This is a set of utilities for Linux networking.

Information:
    http://www.linuxfoundation.org/collaborate/workgroups/networking/iproute2

Download:
    http://devresources.linuxfoundation.org/dev/iproute2/download

Repository:
    git://git.kernel.org/pub/scm/linux/kernel/git/shemminger/iproute2.git

How to compile this.
--------------------
1. libdbm

arpd needs to have the db4 development libraries. For Debian
users this is the package with a name like libdb4.x-dev.
DBM_INCLUDE points to the directory with db_185.h which
is the include file used by arpd to get to the old format Berkeley
database routines.  Often this is in the db-devel package.

2. make

The makefile will automatically build a Config file which
contains whether or not ATM is available, etc.

3. To make documentation, cd to doc/ directory , then
   look at start of Makefile and set correct values for
   PAGESIZE=a4		, ie: a4 , letter ...	(string)
   PAGESPERPAGE=2	, ie: 1 , 2 ...		(numeric)
   and make there. It assumes, that latex, dvips and psnup
   are in your path.

4. This package includes matching sanitized kernel headers because
   the build environment may not have up to date versions. See Makefile
   if you have special requirements and need to point at different
   kernel include files.

Stephen Hemminger
shemminger@linux-foundation.org

Alexey Kuznetsov
kuznet@ms2.inr.ac.ru

Here are a few quick points about DECnet support...

 o iproute2 is the tool of choice for configuring the DECnet support for
   Linux. For many features, it is the only tool which can be used to
   configure them.

 o No name resolution is available as yet, all addresses must be
   entered numerically.

 o Remember to set the hardware address of the interface using:

   ip link set ethX address xx:xx:xx:xx:xx:xx
      (where xx:xx:xx:xx:xx:xx is the MAC address for your DECnet node
       address)

   if your Ethernet card won't listen to more than one unicast
   mac address at once. If the Linux DECnet stack doesn't talk to
   any other DECnet nodes, then check this with tcpdump and if its
   a problem, change the mac address (but do this _before_ starting
   any other network protocol on the interface)

 o Whilst you can use ip addr add to add more than one DECnet address to an
   interface, don't expect addresses which are not the same as the
   kernels node address to work properly with 2.4 kernels. This should
   be fine with 2.6 kernels as the routing code has been extensively
   modified and improved.

 o The DECnet support is currently self contained. It does not depend on
   the libdnet library.

Steve Whitehouse <steve@chygwyn.com>

Iproute2 development is closely tied to Linux kernel networking
development. Most new features require a kernel and a utility component.

Please submit both to the Linux networking mailing list
   <netdev@vger.kernel.org>

The current source is in the git repository:
    git://git.kernel.org/pub/scm/linux/kernel/git/shemminger/iproute2.git

The master branch contains the source corresponding to the current
code in the mainline Linux kernel (ie follows Linus). The net-next
branch is a temporary branch that tracks the code intended for the
next release; it corresponds with networking development branch in
the kernel.

I. About the distribution tables

The table used for "synthesizing" the distribution is essentially a scaled,
translated, inverse to the cumulative distribution function.

Here's how to think about it: Let F() be the cumulative distribution
function for a probability distribution X.  We'll assume we've scaled
things so that X has mean 0 and standard deviation 1, though that's not
so important here.  Then:

	F(x) = P(X <= x) = \int_{-inf}^x f

where f is the probability density function.

F is monotonically increasing, so has an inverse function G, with range
0 to 1.  Here, G(t) = the x such that P(X <= x) = t.  (In general, G may
have singularities if X has point masses, i.e., points x such that
P(X = x) > 0.)

Now we create a tabular representation of G as follows:  Choose some table
size N, and for the ith entry, put in G(i/N).  Let's call this table T.

The claim now is, I can create a (discrete) random variable Y whose
distribution has the same approximate "shape" as X, simply by letting
Y = T(U), where U is a discrete uniform random variable with range 1 to N.
To see this, it's enough to show that Y's cumulative distribution function,
(let's call it H), is a discrete approximation to F.  But

	H(x) = P(Y <= x)
	     = (# of entries in T <= x) / N   -- as Y chosen uniformly from T
	     = i/N, where i is the largest integer such that G(i/N) <= x
	     = i/N, where i is the largest integer such that i/N <= F(x)
	     		-- since G and F are inverse functions (and F is
	     		   increasing)
	     = floor(N*F(x))/N

as desired.

II. How to create distribution tables (in theory)

How can we create this table in practice? In some cases, F may have a
simple expression which allows evaluating its inverse directly.  The
Pareto distribution is one example of this.  In other cases, and
especially for matching an experimentally observed distribution, it's
easiest simply to create a table for F and "invert" it.  Here, we give
a concrete example, namely how the new "experimental" distribution was
created.

1. Collect enough data points to characterize the distribution.  Here, I
collected 25,000 "ping" roundtrip times to a "distant" point (time.nist.gov).
That's far more data than is really necessary, but it was fairly painless to
collect it, so...

2. Normalize the data so that it has mean 0 and standard deviation 1.

3. Determine the cumulative distribution.  The code I wrote creates a table
covering the range -10 to +10, with granularity .00005.  Obviously, this
is absurdly over-precise, but since it's a one-time only computation, I
figured it hardly mattered.

4. Invert the table: for each table entry F(x) = y, make the y*TABLESIZE
(here, 4096) entry be x*TABLEFACTOR (here, 8192).  This creates a table
for the ("normalized") inverse of size TABLESIZE, covering its domain 0
to 1 with granularity 1/TABLESIZE.  Note that even with the granularity
used in creating the table for F, it's possible not all the entries in
the table for G will be filled in.  So, make a pass through the
inverse's table, filling in any missing entries by linear interpolation.

III. How to create distribution tables (in practice)

If you want to do all this yourself, I've provided several tools to help:

1. maketable does the steps 2-4 above, and then generates the appropriate
header file.  So if you have your own time distribution, you can generate
the header simply by:

	maketable < time.values > header.h

2. As explained in the other README file, the somewhat sleazy way I have
of generating correlated values needs correction.  You can generate your
own correction tables by compiling makesigtable and makemutable with
your header file.  Check the Makefile to see how this is done.

3. Warning: maketable, makesigtable and especially makemutable do
enormous amounts of floating point arithmetic.  Don't try running
these on an old 486.  (NIST Net itself will run fine on such a
system, since in operation, it just needs to do a few simple integral
calculations.  But getting there takes some work.)

4. The tables produced are all normalized for mean 0 and standard
deviation 1.  How do you know what values to use for real?  Here, I've
provided a simple "stats" utility.  Give it a series of floating point
values, and it will return their mean (mu), standard deviation (sigma),
and correlation coefficient (rho).  You can then plug these values
directly into NIST Net.

iproute2+tc*

It's the first release of Linux traffic control engine.


NOTES.
* csz scheduler is inoperational at the moment, and probably
  never will be repaired but replaced with h-pfq scheduler.
* To use "fw" classifier you will need ipfwchains patch.
* No manual available. Ask me, if you have problems (only try to guess
  answer yourself at first 8)).


Micro-manual how to start it the first time
-------------------------------------------

A. Attach CBQ to eth1:

tc qdisc add dev eth1 root handle 1: cbq bandwidth 10Mbit allot 1514 cell 8 \
avpkt 1000 mpu 64

B. Add root class:

tc class add dev eth1 parent 1:0 classid 1:1 cbq bandwidth 10Mbit rate 10Mbit \
allot 1514 cell 8 weight 1Mbit prio 8 maxburst 20 avpkt 1000

C. Add default interactive class:

tc class add dev eth1 parent 1:1 classid 1:2 cbq bandwidth 10Mbit rate 1Mbit \
allot 1514 cell 8 weight 100Kbit prio 3 maxburst 20 avpkt 1000 split 1:0 \
defmap c0

D. Add default class:

tc class add dev eth1 parent 1:1 classid 1:3 cbq bandwidth 10Mbit rate 8Mbit \
allot 1514 cell 8 weight 800Kbit prio 7 maxburst 20 avpkt 1000 split 1:0 \
defmap 3f

etc. etc. etc. Well, it is enough to start 8) The rest can be guessed 8)
Look also at more elaborated example, ready to start rsvpd,
in rsvp/cbqinit.eth1.


Terminology and advices about setting CBQ parameters may be found in Sally Floyd
papers.


Pairs X:Y are class handles, X:0 are qdisc handles.
weight should be proportional to rate for leaf classes
(I choosed it ten times less, but it is not necessary)

defmap is bitmap of logical priorities served by this class.

E. Another qdiscs are simpler. F.e. let's join TBF on class 1:2

tc qdisc add dev eth1 parent 1:2 tbf rate 64Kbit buffer 5Kb/8 limit 10Kb

F. Look at all that we created:

tc qdisc ls dev eth1
tc class ls dev eth1

G. Install "route" classifier on root of cbq and map destination from realm
1 to class 1:2

tc filter add dev eth1 parent 1:0 protocol ip prio 100 route to 1 classid 1:2

H. Assign routes to 10.11.12.0/24 to realm 1

ip route add 10.11.12.0/24 dev eth1 via whatever realm 1

etc. The same thing can be made with rules.
I still did not test ipchains, but they should work too.

Setup of rsvp and u32 classifiers is more hairy.
If you read RSVP specs, you will understand how rsvp classifier
works easily. What's about u32... That's example:



#! /bin/sh

TC=/home/root/tc

# Setup classifier root on eth1 root (it is cbq)
$TC filter add dev eth1 parent 1:0 prio 5 protocol ip u32

# Create hash table of 256 slots with ID 1:
$TC filter add dev eth1 parent 1:0 prio 5 handle 1: u32 divisor 256

# Add to 6th slot of hash table rule to select tcp/telnet to 193.233.7.75
# direct it to class 1:4 and prescribe to fall to best effort,
# if traffic violate TBF (32kbit,5K)
$TC filter add dev eth1 parent 1:0 prio 5 u32 ht 1:6: \
	match ip dst 193.233.7.75 \
	match tcp dst 0x17 0xffff \
	flowid 1:4 \
	police rate 32kbit buffer 5kb/8 mpu 64 mtu 1514 index 1

# Add to 1th slot of hash table rule to select icmp to 193.233.7.75
# direct it to class 1:4 and prescribe to fall to best effort,
# if traffic violate TBF (10kbit,5K)
$TC filter add dev eth1 parent 1:0 prio 5 u32 ht 1:: \
	sample ip protocol 1 0xff \
	match ip dst 193.233.7.75 \
	flowid 1:4 \
	police rate 10kbit buffer 5kb/8 mpu 64 mtu 1514 index 2

# Lookup hash table, if it is not fragmented frame
# Use protocol as hash key
$TC filter add dev eth1 parent 1:0 prio 5 handle ::1 u32 ht 800:: \
	match ip nofrag \
	offset mask 0x0F00 shift 6 \
	hashkey mask 0x00ff0000 at 8 \
	link 1:


Alexey Kuznetsov
kuznet@ms2.inr.ac.ru

lnstat - linux networking statistics
(C) 2004 Harald Welte <laforge@gnumonks.org
======================================================================

This tool is a generalized and more feature-complete replacement for the old
'rtstat' program.

In addition to routing cache statistics, it supports any kind of statistics
the linux kernel exports via a file in /proc/net/stat.  In a stock 2.6.9
kernel, this is
	per-protocol neighbour cache statistics
		(ipv4, ipv6, atm, decnet)
	routing cache statistics
		(ipv4)
	connection tracking statistics
		(ipv4)

Please note that lnstat will adopt to any additional statistics that might be
added to the kernel at some later point

I personally always like examples more than any reference documentation, so I
list the following examples.  If somebody wants to do a manpage, feel free
to send me a patch :)

EXAMPLES:

In order to get a list of supported statistics files, you can run

	lnstat -d

It will display something like

/proc/net/stat/arp_cache:
         1: entries
         2: allocs
         3: destroys
[...]
/proc/net/stat/rt_cache:
         1: entries
         2: in_hit
         3: in_slow_tot

You can now select the files/keys you are interested by something like

	lnstat -k arp_cache:entries,rt_cache:in_hit,arp_cache:destroys

arp_cach|rt_cache|arp_cach|
 entries|  in_hit|destroys|
       6|       6|       0|
       6|       0|       0|
       6|       2|       0|


You can specify the interval (e.g. 10 seconds) by:

	lnstat -i 10

You can specify to only use one particular statistics file:

	lnstat -f ip_conntrack

You can specify individual field widths

	lnstat -k arp_cache:entries,rt_cache:entries -w 20,8

You can specify not to print a header at all

	lnstat -s 0

You can specify to print a header only at start of the program

	lnstat -s 1

You can specify to print a header at start and every 20 lines:

	lnstat -s 20

You can specify the number of samples you want to take (e.g. 5):

	lnstat -c 5