Qdisc step on ingress is incorrect

[shemminger]

The step about qdisc on ingress side is incorrect. By default there is no ingress queue discipline.

It is possible to add ingress qdisc, but it is not the default qdisc. The ingress qdisc is special, it never queues packet it only does actions like policing or redirect.

[leandromoreira]

Hi there @shemminger thank you for pointing this out. Is it fair to assume that the default behavior is pfifo_fast for the default_qdisc? I'm thinking about ways to change to text and graphic to be more accurate.

source: https://www.kernel.org/doc/Documentation/sysctl/net.txt

[shemminger]

Egress qdisc has an actual packet queue, and the choice of qdisc determines how it schedules those packets. See pfifo, sfq, codel, etc.

Ingress qdisc has no actual queue of packets. It is only useful as a place to apply filtering rules. By default the ingress qdisc is not enabled. When it is enabled, some options are to police which limits traffic to a certain maximum. Packets over that max rate get dropped.

Another ingress usage is to forward the packets to a pseudo device call ifb. With ifb the ingress packets become egress packets so more complex egress qdisc can be applied.

Lots of details here: https://tldp.org/HOWTO/html_single/Traffic-Control-HOWTO/

[leandromoreira]

Thanks @shemminger :) do you mind to review my PR in the future? (when I create one to solve this)

[shemminger]

Sure, just add me as reviewer if you want