Linux-Networking Digest #670, Volume #10         Mon, 29 Mar 99 15:13:31 EST

Contents:
  Re: Using Linux instead of NT Server in home environment.... (jedi)
  Re: kernel: ip_masq_new(proto=TCP): no free ports. (Brian Turner)
  Re: Two NIC's in 1 machine for double bandwidth? (Matt Corddry)
  Microsoft Proxy server 2.0 and Red Hat 5.2 ("Marat Garafutdinov")
  Re: AFS for linux (Magnus Ahltorp)
  XDR problem (on Linux): xdr_hyper (and xdr_uint) nonexistent. (terry jones)
  Re: IP Masquerading almost successful (Matt Corddry)
  Re: POP3 Errors (Matt Corddry)
  Re: Dialup server setup (Matt Corddry)
  Re: internal modem (Clifford Kite)
  PPP IP address ("R. Brooks")
  Re: winmodems ([EMAIL PROTECTED])
  Re: Bind 8.1.2 and MX records ([EMAIL PROTECTED])
  Re: Machine name themes - what do you use? (Tony Wright)
  Problem with arp reply
  Help: Winbond W89C84F (Patrick Griffon)
  Re: Linux under NT40 Proxy Server 2.0 ("Klas Eliasson")

----------------------------------------------------------------------------

From: [EMAIL PROTECTED] (jedi)
Crossposted-To: 
microsoft.public.windowsnt.misc,microsoft.public.windowsnt.setup,comp.os.linux.misc,comp.os.linux.setup,micorosft.public.outlook
Subject: Re: Using Linux instead of NT Server in home environment....
Date: Mon, 29 Mar 1999 09:05:35 -0800

On Mon, 29 Mar 1999 17:51:02 +0300, Alexander I. Butenko <[EMAIL PROTECTED]> wrote:
>I finally explain that I  mean that it's ANYWAY more easier to install NT
>than Linux for a novice user.

        ...not for any task that has it's complexity 'built in'
        regardless of GUI (or not).

>David Martin <[EMAIL PROTECTED]> wrote in message
>news:7dn8fq$ntm$[EMAIL PROTECTED]...
>> In article <#H7f6dVe#GA.267@cppssbbsa03>,
>> "Alexander I. Butenko" <[EMAIL PROTECTED]> writes:
>> > Well,, I'd beter think that there is a Server for the client, not
>> > vice-versa. And one more - the question was about the OS for the home.
>MOst
>> > hom eusers can't even properly configure Win98, so the most correct
>answer
>> > about the Server was not linux but WIndows NT. Anyway I do not wasn to
>> > continue this flame as most of this group members seem to gain nothing
>from
>> > it.
>>
>> If most people can't even set up win98 at home, what hope have they with
>NT?
>>
>> If you want to see how easy it is to set up a home network, with IP
>> masquerading, firewalls etc. I would reccommend John Sery's book 'Linux
>> Network  Toolkit' (IDG) which comes with all the software you need to get
>it
>> going. Starts off with simple networks and works up to workgroup size
>complex
>> ones.
>>
>> ..d
>


-- 

  "I was not elected to watch my people suffer and die     |||
   while you discuss this a invasion in committe."        / | \

        In search of sane PPP docs? Try http://penguin.lvcm.com

------------------------------

From: Brian Turner <[EMAIL PROTECTED]>
Subject: Re: kernel: ip_masq_new(proto=TCP): no free ports.
Date: Mon, 29 Mar 1999 18:18:41 GMT

Leopold Toetsch <[EMAIL PROTECTED]> wrote:

> How does your /etc/route.conf look like?
(192.168.192.0 is, naturally, our internal network)

external.network           0.0.0.0 255.255.255.0 eth1
192.168.192.0           0.0.0.0                 255.255.255.0
eth0
default                 ISPs.default.gateway

> # route -n
> before you get connected and
(before ifconfig)

Destination     Gateway         Genmask         Flags Metric Ref    Use
Iface
192.168.192.1   0.0.0.0         255.255.255.255 UH    0      0        0
dummy0
127.0.0.0       0.0.0.0         255.0.0.0       U     0      0        0 lo

> # route -n
> after you get connected.
(after ifconfig, ipfwadm, and insmod)

192.168.192.1   0.0.0.0         255.255.255.255 UH    0      0   
     0 dummy0
external.network   0.0.0.0         255.255.255.0   U     0      0  
     55 eth1
192.168.192.0   0.0.0.0         255.255.255.0   U     0      0    
  162 eth0
127.0.0.0       0.0.0.0         255.0.0.0       U     0      0       
  15 lo
0.0.0.0         ISPs.default.gateway   0.0.0.0         UG    0 
     0 43834 eth1



> How do you start masquerading & firewall (ipfwadm ...) ?
ipfwadm and insmod (for the masq modules)

Thanks for your help,

Brian


------------------------------

Date: Sun, 28 Mar 1999 23:48:10 -0800
From: Matt Corddry <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.ibm.pc.hardware.networking
Subject: Re: Two NIC's in 1 machine for double bandwidth?

Bio Hazard wrote:
> 
> On Sun, 28 Mar 1999 13:42:21 GMT, [EMAIL PROTECTED] (John R.
> Campbell) wrote:
> >       You *can* boost the *aggregate* bandwidth of the network but
> >       you cannot boost the bandwidth of a particular task (session)
> >       by doing this (although, with multiple web server processes you
> >       can subdivide the bandwidth and get bigger numbers).  If you're
> >       doing simple FTP transfers, no, only one of the cards'll be
> >       utilized.
> 
> There must be a way.  I mean, I can hook up two 56K modems to two
> phone lines, and do multiplexing on the Internet to get double the
> bandwidth (sometimes called "Shotgun" technology).  There must be a
> way to do something similar with two 100Base-T network cards in each
> machine.  Let me know if anyone knows if this is possible.

Well... not in linux, yet. The technology to do this is referred to as
Fast EtherChannel (at least by cisco). It involves bonding multiple net
cards & cables (usually 2, up to 4) into one virtual link, with
bandwidth aggregation and hot fail-over. The only PC NICs that do this
are the new Intel Server cards (inexpensive, but they don't work w/ the
standard linux intel driver). Sun quad-fastenet nic's work w/ the cisco
switches and routers that support this technology, and I'd assume that
other big unix platforms support this as well. However, with the recent
interest in linux that intel has shown, it wouldn't surprise me if they
helped out in providing a linux driver for these server cards (they're
nt/novell-only right now).

You can (as previous posts detailed) assign a different IP address to
each NIC on the same subnet, and loadbalance incoming connections either
with a portforwarder (bigIP, whatever) or a DNS round-robin system.
Doesn't help you at all with single-connection bandwidth, tho.

matt

------------------------------

From: "Marat Garafutdinov" <[EMAIL PROTECTED]>
Subject: Microsoft Proxy server 2.0 and Red Hat 5.2
Date: Sun, 28 Mar 1999 23:41:30 -0800

Question:
Is it possible to access  Internet from my linux box from BEHIND the MS
Proxy server (same IP as our default gateway)
I'm running Red Hat 5.2 and the computer MS Proxy is running on has NT 4.0
Thanks.




------------------------------

From: Magnus Ahltorp <[EMAIL PROTECTED]>
Subject: Re: AFS for linux
Date: 29 Mar 1999 20:54:31 +0200

Jerome Fayot <[EMAIL PROTECTED]> writes:

> I am looking for a free software like AFS for linux. Does anyone
> know about this ?

There is a free AFS client (BSD style license) for Linux called
Arla. There is support for both Linux 2.0 and 2.2, but I'd recommend
using 2.2.

You can also run Arla on FreeBSD, OpenBSD, NetBSD, SunOS, Solaris,
AIX, IRIX and Digital UNIX. Most hardware platforms are supported.

Web page:           http://www.stacken.kth.se/projekt/arla/
Latest version:     ftp://ftp.stacken.kth.se/pub/arla/arla-0.23.tar.gz
Mailing list:       [EMAIL PROTECTED]
Subscribe requests: [EMAIL PROTECTED]

/Magnus

------------------------------

From: terry jones <[EMAIL PROTECTED]>
Crossposted-To: 
comp.os.linux.development.system,comp.os.linux.development.apps,comp.unix.questions,comp.unix.internals
Subject: XDR problem (on Linux): xdr_hyper (and xdr_uint) nonexistent.
Date: 28 Mar 1999 23:59:19 -0800


I have Linux 2.2 on a Pentium.

I'm trying to use XDR to implement passing various C types via
TCP/IP. I've made myself a .x file and rpcgen produces .c which I can
compile just fine.

Unfortunately, the resulting .o contains references to xdr_uint and
xdr_hyper (due to my use of unsigned int and hyper in my .x file) and
the linker cannot resolve these.

I worked around some of this via compiling with

  -Dhyper='long long' -Dxdr_uint=xdr_u_int

after finding xdr_u_int in the C library. nm -a doesn't reveal
xdr_hyper (or anything resembling xdr_longlong) in any other libraries
in /lib or /usr/lib.

Is there no xdr function for sending a long long under Linux on a
Pentium (i.e., sending a 64-bit long)?


Has anyone done this before, or do I need to write my own xdr_hyper?

Maybe the reason this doesn't exist in the C library (or libnsl) is
because long long isn't very well defined acrosss architectures (in
terms of size)? If I do write it myself, do I have to resign myself to
it only working on some fixed architectures (e.g., 32-bit with long
long = 64 bits)? Presumably 64-bit architectures have sizeof(long
long) > 8, so could I solve the problem (portably) by implementing
xdr_hyper to use (say) 128 bits?


Thanks for any help.

Terry Jones ([EMAIL PROTECTED]).

------------------------------

Date: Mon, 29 Mar 1999 00:02:18 -0800
From: Matt Corddry <[EMAIL PROTECTED]>
Subject: Re: IP Masquerading almost successful

Hmmm... some general addressing problems I see here...

You said your router has the IP address 192.168.1.254 -- is it running
NAT (network address translation, similar to ipfw)? If not, then it's
being treated as a device directly on the internet, and it can't have a
192.168.x.x IP address, since those are nonrouting reserved IPs (what
you use behind a firewall). So... assuming you're running a standard
router with no address translation capability (model #, wan type, config
would help), the router needs to have a 139.x.x.x address, as does eth1
in the linux box. You can then run the 192.168.0.x/24 network behind the
ipfw box. Oh, and I don't think ICMP (ping) masquerading is enabled in
the kernel by default -- it's a seperate option when you compile. So
your pings won't get thru regardless.

The linux config would then look something like (I'll let you translate
into RH config files):

ipfwadm -F -a m -S 192.168.0.0/24 -D 0.0.0.0/0
default gw: 139.x.x.<addr of router>
eth1: 139.x.x.<addr of linux box>
eth0: 192.168.0.1

This is all assuming that your ISP is routing a range of IP addresses to
you (something like 139.1.1.0/24), so you have an IP address for the
router, and a different one for the linux box.

If you are running NAT on the router, then I really don't see any reason
to run ipfw at all, since you should already have all the addresses you
need from the NAT setup, and you should also have decent security.

hope this helps

matt

[EMAIL PROTECTED] wrote:
> 
> Cheers everyone, I hope you can help me.
> 
>         This is the network design I want to implement:
> 
>         Internet
>                 I
>                 I
>         ISP Router 139.X.X.X DNS
>                 I
>                 I  139.X.X.X Dynamically assigned
>         Our Router 192.168.1.254
>                 I
>              minihub-----WWW Server 192.168.1.1
>                 I
>         Linux Box eth1  192.168.1.253  RedHat 5.2
>                 I eth0  192.168.0.254  Kernel 2.0.36
>                 I
>                 I
>   192.168.0.3--LAN Hub ------192.168.0.1
>                 etc.---192.168.0.2
> 
> - I've set up the LAN clients (NT and W9x boxes) with
>   192.168.0.254 as default gateway, and 139.X.X.X as DNS
> 
> - On the Linux, both NICs are correctly detected and show on dmesg
> 
> - I've added the /sbin/depmod & modprobe ip_masq lines to   /etc/rc.d/rc.local
> and also echo "1" > /proc/sys/net/ipv4/ip_forward
> 
> - My /etc/sysconfig/network:
> 
>         NETWORKING=yes
>         FORWARD_IPV4=TRUE
>         HOSTNAME=host.gw
>         DOMAINAME=gw
>         GATEWAY=0.0.0.0  - Are these
>         GATEWAYDEV=eth0  - Ok?
> 
> - My /etc/resolv.conf:
> 
>         search gw
>         nameserver 139.X.X.X
> 
> - My /etc/sysconfig/network-scripts/ifcfg-eth0
> 
>         DEVICE=eth0
>         IPADDR=192.168.0.254
>         NETMASK=255.255.255.0
>         GATEWAY=192.168.1.253 - Is this Ok?
>         BROADCAST=192.168.0.255
>         NETWORK=192.168.0.0
>         ONBOOT=yes
> 
> - My /etc/sysconfig/network-scripts/ifcfg-eth1
> 
>         DEVICE=eth1
>         IPADDR=192.168.1.253
>         NETMASK=255.255.255.0
>         GATEWAY=192.168.1.253 - Is this Ok?
>         BROADCAST=192.168.1.255
>         NETWORK=192.168.1.0
>         ONBOOT=yes
> 
> - I've tried with this rules:
> 
>         ipfwadm -F -p deny
>         ipfwadm -F -a m -S 192.168.0.0 eth1
> 
> - Results:
> 
>  I can't see our router, 192.168.1.254 pinging from the Linux.
>  Should I manually add any route command?, such as:
> 
>         route add -host 192.168.1.254 dev eth1
> 
>  What's the point I'm missing? Are Ok those Gateway's IPs? Are ok those
> rules?
> 
>   I'd be grateful if anyone here could enlighten me.
> 
>   Please, reply either to the group as well as to my private e-mail.
> 
>   Thanks in advance.
> 
> -----------== Posted via Deja News, The Discussion Network ==----------
> http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own

------------------------------

Date: Mon, 29 Mar 1999 00:03:27 -0800
From: Matt Corddry <[EMAIL PROTECTED]>
Subject: Re: POP3 Errors

Sean Fancey wrote:
> 
> I have a Linux 3.3/Kernel 2.0.33 whihc has been running for over year.
> Everything is working smoothly except for one little thing.  Everytime I
> try to access any of the e-mail accounts using a POP3 client (Such as
> Netscape), in.pop3d sends the following error:
> 
>  being read already /usr/spool/mail/frodo
> 
> Is there something not quite correct about my POP3 setup?  Also, are
> there other daemons such as rsmtp that have to running in order for this
> to work.
> 
> Thanks in advance
> ---------------------------------------------------------------------------
> 
> "If GM had developed technology like Microsoft, for no reason whatsoever
> your car would crash twice a day." - Mr. Welch


I was having these problems on a similar setup awhile ago. I think it
was a problem in the pop3 daemon, because I installed the qualcomm pop3d
and the problem went away.

matt

------------------------------

Date: Mon, 29 Mar 1999 00:08:08 -0800
From: Matt Corddry <[EMAIL PROTECTED]>
Subject: Re: Dialup server setup

"�v����" wrote:
> 
> Hi to all,
> 
> Is there any dialup server program/tools for Linux (X11 or console)? Or i
> can directly use commands/tool in Linux
> I'm trying to build a RAS server using linux (redhat 5.2), but cannot find
> any resources to setup...
> Thx

check out mgetty at http://www.leo.org/~doering/mgetty/ ... also be sure
to check out autoPPP (linked from the mgetty page) to automagically
start PPP with windoze clients.

matt

------------------------------

From: [EMAIL PROTECTED] (Clifford Kite)
Subject: Re: internal modem
Date: 29 Mar 1999 12:03:49 -0600

Olle Soderqvist ([EMAIL PROTECTED]) wrote:

: I have a Rockwell internal v90 modem and I have the problem that I have seen
: that a whole bounce of other also have!
: Linux can find my modem on COM3 but everything takes a very, very long
: time....for exampel writing the letters ATDT takes 20 seconds before they
: appear on the screen! And everything goes so damn slow! So does anyone know
: it that's because I have a winmodem or is this problem fixeble??

It's likely that the modem's IRQ is not the one configured for /dev/ttyS2.


--
Clifford Kite <[EMAIL PROTECTED]>                       Not a guru. (tm)
/* The wealth of a nation is created by the productive labor of its
 * citizens. */

------------------------------

From: "R. Brooks" <[EMAIL PROTECTED]>
Subject: PPP IP address
Date: Mon, 29 Mar 1999 14:22:26 -0500

I connect with PPP and am going to Cable Modem with
DHCP.
I need a way to obtain the IP address as a system variable
so I can use IP chains to block connections to my box.
Does anyone know if it is stored in some variable or
a util to get it stored?

Thanks

Randall

--
_____________________________________________
Randall Brooks


------------------------------

From: [EMAIL PROTECTED]
Subject: Re: winmodems
Date: Mon, 29 Mar 1999 19:04:42 GMT

On Sat, 27 Mar 1999 22:43:25 -0600, Sami Yousif <[EMAIL PROTECTED]>
wrote:

<--snip-->
>
>So  yes... A winmodem driver would be a good thing just so that we can say
>"Linux supports most hardware, even those pesky winmodems"
>
>--
>-
>---
>Sami Yousif

that and with the better design and performance of linux, who knows,
the winmodem might prove to work better under linux with a good
driver.  then I guess they would have to call it a  lin-modem :)

tng

------------------------------

From: [EMAIL PROTECTED]
Subject: Re: Bind 8.1.2 and MX records
Date: Mon, 29 Mar 1999 19:07:41 GMT

On Sun, 21 Mar 1999 12:23:15 +0100, Carles Arjona
<[EMAIL PROTECTED]> wrote:

>Bill Dunn <[EMAIL PROTECTED]> wrote:
>> 
>>     I have always had the MX record set as mail.whatever.com, which is set
>> as a CNAME to a previous A record.
>
>That's wrong. From the http://www.sendmail.org/virtual-hosting.html web
>page:
>
>"3.Configure MX records for your domain (Note: CNAME records can not be
>used; see � 5.2.2 of RFC 1123 for details.)"
>

simple way around that though... you can use IN A records to assign
multiple names to the same machine example

host1   IN      A       192.168.0.1
mail            IN      A       192.168.0.1

this might not be the prefered way to do it and probably not
recomended, but it works.

tng
>http://www.sendmail.org/rfc/1123.html#Section5.2.2
>
>
>Regards.
>-
>Carles Arjona    [EMAIL PROTECTED]
>-
>Please, just remove the second NOSPAM for replies from newsgroups.
>(Yes, NOSPAM is my real username)
>[Don't send me e-mail copies of usenet postings, please]


------------------------------

Date: Mon, 29 Mar 1999 20:17:52 +0000
From: Tony Wright <[EMAIL PROTECTED]>
Crossposted-To: 
vmsnet.networks.misc,comp.unix.solaris,comp.os.os2.networking.server,comp.os.ms-windows.nt.admin.networking,comp.infosystems.www.servers.unix,comp.protocols.tcp-ip.domains
Subject: Re: Machine name themes - what do you use?

James Thompson wrote:

> I tend to use endangered or extinct animals:
>
> e.g: uakari, indris, dingo, etc.
>

Shouldn't OS/2 be included then...?

--
Rgds Tony W   Email: [EMAIL PROTECTED]

"humanum est errare: To err is human
.... and to fail is to be a Project Manager...
...but to foul things up completely needs a computer!"

------------------------------

From: [EMAIL PROTECTED] ()
Subject: Problem with arp reply
Date: 29 Mar 1999 05:21:25 GMT

Hi,

I'm looking for a solution to aproblem I've been tearing my hair out over
the last few days.  

First, a little background.  I have a wireless ethernet link using a pair
of 900MHz wavelan cards in linux boxes.  Both are acting as routers in
a setup soemthing like this:

               Box 1                         Box 2
Internet <->  ethernet                       ethernet <-> Home network
              wavelan  <-------------------> wavelan

Box 1 is running kernel 2.2.3 and using ipchains to to IP MASQ for
everything that comes in over the wavelan link.  

Box 2 is running kernel 2.0.33 and has a default route out the wavelan
card to box1 (routes for the home network)

Problem is such.  Sometimes but not all the time(ARGH!) box2 can't seem to
pass traffic over to box1.  So, with an alternate link to box1, I run
tcpdump on both machines.  What I see confuses me.  

On box 2, I see "arp who has box 2" messages received on the wireless
interface, and then an "arp reply ..." being sent back out the interface
with box2's hardware address.

On box 1, at the same time, I see each of those "arp who has box2" packets
being sent out, but I never see the "arp reply" messages arriving from the
far end.  Okay, I think, there's something wrong at the RF level and box 2
can hear box1, but box 1 can't hear box 2.  Bzzzzt.  cat
/proc/net/wireless shows that packets are being received on both ends and
link stats look good.  Also, when sending a ping from box 2 to box 1, I
see the ICMP packet transmitted on box2, and then I see it arrive at box1
-- but box 1 can't reply to it becuase it never seems to get the "arp
reply" it's begging for about once a second.

Box 2's arp table shows box 1's hardware address, but box 1 shows
(incomplete) for box2.

Hoping to outsmart the kernel, I manually used arp on box 1 to tell it
box2's hardware address.  This allows traffic to start flowing between
box1 and box 2.  Unfortunately, though, IP_MASQ doesn't seem to work right
in this scenario.  PINGs to the net at large work fine, but things like
telnet, ssh, ftp, and smtp seem to have some kind fo strange flow-control
problem -- where packets from the remote don't seem to get sent back to
the home network correctly.  For instance, a telnet connection takes a
very long time to produce a login prompt, and when it does, the seesion is
very slow and output from the remote end only seems to make it to the home
network end immediately when a packet is sent from the home network end
(I can type, say, 'w <enter>', and get the first line of output of the 'w'
command, then it just stops.  If I press enter again, I get both the
rest of the output of my command and two prompts.

This stuff is driving me bonkers becuase 1) it was working just fine last
week before I rebooted box 1 and 2) the problem is intermittant... I can
sit here watching tcpdump for hours and never see an arp reply on box1,
and then I'll turn around and all of a sudden it's recieved one and
everything works for a while... until the arp entry times out I think.

any suggestions would be greatly appreciated.  I can provide any
additional info as well.

Thanks,

Greg Romaniak


--

------------------------------

From: Patrick Griffon <[EMAIL PROTECTED]>
Subject: Help: Winbond W89C84F
Date: Mon, 29 Mar 1999 09:48:13 +0000

Hi,

Can anyone tell me what drivers I should be using ?

cat /proc/pci shows
ethernet winbond ... IRQ 10,E400
but the boot don't detect the card, nothing by dmesg.

The ne2k-pci seems not support this card just teh W89C940

Thanks


------------------------------

From: "Klas Eliasson" <[EMAIL PROTECTED]>
Subject: Re: Linux under NT40 Proxy Server 2.0
Date: 29 Mar 99 19:50:19 GMT

Its the software wich is going to be configured to use a proxy, not the OS.
So - just set up your email-, www- and newsclients to use the proxy -
exactly like you do on your win95 clients.

//Klas Eliasson, sweden

Dante A. Garcia <[EMAIL PROTECTED]> skrev i inl�gg
<[EMAIL PROTECTED]>...
> Hi,
> I am running a small home network, i have one physical connection to the
> Internet by Mediaone (cable), this connection go to and NT40 Proxy
> Server, from the Proxy server go to a HUB connecting 4 clients, a mix of
> Windows NT and Windows 95. I recently purchased Linux from Red Hat 5.2 i
> also add and extra HD dedicated to Linux, installed Linux and now i have
> to get access to the Internet trought the NT Proxy Server, is this
> possible? if so what i have to do for this to work.
> The Proxy Server configuration is as follow, the first NIC is the
> default setup from MediaOne, the second is TCP/IP number: 192.168.0.1,
> Subnet Mask: 255.255.255.0, Default Gateway: Blank, under DNS tab
> Hostname: ProxyNetServer, Domain: MyDomain.
> 
> Thank You.
> 
> Dante.
> 
> 
> 

------------------------------


** 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
******************************

Reply via email to