Hi,
Is there any documentation about the u32 classifiers except for the
source code? The exampels I have seen on it's usage (for example in
the paper "Quality of Service Support in Linux" by "Saravanan
Radhakrishnan") really looks like magic to me. As far as I have understood
a filter should do the following:
1) Look at the packet
2) Tell the qdisc it belongs to in which of it's classes it should put the
packet.
That is, it should be a mapping from packets to class ids. That sounds
easy enough... As I can figure out u32 makes it possible to setup a rule
for each class. That is, you can say, for example, that packets to this
and this IP address should go to this class, packets from that and that
TCP port should go to that class and so on.
Okay, that sounds just fine. But how does it work and how do I specify it?
I understand that u32 somehow uses a hash-tables and that I need to set
that one up. It is, however, not obvious to me how a hash table can be
used to solve this problem.
At the first glance one might thing that a hash-key is calculated from
the packet and then looked up in the hash table which then contains the
class that the packet should go to (or a list of possible classes). But
this is not possible because which parts of the packets that should
contribute to such a hashkey will be different for the different classes
(depending on weather it describes source IP or destination TCP port,
for example).
So what is the point? If someone can describe it so that I can understand
it I will be happy to make a general user-friendly description on how to
setup trafic-control. I think I already have understood alot of how the
trafic control works and I have written my own qdisc which also contains
classes that can be manipulated with tc. But it uses it's own build-in
filter.
Christian
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
