Stephen Hemminger wrote:
On Sun, 26 Apr 2009 07:26:16 -0400 (EDT)
"sting" <sting@xxxxxxxxxxxxx> wrote:
Obviously, traffic control and iptables operate at the IP layer.
But if I was to do any traffic shaping based on the known MAC address for
hosts that are on my subnets, what would make the most sense?
The u32 classifier and Qos work with MAC addresses. You just need
to manually give proper offsets and masks (or use latest iproute utilities,
I just added that enhancement).
Thank you, I've played around with it and it works (using offsets, not
your update to the iproute toolchain yet).
To ensure I'm not doing something grossly wrong, the idea is I want to
preserve the traffic shaping (for me, network simulation mostly) I've
been doing so far, but want to skip all of it for certain MAC addresses.
Would it be a bad idea to insert a new prio qdisc at root, filter all
traffic from desired MAC addresses to the first band/class, and filter
remaining traffic to the second band/class, with the second band/class
containing my original root queuing discipline?
There isn't a way around having to add one filter per MAC address,
correct? (not that I think this would be a problem)
Also, is it a bad idea to be using the prio qdisc for non-QoS purposes?
All I use it for is to selectively send traffic into further queuing
disciplines using filters. But if a packet happens not to match any of
the filters, will it then be sent to a specific band based on it's QoS
flag like it would if I had no filters? If so I could accidently get
packets queued where I don't intend to.
(In case you wonder certain game console devkits have two internal NICs
interfaced via a single ethernet jack, and when performing network
simulation I obviously don't want to affect deploy times to the target,
or interactivity with the debugger)
--
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
