[techtalk] Telnet, solaris and .profile

2000-08-17 Thread Britta Koch

Hi!

My home directory at work is mounted via NFS on my Linux box. It's got a
.profile and a .bash_profile file. When I login from Linux, it reads my
.bash_profile file. When I telnet into the Solaris 5.6 box that hosts my home
directory, my .profile file is not read (and I don't want it to read my
.bash_profile - they're different files for a reason). It's annoying me that I
always have to . .profile after logging in. Is there any way I can automate
this?

Thanx for any answer

Britta


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



Re: [techtalk] Netstat weirdness

2000-08-17 Thread Conor Daly

On Wed, Aug 16, 2000 at 04:19:33PM -0500 or so it is rumoured hereabouts,
 [EMAIL PROTECTED] thought...
> 
> Here's the route list you gave earlier:
> |Destination Gateway Genmask Flags   MSS Window  irtt Iface
> |63.251.67.560.0.0.0 255.255.255.248 U 0 0  0 eth0
> |192.168.0.0 0.0.0.0 255.255.255.0   U 0 0  0 eth1
> |127.0.0.0   0.0.0.0 255.0.0.0   U 0 0  0 lo
> |0.0.0.0 63.251.67.570.0.0.0 UG0 0  0 eth0 
> 
> The way the kernel routes packets is, for each route table entry, in
> order, it takes the IP address you're trying to reach, binary-ANDs it
> with the genmask, and then compares the result with the destination.
> If it matches, it sends the packet to the specified interface.  If the
> G flag is NOT set, then this is a direct route; if G is not set, then
> you're going out a gateway.  So, in your case, 63.251.67.56 thru
> 63.251.67.63 all go direct to eth0, 192.168.0.0 through 192.168.0.255
> all go direct to eth1, 127.0.0.0 thru 127.255.255.255 all go direct
> thru lo (the loopback interface), and everything else is gatewayed to
> 63.251.67.57 via eth0.  This is perfectly consistent with your
> ifconfig settings.
> 
> The network on eth0 has the network address 63.251.67.56.  This should
> not be the IP address of any machine on that network, since subnet
> address 0 is reserved.  .56 is the network address and .63 is the
> broadcast address for this subnet.
> 
> In short, there's doesn't appear to be anything at all wrong with your 
> routing.
> 

Question for you...

I'm using an old 486 as internet gateway / router / firewall.  I'm using a
P200 as server / DNS / squid proxy (Well, I will be once I get around to 
configuring squid).  I want to set things up so that clients need one IP
address only for gateway, DNS, proxy etc.  How do I set up the routing?

Firewall192.168.1.254
Server  192.168.1.1
Clients 192.168.1.2-5

Client Routing table
Destination Gateway Genmask Flags   MSS Window  irtt Iface
192.168.1.254   0.0.0.0 255.255.255.255 U 0 0  0 eth0
192.168.1.0 0.0.0.0 255.255.255.0   U 0 0  0 eth0
127.0.0.0   0.0.0.0 255.0.0.0   U 0 0  0 lo
0.0.0.0 192.168.20.10.0.0.0 UG0 0  0 sl0


Server Routing table
Destination Gateway Genmask Flags   MSS Window  irtt Iface
192.168.1.1 0.0.0.0 255.255.255.255 U 0 0  0 eth0
192.168.1.0 0.0.0.0 255.255.255.0   U 0 0  0 eth0
127.0.0.0   0.0.0.0 255.0.0.0   U 0 0  0 lo
0.0.0.0 192.168.1.254   0.0.0.0 UG0 0  0 eth0

If I point a client to the server as it's gateway, nothing gets routed out
through the gateway.  Only if I give the actual gateway machine as gateway
does the routing happen but I want to use the squid proxy transparently and
supply only one IP address to be used for gateway / dns .  

Is it simply a matter of using ipchains to set up forwarding from the server
to the gateway (which, itself, uses ipchains for the firewall / Ip masq)?

-- 
Conor Daly <[EMAIL PROTECTED]>

Domestic Sysadmin :-)


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



Re: [ILUG] Re: [techtalk] How a script is called

2000-08-17 Thread Conor Daly

On Thu, Aug 17, 2000 at 01:24:25AM +0100 or so it is rumoured hereabouts,
 Fergal Daly thought...
> At 19:48 16/08/00, Conor Daly wrote:
> >
> >Nice answers everyone but, alas, there is an implicit assumption that there's
> >only one arg or none involved.  I'm accepting up to three args in any order at
> >present and have no control over how the script is called in the case of a
> >pipe.
> 
> Not trying to start (another ;-) fight but did you read my reply? Using if 
> [ -t 0 ] makes no assumptions about arguments, it is a common and standard 
> way of checking whether you're being piped or not. You can find it in no 
> less an authority than the Perl Cookbook and I'm sure many other places,
> 
Fergal,

Your (sexy) answer didn't appear at my mail client until after I posted the
above response.  I responded to that somewhere in this thread also.

Thanks again!! :-)

-- 
Conor Daly <[EMAIL PROTECTED]>

Domestic Sysadmin :-)


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



Re: [techtalk] Netstat weirdness

2000-08-17 Thread Malcolm Tredinnick

On Thu, Aug 17, 2000 at 08:49:57AM +0100, Conor Daly wrote:
> Question for you...
> 
> I'm using an old 486 as internet gateway / router / firewall.  I'm
> using a P200 as server / DNS / squid proxy (Well, I will be once I get
> around to configuring squid).  I want to set things up so that clients
> need one IP address only for gateway, DNS, proxy etc.  How do I set up
> the routing?

[...snip...]

> If I point a client to the server as it's gateway, nothing gets routed
> out through the gateway.  Only if I give the actual gateway machine as
> gateway does the routing happen but I want to use the squid proxy
> transparently and supply only one IP address to be used for gateway /
> dns .  
> 
> Is it simply a matter of using ipchains to set up forwarding from the
> server to the gateway (which, itself, uses ipchains for the firewall /
> Ip masq)?

It's even easier than that: you just need to tell the kernel to allow
forwarded packets. If you are using a recent RedHat system (I think it
needs to be 6.2 or later), then edit /etc/sysctl.conf and change the
line that say net.ipv4.ip_forward=0 to be net.ipv4.ip_forward=1.

On other distributions, there may be a similar method, but the general
solution is just to do:

echo 1 > /proc/sys/net/ipv4/ip_forward

This simply tells the kernel that it is allowed to forward packets (you
would have had to turn this on for your firewall box, but it may have
been done magically for you by whatever you used to set up the
firewall).

Cheers,
Malcolm

-- 
Malcolm Tredinnickemail: [EMAIL PROTECTED]
CommSecure Pty Ltd

 PGP signature


Re: [techtalk] cron help needed!

2000-08-17 Thread Malcolm Tredinnick

On Wed, Aug 16, 2000 at 03:17:32PM -0700, Amanda LeNay wrote:
> We upgraded to Red Hat 6.2 (from 5.0) and now our cron jobs aren't
> running. The cron log file isn't showing any errors - in fact it looks
> like everything is running - only it isn't - nothing is executing. I
> have tried restarting crond but I don't know what else to do at this
> point.
> 
> Any help would be much appreciated.

Two suggestions to help diagnose your problem:

(1) Look in /var/log/messages and see if anything was reported by crond.

(2) Try to run the various scripts by hand (/var/log/cron shows the
commands cron was trying to run) and see what appears.

Apologies if these are obvious things that you've already tried, but I
can't guess what else might be going wrong. My first thought is that
some crucial file you are relying on has moved in the upgrade (some
system thing that RH moved to a "better" location and you've hardcoded
the path), but it's only a guess.

Cheers,
Malcolm

-- 
Malcolm Tredinnickemail: [EMAIL PROTECTED]
CommSecure Pty Ltd

 PGP signature


Re: [techtalk] Telnet, solaris and .profile

2000-08-17 Thread Conor Daly

On Thu, Aug 17, 2000 at 09:33:50AM +0200 or so it is rumoured hereabouts,
 Britta Koch thought...
> Hi!
> 
> My home directory at work is mounted via NFS on my Linux box. It's got a
> .profile and a .bash_profile file. When I login from Linux, it reads my
> .bash_profile file. When I telnet into the Solaris 5.6 box that hosts my home
> directory, my .profile file is not read (and I don't want it to read my
> .bash_profile - they're different files for a reason). It's annoying me that I
> always have to . .profile after logging in. Is there any way I can automate
> this?
> 

There's a way to pick up the Ip of the machine you're coming in from (can't
think of it now).  You could have something that checks that IP and does a 
source .profile
on the basis of the IP.  Does a .login or anything run?

-- 
Conor Daly <[EMAIL PROTECTED]>

Domestic Sysadmin :-)


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



Re: [techtalk] Netstat weirdness

2000-08-17 Thread Conor Daly

On Thu, Aug 17, 2000 at 08:41:00PM +1000 or so it is rumoured hereabouts,
 Malcolm Tredinnick thought...
> 
> On other distributions, there may be a similar method, but the general
> solution is just to do:
> 
>   echo 1 > /proc/sys/net/ipv4/ip_forward
> 
> This simply tells the kernel that it is allowed to forward packets (you
> would have had to turn this on for your firewall box, but it may have
> been done magically for you by whatever you used to set up the
> firewall).
> 

The "echo 1 > ... " line is indeed in use on the gateway in the firewall
script, not sure if it is there on the server (both running RH6.2).  I seem to
remember (when using linuxconf) checking the "allow packet forwarding" on the
server but doing it in the config files always seems to be the better route.  

I have found at times that, after modifying a config file with vi, linuxconf
doesn't seem to read back the config from the same file(s).  Does it keep a
database of settings elsewhere or something?

-- 
Conor Daly <[EMAIL PROTECTED]>

Domestic Sysadmin :-)


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



Re: [techtalk] Netstat weirdness

2000-08-17 Thread Malcolm Tredinnick

On Thu, Aug 17, 2000 at 12:54:55PM +0100, Conor Daly wrote:
> On Thu, Aug 17, 2000 at 08:41:00PM +1000 or so it is rumoured hereabouts,
>  Malcolm Tredinnick thought...
> > 
> > On other distributions, there may be a similar method, but the general
> > solution is just to do:
> > 
> > echo 1 > /proc/sys/net/ipv4/ip_forward
> 
> The "echo 1 > ... " line is indeed in use on the gateway in the
> firewall script, not sure if it is there on the server (both running
> RH6.2).  I seem to remember (when using linuxconf) checking the "allow
> packet forwarding" on the server but doing it in the config files
> always seems to be the better route.  
> 
> I have found at times that, after modifying a config file with vi,
> linuxconf doesn't seem to read back the config from the same file(s).
> Does it keep a database of settings elsewhere or something?

No. Linuxconf tries to parse the various configuration files to work out
what the current settings are. In general, I think it is safer to use
linuxconf all the time, just to keep everything sane.

In this case, though, linuxconf (on RH6.2) will change the
/etc/sysctl.conf file. If you do the same thing, I would recommend
altering the comment before the line that sets up ip_forwarding (it is a
particularly *stupid* comment, imho, since it says what the default
setting does and if you change that setting without altering the
comment, you are left wondering whether 1 means "enabled" or
"disabled").

Cheers,
Malcolm

-- 
Malcolm Tredinnickemail: [EMAIL PROTECTED]
CommSecure Pty Ltd

 PGP signature


RE: [techtalk] Telnet, solaris and .profile

2000-08-17 Thread Fan, Laurel

Britta Koch, [EMAIL PROTECTED], said:
> My home directory at work is mounted via NFS on my Linux box. 
> It's got a .profile and a .bash_profile file. When I login
> from Linux, it reads my .bash_profile file. When I telnet into
> the Solaris 5.6 box that hosts my home directory, my .profile
> file is not read (and I don't want it to read my .bash_profile
> - they're different files for a reason). It's annoying me that I
> always have to . .profile after logging in. Is there any way 
> I can automate this?

What shell are you using, and is it one that normally runs .profile?

One way to do things differently on different computers is to have it
check what os/hostname it is, and run a different file depending on
what it comes up with.  Here's a way that works in both bash and csh:

in the 'main' script (say, .profile), have these two lines:
---cut---
source script.`uname`
source script.`hostname`
---cut---

(and you can of course have stuff that you want to do all the time
in the main file as well)
the `` (backticks) mean "run this program and put the results here",
so if I was on Linux, running uname gives me "Linux", so that would
turn into "source script.Linux", and if I was on solaris, running
uname gives me "SunOS", so that would turn into "source script.SunOS".
Of course, change the "script" part of it to something that makes sense,
like .profile if you're doing your profile.[1]

So now, for each system you need os- or host- specific stuff in[2], create
a file for the script.`uname` and script.`hostname`.  One easy way to do
this is to actually log in to the system you want it for, and do a
emacs/vi/cat > or whatever literally script.`uname`, which prevents
you from making any typos.  You might also want to do an
echo script.`uname` beforehand, just to make sure it makes sense.

One disadvantage is that if you don't have a system-specific file for
a particular system it'll give an error.  I actually like this, because
it lets me see if I'm missing any systems.

hope this made sense and/or was useful..


[1] an alternative is to put the specific files in a different directory,
like .profiles/ (you can probably figure out how to call them from the
above)

Another alternative is to run the files instead of sourcing them.  This
probably won't work very well for .profile kinds of things, but may work
for other scripts, especially if for some reason you wrote them in
different languages, or if some of them are symlinks to the 'real'
programs and some of them are programs you wrote.

[2] if you only want to do it on hostnames or only want to do it on
unames, you can obviously just take the line out.  You can also do it
with other things, like possibly whether your home dir is local or
nfs'd (I don't know if or how to do this, I haven't used nfs much)


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



RE: [techtalk] Telnet, solaris and .profile

2000-08-17 Thread Britta Koch

On 17 Aug, Fan, Laurel wrote:

> 
> What shell are you using, and is it one that normally runs .profile?

I'm using the bash on a Solaris 5.6 machine. But it doesn't read
..bash_profile, either. Maybe I'll have to look in /etc/passwd or the
Solaris equivalent.
> 
> One way to do things differently on different computers is to have it
> check what os/hostname it is, and run a different file depending on
> what it comes up with.  Here's a way that works in both bash and csh:
> 
> in the 'main' script (say, .profile), have these two lines:
> ---cut---
> source script.`uname`
> source script.`hostname`
> ---cut---
> 

Now that's an elegant solution! Thanx!

Britta
-- 
  /"\  . . . . . . . . . . . . . [EMAIL PROTECTED] . . . . . . . . . . . . .
  \ /   ASCII Ribbon Campaign "Wer mit der Speckseite nach der Wurst wirft,
   X- NO HTML/RTF in e-mailmuss sich nicht wundern, wenn sie dem Fass
  / \   - NO Word docs in e-mail   die Krone ins Gesicht schlägt."
 




___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk



[techtalk] I asked my resident alphageek...

2000-08-17 Thread J-Mag Guthrie

Date: Thu, 17 Aug 2000 09:56:54 -0500 (CDT)
From: Jonathan Guthrie <[EMAIL PROTECTED]>
To: J-Mag Guthrie <[EMAIL PROTECTED]>
Subject: Re: Check this out

On Thu, 17 Aug 2000, J-Mag Guthrie wrote:

> http://www.abcnews.go.com/sections/tech/FredMoody/moody.html

Oh, okay.  He's the guy who added up all the columns listing defect counts
for Linux distributions, including "aggregate" and concluded that Linux
was the "worst operating system ever".

He's got a lot of nerve complaining about people who do "not seem to
read, understand, or otherwise know how to employ the English language"
when he adds the aggregate total to all the other columns in the
statistics to reach his conclusion.

He's also lying in this column.  He said nothing about relative market
share in his previous message.  He also quoted not ONE "security
expert" as the basis of his conclusion The conclusion was based solely on
the total number of defects reported.

Based on the fact that he used "fighting words" in his original message,
that he doesn't know what he's talking about, and the fact that he lies
through his teeth about the focus of the previous article, I have to
conclude that he's a moron.
-- 
Jonathan Guthrie ([EMAIL PROTECTED])
Brokersys  +281-580-3358   http://www.brokersys.com/
12703 Veterans Memorial #106, Houston, TX  77014, USA



___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk