Packet loss only on Tx using tc(Traffic Control) in Linux

I'm aware that one can introduce packet loss in a Linux machine using tc. For example:

tc qdisc change dev eth0 root netem loss 10%

However, is it possible to specify a traffic direction where the packet loss occurs? I would like to have no packet loss on receive, but 10% packet loss on transmit for testing purposes.

Best Answer

As far as I know, specifying a qdisc in the above way means an egress qdisc only. So you should only see 10% loss on egress traffic from eth0.

To do the same for ingress traffic you'd need to use something like an intermediate functional block pseudo device, where you redirect ingress traffic on eth0 to the egress queue of a pseudo device (e.g. ifb0) and shape/queue on that device. See here for more details http://www.linuxfoundation.org/collaborate/workgroups/networking/netem

Related Solutions

Linux Traffic Control: qdisc mq

A google search would have easily yielded this article from the vger kernel mailing list:

These patches contain a classful multiqueue ("mq") dummy scheduler to fix a couple of problems with the current multiqueue TC API integration. The changelogs of patch 05 and 07 contain more details.

The mq scheduler does two things:

present device TX queues as classes, allowing to attach different qdiscs to them, which are grafted to the TX queues

present accumulated statistics of all device queue root qdiscs

Its used by default for multiqueue devices instead of the regular pfifo_fast qdisc, but can also be attached manually to restore multiqueue behaviour after attaching a non-multiqueue (shared) qdisc.

Multiqueue is a feature is otherwise referenced as RSS (Receive-Side-Scaling) which basically is distributing the load of packet processing among several cores. Further reading: https://elixir.bootlin.com/linux/latest/source/Documentation/networking/scaling.rst

Iptables – How to add delay to incoming traffic

Adding delay can be done on incoming (ingress) or outgoing (egress) packets (or both). Delaying ingress packets is a bit harder (cause the packet has already arrived) but is achievable with Intermediate Functional Block device (have a look here. search for "netem on incoming traffic").

Because your goal can be achieved by delaying ingress or egress traffic and delaying egress is easier, I'd recommend to do that. Your configuration is almost correct: you just need to replace the iptables mangle line with

sudo iptables -t mangle -A OUTPUT -p tcp -m tcp --sport 80 -j MARK --set-mark 11

Best Answer

Related Solutions

Linux Traffic Control: qdisc mq

Iptables – How to add delay to incoming traffic

Related Topic