freebsd-skq/share/examples/ppp/ppp.conf.span-isp
1999-09-07 07:56:54 +00:00

193 lines
6.6 KiB
Plaintext

# $FreeBSD$
# This advanced ppp configuration file explains how to implement
# the following:
#
# ------------- ------------- -------------
# | host1 | | host2 | | host3 |
# ------------- ------------- -------------
# | | |
# |---------------------- LAN ----------------------|
# |
# -------------
# | Gateway |
# -------------
# |
# -----------------------------------
# | | | |
# isp1 isp2 isp3 ispN
# | | | |
# -----------------------------------
# |
# ------------
# | Receiver |
# ------------
# |
# Internet
#
# The connection is implemented so that any ISP connection can go down
# without loss of connectivity between the LAN and the Internet. It is
# of course also possible to shut down any link manually.
#
# There is a working example in ppp.*.span-isp.working that can be tested
# on a single machine !
#
#
# Prerequisites:
#
# o The Receiver machine must be in the outside world and must be willing
# to accept a multilink ppp connection over UDP, assigning a routable IP
# number to the Gateway machine. This probably means that it must be
# a *BSD box as I know of no other ppp implementations that can use UDP
# as a transport.
#
# o The Receiver machine must be multi-homed with at least N+1 addresses
# where N is the maximun number of ISPs that you wish to use
# simultaneously. We assume the IP numbers to be RIP1, RIP2 ... RIPN.
# REAL-LOCAL-IP is the real IP number of the Receiver machine (and must
# not be the same as any of the RIP* numbers).
#
# o Both the Gateway and the Receiver machines must have several tun
# interfaces configured into the kernel (see below).
#
# o Both the Gateway and the Receiver machines must have the following
# entry in /etc/services:
#
# ppp 6671/udp
#
# The port number isn't important, but it must be consistent across
# machines.
#
# o The Receiver machine must have the following entry in
# /etc/inetd.conf:
#
# ppp dgram udp wait root /usr/sbin/ppp ppp -direct vpn-in
#
# Note: Because inetd ``wait''s for ppp to finish, a single ppp
# invocation receives all incoming packets. This creates
# havoc with LQR magic number checks, so LQR *must not* be
# enabled.
# Also, -direct invocations of ppp do sendto()s using the
# address that was last recvfrom()d. This means that the
# returning traffic is a bit unbalanced. Perhaps ppp should
# be smart enough to automatically clone an existing link
# when it detects a new incoming address.... tricky !
#
# If you use ppp to connect to your ISPs, the isp* profiles shold be used,
# resulting in the vpn* profiles being called from ppp.linkup.span-isp.
# These invocations will bond together into a MP ppp invocation.
#
# If the link to your ISP is via another type of interface (cable modem
# etc), simply configure the interface with a netmask of 0xffffffff and
# add a route to RIPN via the interface address (no default). You can
# then start ppp using the vpn-nic label.
#
# The Receiver machine should have N tun interfaces (where N is the maximum
# number of ISPs that you wish to use simultaneously). The Gateway machine
# requires N interfaces plus an additional N interfaces (total 2 * N) if
# you're using ppp to talk to the ISPs.
# Using ppp to connect to your ISPs (PPP over UDP over PPP):
#
# When we connect to our ISPs using ppp, we start the MP ppp invocation
# from ppp.linkup (see ppp.linkup.span-isp) for each link. We also remove
# the link from ppp.linkdown (see ppp.linkdown.span-isp). This is necessary
# because relying on our LQR strategy (dropping the link after 5 missing
# replies) is just too slow to be practical in this environment.
#
# This works because the MP invocations are smart enough to recognise that
# another process is already running and to pass the link over to that
# running version.
#
# Only the ISP links should be started manually. When they come up, they'll
# start the MP invocation.
default:
set speed 115200
set device /dev/cuaa0 /dev/cuaa1 /dev/cuaa2 /dev/cuaa3
set dial "ABORT BUSY ABORT NO\\sCARRIER ABORT NO\\sDIAL\\sTONE TIMEOUT 4 \
\"\" ATZ OK-ATZ-OK ATDT\\T TIMEOUT 60 CONNECT \\c \\n"
set login
set redial 3 5
set timeout 0
enable lqr
set lqrperiod 15
isp1:
set phone "1234567"
set authname isp1name
set authkey isp1key
add! RIP1/32 HISADDR
isp2:
set phone "2345678"
set authname isp2name
set authkey isp2key
add! RIP2/32 HISADDR
ispN:
set phone "3456789"
set authname ispNname
set authkey ispNkey
add! RIPN/32 HISADDR
# Our MP version of ppp. vpn is a generic label used by each of the
# other vpn invocations by envoking ppp with both labels (see
# ppp.linkup.span-isp).
# Each ``set device'' command tells ppp to use UDP packets destined for
# the given IP/port as the link (transport). The routing table will
# ensure that these UDP packets use the correct ISP connection.
vpn:
set enddisc LABEL
set speed sync
set mrru 1500
nat enable yes
set authname vpnname
set authkey vpnkey
add! default HISADDR
disable deflate pred1 lqr
deny deflate pred1
vpn1:
rename 1
set device RIP1:ppp/udp
vpn2:
rename 2
set device RIP2:ppp/udp
vpnN:
rename N
set device RIPN:ppp/udp
vpn-nic:
load vpn
clone 1 2 N
link deflink rm
link 1 set device RIP1:ppp/udp
link 2 set device RIP2:ppp/udp
link N set device RIPN:ppp/udp
# The Receiver profile is a bit more straight forward, as it doesn't need
# to get bogged down with sublinks. Replace REAL-ASSIGNED-IP with the
# IP number to be assigned to the Gateway machine. Replace REAL-LOCAL-IP
# with the real IP number of the Receiver machine.
#
# No other entries are required on the Receiver machine, and this entry
# is not required on the Gateway machine. The Receiver machine also
# requires the contents of ppp.secret.span-isp.
#
# Of course it's simple to assign an IP block to the client with a simple
# ``add'' command, and then have the client use those IP numbers on its
# LAN rather than using ``nat enable yes''.
vpn-in:
set enddisc label
set speed sync
set mrru 1500
enable chap
disable lqr
set ifaddr REAL-LOCAL-IP REAL-ASSIGNED-IP