Linux-Networking Digest #230, Volume #12 Sun, 15 Aug 99 07:13:31 EDT
Contents:
Re: please help can't see the host when bridging (Vidar Andresen)
Linux vs NT speeds in Halloween document - TROLL - Re: Linux crashes all (David
Crooke)
Samba Setup (Bond Robbins)
Adding a Networking Card ("Suddn")
Re: Gateway telepath modem ("Edward M. Babb")
Samba, Windows Workstation hangs (Stefan Frings)
Mounting ZIP drive (Stefan Frings)
Re: Help on subnetting (David Crooke)
Linux and TCP data compression (Victor Vjunnik)
Re: Connecting Linux to a Mac network (David Crooke)
Re: socket++ for linux (Erwin Hogeweg)
Re: I'm lost in this Masquerade! ("Steve Cowles")
Re: I can't believe I've forgotten.... (David Crooke)
Re: Home network and cable modem (David Crooke)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Vidar Andresen)
Subject: Re: please help can't see the host when bridging
Date: Sun, 15 Aug 1999 09:57:40 GMT
In article <7p3mj7$ine$[EMAIL PROTECTED]>, [EMAIL PROTECTED] wrote:
>I have followed the bridging mini-howto word for word except i have
What kernel-version are you running? 2.2.*
Please look at Message-ID: <[EMAIL PROTECTED]>
Date: Tue, 10 Aug 1999 03:46:13 +0200
Subject: Re: Bridge using Linux
[...]
Dont know, dont run 2.2.* kernel. But.
http://support.broadjump.com/cerberus/ is a bridge
rewritten for 2.2.* I belive. Got some info.
There is two older versions of bridge out there, they
both have a script 'bridge' which is to be placed in
'/etc/rc.d/init.d/' and to be called as
'/etc/rc.d/init.d/bridge start' or
'/etc/rc.d/init.d/bridge stop'
[...]
>given one interface an ip. When I give both interfaces an IP only
>people on the hub connected to the interface which is brought up first
>can see the host but the rest of the network can talk easily. If i
>bring the interfaces up in a different order the problem reverses.
The host itself get a route out on the last interface activated, 'ifup
eth1' for that side.
>I have tried with both interfaces having and IP, on interface having an
>IP and none having an ip but giving one an alias. No matter what I seem
>to do only one half of the network can see the host at one time.
>
>The closest I have come to getting it to work is giving my interfaces
>192.168.1.1 and 192.168.2.1 and having the people on one side have an
>ip coresponding to the interface. This option is less desirable but
>also has problems. The people on the side of the interface brought up
>first can ping the host and everyone else but the people on the second
>interface can ping the host but not the people on the other side. I
>then changed the subnets on all windows machines to 255.255.0.0 and
>pinging worked fine, but they were unable to browse using smb on any
>computer with a different subnet. Upping the subnet on the linux host
>returns me to the original problem of only one side seeing the host.
If you need the host acting as a bridge to be abel to speak to hosts
on either side of the bridge, and answer to both sides,
simultaneously, you need to set routes to the hosts on both side.
As you would if no bridging where there.
But no need to do so if the machine is to act only as a bridge.
No real need to split up the sides of the bridge. (if there not is a
reason like need to minimize broadcast.)
If you have settled for a ip-range, let say 192.168.1.* on the whole
network, give your 1st nic, the eth0 the 192.168.1.2 address and the
eth1 the 192.168.1.3 address. In other words, give them real
adresses (whatever within range), and valid netmask.
If it is to bridge a lot of win-machines, the automagic (and old
correct netmask for the 192.168.*.*) is that a netmask of
255.255.255.0 is assigned (in win95 it is, if nothing chosen). To
save yourself trouble, use that.
(if you dont mean to do dns or being a gateway, redhat atutomagically
sets 192.168.1.1 and 192.168.1.254 for that (dont remember whats
what). And _stay_ _off_ 192.168.1.0 and 192.168.1.255. (Sorry, just
try to take everything into acount..;)
So start with 192.168.1.2 and up on the bridging machine.
(You could do the subnetting thing in the 'bridge-howto' or was it the
'bridge+ipfwadm-howto', I'm not into that.. And what you then get is
routing done manually.. take it or go for the subnetting-thing. If
you only need the bridge to expand a segment-lenght, 'switching'
(which is bridging..) or set togheter a 10 an 100Mbps net without a
hardware 'bridge' aka 'switch', I think this should do.)
The first thing you need is being abel to ping both ip-adresses on the
bridge from the bridge-machine itself. Without doing any 'ifup *' to
get contact with the spesific nic. This is from a 'route.do' I used
to put into the same script as the bridge, but it could also be done
before the bridge is started, that way the bridge host itself will
have a route out all ways, even if not bridging is up.
This is probably not needed, dont do any harm, from this on you shold
be abel to ping any of the (two) _internal_ interfaces:
route add default lo
route add 192.168.1.2 eth0
route add 192.168.1.3 eth1
#
#This make a route _out_ to other machines over given interface:
#
route add 192.168.1.4 eth0
route add 192.168.1.5 eth1
route add 192.168.1.6 eth0
[...]
Make 'that-list' as long as you like, save it to a file and
'chmod +x that-list'
Watch out for double routes.
If you start bridging from like say /etc/rc.d/rc.local
(the thing under is to complicated '/etc/rc.d/init.d/bridge start' is
enough, i use the 'bridge' script which follows the bridge packages,
and dont mess with 'brcfg -ena', place brcfg in /sbin/, the script
check for it, and set which interfaces to be bridged in the
/etc/rc.d/init.d/bridge script)
To complicated, but:
/etc/rc.d/init.d/bridge stop
/sbin/ifup eth1
/sbin/ifup eth0
/etc/rc.d/init.d/bridge start
/sbin/ifup eth1
/sbin/ifup eth0
And you could add at any side,
/sbin/route add default lo
/sbin/route add 192.168.1.2 eth0
/sbin/route add 192.168.1.3 eth1
/sbin/route add 192.168.1.4 eth0
/sbin/route add 192.168.1.5 eth1
/sbin/route add 192.168.1.6 eth0
Or simply
/path-to/that-list-mentioned-above
To execute the list from /etc/rc.d/rc.local, that is every time the
machine is started. (or you could add it to som cron-thing..)
That mean you have to know which interface connects to which
adresses, if they want to speak to the host doing the bridging.
(Even if the machines connected to any of the sides change side _without_
this routing beeing updated,
(that is
'route delete ip-nr'
'route add ip-nr other-eth?'
like:
route delete 192.168.1.4
route add 192.168.1.4 eth1
)
then the bridge itself should have no problem 'learn' the new position
of the machine, and keep working, doing the 'bridge' stuff as usual,
even if routing is broken..
But not the host, which need to be told the routing stuff... as shown
above.
Why this is so? Bridging happens on a low level, layer 2, i think.
Routing happens on a higer level. (if there was something which
looked into the bridging 'learned mac-adresses' and did the routing
for the host machine, automagically, it would have been nice, maybe
the:
http://support.broadjump.com/cerberus/
does that, I dont know..)
No need to 'export the route', that is, acting as a router, this
routing is only needed for the host itself.
This far you should be able to contact and have business with machines
on both sides.
Except for one thing. The eth0 will answer to 192.168.1.2 and the
eth1 will answer to 192.168.1.3.
To make both nic answer to one ip-nr, or two, if you like, you need
ip-aliasing. (maybe that is a kernel option, check for it if not
working.)
The easy way to get that is starting linuxconf (where all this
routing-thing also could have been done) and go to:
Config > networking > Server Task > IP aliases for virtual hosts :
And set your Network device: both eth0 and eth1 to:
IP aliases 192.168.1.2-3
Then you should be set.
Mvh Vidar Andresen
------------------------------
From: David Crooke <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Crossposted-To: alt.linux,alt.os.linux,alt.uu.comp.os.linux.questions
Subject: Linux vs NT speeds in Halloween document - TROLL - Re: Linux crashes all
Date: Sun, 15 Aug 1999 10:03:17 GMT
"George P. Staplin" wrote:
>
> Daniel Grace wrote in message <[EMAIL PROTECTED]>...
> >"I previously had IE4/NT4 on the same box and by comparison the
> >combination of Linux / Navigator ran at least 30-40% faster when
> >rendering simple HTML + graphics."
> >
> > -- Vinod Valloppillil, MS Employee.
>
> That's very strange and also FUD. Networking and browsing is much faster in
> NT for me. That's crazy to say that Navigator is 30-40% faster with Linux.
> I seriously doubt that Vinod Valloppillil is a Microsoft employee.
Actually, I believe Vinod is no longer a Microsoft employee. The quote
is taken from one of the so-called "Halloween" documents, internal white
papers written by engineers in Microsoft around a year ago to assess the
competitive threat from Linux, and leaked at the end of October, hence
the name.
For the configuration hew was testing (IIRC something like a 64Mb 200MHz
Pentium, ATI Rage video card) Netscape 4.0 on Linux 2.0 with XFree86
*is* noticeably faster than IE4 on NT4 when displaying simple HTML and
graphics. He was merely stating an observation of fact.
The biggest variation between Windows and Linux in this situation will
be how well they drive the graphics hardware - some accelerated graphics
cards are not well supported by XFree86 and so Linux graphics
performance will frankly suck on them, but on cards which it does
support well, it will visibly outperform an NT video driver for this
sort of work.
As for networking, Linux will drive a single network card more
efficiently and closer to its peak throughput, while NT supports
multiple cards better (as exploited by the design of the famous
Mindcraft webserver benchmark).
Netscape is faster than IE at rendering simple HTML, while IE is faster
at rendering complex nested table layouts.
In short, it depends on the exact circumstances.
I hope this nips your troll in the bud :-)
Dave
--
David Crooke, Austin TX, USA. +1 (512) 656 6102
"Open source software - with no walls and fences, who needs Windows
and Gates?"
------------------------------
From: Bond Robbins <[EMAIL PROTECTED]>
Subject: Samba Setup
Date: Sun, 15 Aug 1999 20:00:36 +1000
When I go into linuxconf and shared drives in samba I can't scroll all
the way down the screen it seems to be dropping off the bottom I was
wondering how to
make it so I could see al the screen??
or is there a way to set samba through the command line?
Thank you
Bond Robbins
------------------------------
From: "Suddn" <[EMAIL PROTECTED]>
Subject: Adding a Networking Card
Date: Sun, 15 Aug 1999 08:03:47 GMT
My networking card isn't working even after I have compiled the module and
used linuxconf and netcfg to set it up.
The card fails to come up at boot time and I can't force it to work with the
modprobe command. I seem to remember that there is a file where I must
define eth0 as an alias. Can anyone tell me what file that is?
If you have any other ideas why the damn thing won't work please let me
know. (I had it working but I had to reinstall Linux and now I can't seem
to get it to work.)
Thanks.
------------------------------
Date: Sat, 14 Aug 1999 23:06:01 -0700
From: "Edward M. Babb" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.hardware
Subject: Re: Gateway telepath modem
I also have had no luck. But my Gateway Telepath had to be installed by
first disabling a com port in bios, then installing the modem to that
com port from within win95. Now the port never even shows up under linux
since it is disabled in the bios, so I think I'm just going to trash the
thing and get a supported modem.
Ed
ger wrote:
>
> Hi,
> has anyone had any luck getting a Gateway "Telepath 56K" modem
> working under Linux?
> It was the only modem Gw had in stock, and I had to get an upgrade as
> the factory-installed modem was a 3COM Winmodem (ack!), despite the fact
> that I specifically ordered a non-winmodem in the system. D'oh!
>
> /proc/pci reports it as a Lucent communications device, unknown.
>
> Please respond with experiences by email, if possible.
>
> Regards, Ger.
------------------------------
From: Stefan Frings <[EMAIL PROTECTED]>
Date: Sun, 15 Aug 1999 10:15:24 GMT
Subject: Samba, Windows Workstation hangs
Hello,
I'cv got 1 Linux server and 2 Windows 98 workstations. Sometimes thy=20=
cannot access files stored on my Linux PC. My data directories are=20
public (rwx for all) and this problem occurs only a few time on a day.
I think that my network cable and card are ok because Windows had=20
never problems with the network and my internet router shows no=20
collisions.
Is there anybody who can help me?
Thanx
------------------------------
From: Stefan Frings <[EMAIL PROTECTED]>
Date: Sun, 15 Aug 1999 10:17:58 GMT
Subject: Mounting ZIP drive
Hello,
If I mount my zip drive (mount /dev/sda1 /zip =96t vfat) only root has=20=
got write access. The directory /zip has rwx for all. How can I set=20
rwx for all users?
My cdrom works fine but not the zip drive. I do not understand this.
I use Windows 98 vfat on the zip drive. Original zip disks use=20
/dev/sda3 but I reformatted them do /dev/sda1 so I am able to boot=20
them.
------------------------------
From: David Crooke <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: Help on subnetting
Date: Sun, 15 Aug 1999 10:35:06 GMT
CHAN Kin Poon wrote:
>
> Greetings,
>
> Appreciate any help on subnetting a private LAN.
>
> This is my environment: -
> 1) My LAN consists of 20+ departments, each with about 10 PCs.
> 2) The LAN consists of only Hubs and Switches in Wiring Closets and our
> Computer room, no Routers.
> 3) All our servers are located in a server farm.
> 4) All users access 1 or more servers in the server farm
> 5) I use private Class B address
>
> Questions: -
> 1) Is it necessary to subnet my environment?
No. The key thing to avoid is having one single broadcast segment at the
ethernet (MAC) level, which limits your aggregate bandwidth to the speed
of the slowest protocol (usually 10BaseT).
Switches are generally a bit cheaper than routers, and are not protocol
specific (i.e. they work on brain dead level 3 protocols like NetBEUI,
IPX and DECnet just as well as TCP/IP).
I have had very good results using multi-standard switches (that can
switch traffic between mutliple FDDI, 100BaseT and 10BaseT segments).
Cleverer ones can split long FDDI TCP/IP frames into mutliple ethernet
packets.
In this day and age, I wouldn't bother with FDDI, use 100BaseTX as your
default system, and use 10BaseT where you need to support older
equipment. You can push 100Mbit ethernet down fibre optic (100BaseFX) if
you need cable runs over 100 metres.
If you need greater than 100Mbps bandwidth to a single box, look at ATM
and Gigabit Ethernet. I would be inclined to choose the latter, as the
overall cost is lower and the bandwidth higher. ATM only shines over
long haul networks IMHO.
> 2) Can I subnet even if I've no Routers in my LAN? That is, can subnetting
> be deployed in a switched setup? All the documentation I've seen shows only
> Routers in their configurations and examples.
If you have subnets, then something has to act as a router between them.
This need not be a "router" per se, it can be e.g. a Linux PC with 2 or
more network cards. All Unix/Linux computers are capable of acting as
routers.
Don't subnet two portions of a network which are also connected by a
switch - even if the router doesn't freak out, that's just stupid.
> 3) If subnetting can be done, how should I configure the routing table at
> the PC and server ends so that packets will find their way around and back.
Say you have a PC on one subnet which is class C 192.168.1.xxx and a
server on another subnet which is
192.168.7.xxx - the router between them would have two network
interfaces, one on each subnet, let's say their addresses are
192.168.1.14 and 192.168.7.234 respectively.
On the PC:
route add -net 192.168.7.0 gw 192.168.1.14
On the server:
route add -net 192.168.1.0 gw 192.168.7.234
Note - I have used the Linux route syntax :-) Other OS'es may vary a bit
(NT in particular). I have no idea how (or if) you can configure routing
from the "pointy clicky" idiot GUI in NT, so don't ask - use the CLI
like a grown-up :-) :-)
> 4) Can the network portion of an IP address be masked as the host portion?
> That is, if I've a Class C address of 192.168.1.0 can is use a mask value of
> 255.255.0.0?
You are asking "can I use a class B netmask with a class C address
range" to which the answer is yes in certain cases, no in others, and in
general it isn't clever. Using a *smaller* netmask (e.g.
255.255.255.128) is on the other hand perfectly workable. If you are on
a private net and need a larger range, you can use 10.x.x.x with netmask
255.0.0.0 and have a whole class A to yourself :-)
Note that if you use non-standard netmask sizes you will need to
explicitly specify netmasks in a number of places, such as the route
commands in answer 3 above.
>
> Thank you in advance.
>
> --
> [EMAIL PROTECTED]
--
David Crooke, Austin TX, USA. +1 (512) 656 6102
"Open source software - with no walls and fences, who needs Windows
and Gates?"
------------------------------
From: [EMAIL PROTECTED] (Victor Vjunnik)
Subject: Linux and TCP data compression
Date: Sun, 15 Aug 1999 13:15:56 +0300
�� ��p�������� All !
I have two networks connected through leased line.
network1--linux1--cisco1--modem1==leased
line==modem2--cisco2--linux2--network2
Modem that i have is without compression.
How I may use IP packet compression on linux1 and linux2
and increase throughput of leased line ?
PS kernel 2.2.11
Victor [Team McLaren & David Coulthard]
------------------------------
From: David Crooke <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: Connecting Linux to a Mac network
Date: Sun, 15 Aug 1999 10:45:44 GMT
You haven't said which Linux distribution you're using, so advising on
which point 'n' click GUI to use is a moot point (I assume being a Mac
user you don't have a keyboard ;-)
If it's Red Hat, run the "control-panel" program as user root, and click
on the network button. Pick the interface "eth0" and change the
configuration protocol from "none" to "DHCP". Click "OK" and "Save"
where it seems sensible to do so.
>From the command line, you can use "pump" or "dhcpcd" but you'll need to
put it in the startup file /etc/rc.d/rc.local (or similar, again depends
on the distribution)
Dave
--
David Crooke, Austin TX, USA. +1 (512) 656 6102
"Open source software - with no walls and fences, who needs Windows
and Gates?"
------------------------------
From: Erwin Hogeweg <[EMAIL PROTECTED]>
Subject: Re: socket++ for linux
Date: Sun, 15 Aug 1999 10:33:02 +0200
Erwin Hogeweg wrote:
> Does somebody know where to find a linux port of socket++-1.11 (or
> later)?
Forget my question. 1.11 is fine there just seems to be a small problem
with `configure`.
CU Erwin
------------------------------
From: "Steve Cowles" <[EMAIL PROTECTED]>
Subject: Re: I'm lost in this Masquerade!
Date: Sun, 15 Aug 1999 10:48:06 GMT
Ferdinand,
Did you enable IP_FORWARDING? If not, try the following command (as root) on
your linux box and then try pinging a internet address from your WIN9x
client.
echo "1" > /proc/sys/net/ipv4/ip_forward
If the above works, then you will need to locate the startup file that sets
this automatically at boot up. I use RedHat, but I'm sure Mandrake is
similar to the following
# NOTE: On RedHat, you can enable IP_FORWARDING at boot up by
# changing the options in /etc/sysconfig/network from:
#
# FORWARD_IPV4=false
# to
# FORWARD_IPV4=true
Steve Cowles
SWCowles at gte dot net
Ferdinand V. Mendoza <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> Folks,
> I'm running Mandrake 6.0 and started tinkering
> with IP Masquerading. No matter what combination
> of ipchains command and simple rules I specify,
> I's always get the message... Destination host unreachable..
> I can ping both machines WIN95<->Linux and vice versa.
> I can connect my Linux box to the Internet via a modem-
> no problem. From my Linux box I can ping "outside"
> addressess- no problem. From my WIn 95 client, I can
> ping the private address of my Linux box- no problem.
> When connected to the internet, if I ping from my Win95
> any outside address ie, my ISP's IP address I'd get
> the horrible message. Is ping here filtered? Do
> I have to recompile the kernel?
> All suggestion are welcome.
>
> Ferdinand
>
>
>
------------------------------
From: David Crooke <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: I can't believe I've forgotten....
Date: Sun, 15 Aug 1999 10:51:03 GMT
Matt Anderson wrote:
>
> But, I took linux off my box for a while to play with the BE os. (Very
> cool by the way, but not as mature as linux yet)
>
> Anyway, I have a windows 98 machine in my home network and I have the
> stupid encrypted password thing messing up my Samba shares.
> In the past, I turned off encrypted passwords in the 98 box. I'd rather
> enable encyption on the Samba/Linux side. However, I don't remember how
> to do the first, and I'm not sure where to start on the second.
> Any help would really be appreciated!
I'll circumvent the complex explanation as to why Windows SMB passwords
are equally insecure whether encrypted passwords are turned on or not.
Suffice to say that if your dog is smart enough to set up a packet
sniffer on your home network, he's smart enough to dump the password
file off your hard disk.
But I digress....
Do the first, it's a lot less hassle...find the file
"Win95_PlainPassword.reg" or similar supplied with Samba, copy it to the
Win98 box and double click it.
Dave
--
David Crooke, Austin TX, USA. +1 (512) 656 6102
"Open source software - with no walls and fences, who needs Windows
and Gates?"
------------------------------
From: David Crooke <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Crossposted-To:
alt.linux,alt.solaris.x86,athome.users-general,athome.users-unix,athome.users-win95,cableinet.cable_modems,comp.os.ms-windows.networking
Subject: Re: Home network and cable modem
Date: Sun, 15 Aug 1999 10:09:00 GMT
Jim Pickering wrote:
>
> Steve: One of the ports should be identified somehow with an imprint of
> a crossed connection, whether engraved in the plastic or silk screened
> on. This would be the uplink port and would not require a cross over
> cable but just the cable from your cable modem to it. The two computers
> then would go to any of the other ports. Both computers should be able
> to access the internet at the same time and should also be able to
> communicate with one another. Have you set up your network and can you
> see each other in Network Neighborhood? Have you gone to Intel's
"Network Neighbourhood" is a concept which only applies to Microsoft
NetBIOS networking. Depending on how the nwteorking is configured on
Microsoft systems, it may or may not show all machines. The original
poster was also running Linux and Solaris on one of his machines,
neither of which (as standard) uses any NetBIOS protocols. Even with
Samba (the open source NetBIOS server for Unix) running, a Unix system
will not necessarily show up in NetBIOS browsing ("Network
Neighbourhood") unless configured to do so.
Be wary of making assumptions and giving people a bum steer!
--
David Crooke, Austin TX, USA. +1 (512) 656 6102
"Open source software - with no walls and fences, who needs Windows
and Gates?"
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.networking) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Networking Digest
******************************
