore automatically obtained routes                              │
│ │ [ ] Ignore automatically obtained DNS parameters                      │
│ │                                                                       │
│ │ [ ] Require IPv4 addressing for this connection                       │
│ └                                                                       │
│                                                                         │
│ ═ IPv6 CONFIGURATION <Automatic>                              <Show>    │
│                                                                         │
│ [X] Automatically connect                                               │
│ [X] Available to all users                                              │
│                                                                         │
│                                                           <Cancel> <OK> │
└─────────────────────────────────────────────────────────────────────────┘

Let's ip address again. You can see the newly created network device on the Alice's side.

2: vlan2@eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
    link/ether ff:ff:ff:ff:ff:ff brd ff:ff:ff:ff:ff:ff
    inet 192.168.2.1/24 brd 192.168.2.255 scope global noprefixroute vlan2
       valid_lft forever preferred_lft forever

Configure the Priority Mapping

Now here comes the crucial part. We have to tell the vlan2 interface to map socket priority 0 to PCP 0, socket priority 1 to PCP 1, and so on. The command goes like this. The 0:0 term means mapping the socket priority 0 on the left of colon to PCP 0 on the right.

sudo ip link \
    set dev vlan2 \
    type vlan \
    egress 0:0 1:1 2:2 3:3 4:4 5:5 6:6 7:7

To testify our settings, we run the socket priority testing program from the previous section and set the sender priority to 6. Run the tcpdump packet sniffer and packets captured will show that PCP=6. (default value of pcp is 0)

18:06:30.891525 08:26:97:f7:49:c5 (oui Unknown) > 08:26:97:f7:49:c9 (oui Unknown), ethertype 802.1Q (0x8100), length 2966: vlan 1, p 6, ethertype IPv4 (0x0800), (tos 0x0, ttl 64, id 23093, offset 0, flags [DF], proto TCP (6), length 2948)
    192.168.1.2.36196 > ros-RSK.55555: Flags [P.], cksum 0x8eca (incorrect -> 0x27f2), seq 76361265:76364161, ack 0, win 502, options [nop,nop,TS val 3959431146 ecr 1510782482], length 2896

A Little Summary

This section brought you to set up a VLAN device and enable the priority mapping. That's the first step to dive into Time-Sensitive network. In later parts, we will create scheduling rules making use of these priorities. TAPRIO will be our first example.