Discussion:
CTC errors ipl'ing Debian Linux/390
(too old to reply)
Peter J Farley III
2003-01-13 01:22:41 UTC
Permalink
Hi all,

I've recently dl'ed Matt Zimmerman's Debian Linux/390 disk image. It
IPL's nicely, but I'm not getting the CTC definitions right to provide
the L/390 access to (at least) my host (which is RH7.3), and me access
to the L/390 image with telnet or ssh or whatever runs there by
default. I get errors at ipl on ctc0 (see below), and I need help
figuring out what I'm doing wrong. I'm trying this with a slightly
modified herc 2.16.5 (curses changes to panel.c).

My RH system is behind a router, so its address is 192.168.1.100.
Here's my modified copy of Matt's hercules.cnf file (sorry for line
wrap):

CPUSERIAL 002623 # CPU serial number
CPUMODEL 3090 # CPU model number
MAINSIZE 64 # Main storage size in megabytes
XPNDSIZE 0 # Expanded storage size in megabytes
OSTAILOR LINUX # OS tailoring
ARCHMODE ESA/390 # Architecture mode S/370, ESA/390 or ESAME

0009 3215

0300 3390 dasd/Debian-3.0r1.3390
0400 3088 CTCI /dev/net/tun 1500 192.168.10.1 192.168.10.2
255.255.255.0
0401 3088 CTCI /dev/net/tun 1500 192.168.10.1 192.168.10.2
255.255.255.0

Here's my /etc/hosts file on the base RH system (ISP info changed):

# Do not remove the following line, or various programs
# that require network functionality will fail.
127.0.0.1 localhost.localdomain localhost
192.168.1.100 xxxxx.ISP.net xxxxx
192.168.10.1 debian

Here are the errors I get when IPL'ing Debian under Hercules:

Configuring network interfaces:
CTC driver Version: 1.55 with CHANDEV support initialized
SIOCSIFADDR: No such device
ctc0: ERROR while getting interface flags: No such device
SIOCSIFNETMASK: No such device
SIOCSIFDSTADDR: No such device
ctc0: ERROR while getting interface flags: No such device
ctc0: ERROR while getting interface flags: No such device
done.

TIA for any help, info, RTFM you can provide.

Peter


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Matt Zimmerman
2003-01-13 02:06:06 UTC
Permalink
Post by Peter J Farley III
I've recently dl'ed Matt Zimmerman's Debian Linux/390 disk image. It
IPL's nicely, but I'm not getting the CTC definitions right to provide
the L/390 access to (at least) my host (which is RH7.3), and me access
to the L/390 image with telnet or ssh or whatever runs there by
default. I get errors at ipl on ctc0 (see below), and I need help
figuring out what I'm doing wrong. I'm trying this with a slightly
modified herc 2.16.5 (curses changes to panel.c).
Try these:

#0A00 3088 CTCI /dev/net/tun 2000 192.168.10.2 192.168.10.1 255.255.255.252
#0A01 3088 CTCI /dev/net/tun 2000 192.168.10.2 192.168.10.1 255.255.255.252

I've added those lines to the example configuration file next to the DASD
images.
--
- mdz

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Peter J Farley III
2003-01-13 04:08:54 UTC
Permalink
--- In hercules-390-***@public.gmane.org, Matt Zimmerman <***@c...> wrote:
<Snipped>
Post by Matt Zimmerman
#0A00 3088 CTCI /dev/net/tun 2000 192.168.10.2 192.168.10.1 255.255.255.252
#0A01 3088 CTCI /dev/net/tun 2000 192.168.10.2 192.168.10.1 255.255.255.252
I've added those lines to the example configuration file next to
the DASD images.
Thanks, Matt. Those worked nicely. Connectivity established. BTW,
why a netmask with .252 on the end instead of .0, like the herc TCP/IP
page suggests?

Next newbie question: There's no sshd nor telnetd running in the
L/390 image, and I can't find a binary for either of them anywhere.
How is one supposed to login to the system (other than the Herc
console, of course)? I know I'm missing something, just don't know
what. Help appreciated.

My suspicion: The configuration of the system is bare minumum, and
bunches of configuration and/or apt-getting (including allocating and
initializing additional dasd to hold it all) remain to be done before
a usable development system is available. Right or wrong?

If my suspicion is right, where do I read up on how to proceed? I do
have Debian-3/390 CD's, so if it's on there, just point me to it. And
I'm sure I'm going to need help establishing the L/390 image's access
to the outside world through my base system.

What I want to achieve in the end is an L/390 C/C++ development system
with all the GNU tools available. Don't need webservers and such yet,
just a programming development platform I can update as needed and
access to it with ssh or telnet, and a fast way to move files between
the L/390 dasd and my base RH system.

Thanks again for your help.

Peter


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Matt Zimmerman
2003-01-13 04:42:03 UTC
Permalink
Post by Peter J Farley III
<Snipped>
Post by Matt Zimmerman
#0A00 3088 CTCI /dev/net/tun 2000 192.168.10.2 192.168.10.1
255.255.255.252
Post by Matt Zimmerman
#0A01 3088 CTCI /dev/net/tun 2000 192.168.10.2 192.168.10.1
255.255.255.252
Post by Matt Zimmerman
I've added those lines to the example configuration file next to
the DASD images.
Thanks, Matt. Those worked nicely. Connectivity established. BTW,
why a netmask with .252 on the end instead of .0, like the herc TCP/IP
page suggests?
It is a point-to-point link, with only two host addresses (.1 and .2), so
the logical netmask is 255.255.255.252.
Post by Peter J Farley III
Next newbie question: There's no sshd nor telnetd running in the L/390
image, and I can't find a binary for either of them anywhere. How is one
supposed to login to the system (other than the Herc console, of course)?
I know I'm missing something, just don't know what. Help appreciated.
Because the system console cannot be used to configure the base system,
Debian/390 installs with telnetd up and running by default (running from
inetd). If you telnet to the system after it finishes booting, the base
system configuration sequence will begin immediately, and you will be asked
a series of questions, enabling you to set the root password, create a user
account, install additional software, etc.
Post by Peter J Farley III
My suspicion: The configuration of the system is bare minumum, and
bunches of configuration and/or apt-getting (including allocating and
initializing additional dasd to hold it all) remain to be done before
a usable development system is available. Right or wrong?
Yes, but the configuration process more or less walks you through this
process.
Post by Peter J Farley III
If my suspicion is right, where do I read up on how to proceed? I do have
Debian-3/390 CD's, so if it's on there, just point me to it. And I'm sure
I'm going to need help establishing the L/390 image's access to the
outside world through my base system.
What I want to achieve in the end is an L/390 C/C++ development system
with all the GNU tools available. Don't need webservers and such yet,
just a programming development platform I can update as needed and access
to it with ssh or telnet, and a fast way to move files between the L/390
dasd and my base RH system.
ssh is standard, and will be installed by default unless you explicitly
request otherwise. For development, simply select the "C/C++ Development"
task when tasksel runs (the "Select tasks to install" screen).

Since you are new to Debian, I suggest that you do NOT opt to run dselect
under any circumstances. This is the default, but you will be prompted.
--
- mdz

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Willem Konynenberg
2003-01-13 11:19:29 UTC
Permalink
Post by Matt Zimmerman
It is a point-to-point link, with only two host addresses (.1 and .2), so
the logical netmask is 255.255.255.252.
Actually, the point-to-point device itself has effectively no
netmask at all (which is roughly equivalent to 255.255.255.255),
because it is, by itself, not a "(sub)net", and the corresponding
route to the remote end should generally really have a mask of
255.255.255.255, to make it a host route, rather than a net route.

The two endpoint addresses of the point-to-point link need not even
be in the same subnet. It is quite possible to set up a link with
10.42.1.100 on one end, and 192.168.10.200 on the other end.
No "subnet mask" would make any sense on such a link.
(You may, if you wish, logically assign a 4-address subnet to the link
(which would then, logically, have a subnet mask 255.255.255.252),
and assign the two available addresses from it to the two end points,
but even then I still think it is better to use a netmask of
255.255.255.255 for the actual config commands (if only to avoid
confusion).
When you assign two addresses from an existing larger subnet range,
then it is definitely *wrong* to use the subnet mask of the existing
subnet for the PtP link, because the guest end of the link *cannot*
directly reach the other addresses in the subnet. It must route
those to the host end of the link. Thus, both ends of the link
need to have a host route, rather than a net route for the PtP link
to make it work. (having part of this wrong may actually happen to
work, but that doesn't make it right... ;-))

In the case of Peter's config, he could, if he wanted, reuse the
192.168.1.100 address of his host system for the host-end of the
CTC-TUN PtP link, and use only a single, lone IP address for the
guest system, e.g. 192.168.10.1 or 10.1.1.1, or whatever.
As long as he uses a netmask of 255.255.255.255, the routing should
come out just right.
--
Willem Konynenberg <w.f.konynenberg-/NLkJaSkS4VmR6Xm/***@public.gmane.org>
Konynenberg Software Engineering

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Matt Zimmerman
2003-01-13 15:14:25 UTC
Permalink
Post by Matt Zimmerman
It is a point-to-point link, with only two host addresses (.1 and .2),
so the logical netmask is 255.255.255.252.
Actually, the point-to-point device itself has effectively no netmask at
all (which is roughly equivalent to 255.255.255.255), because it is, by
itself, not a "(sub)net", and the corresponding route to the remote end
should generally really have a mask of 255.255.255.255, to make it a host
route, rather than a net route.
The two endpoint addresses of the point-to-point link need not even
be in the same subnet. It is quite possible to set up a link with
10.42.1.100 on one end, and 192.168.10.200 on the other end.
No "subnet mask" would make any sense on such a link.
(You may, if you wish, logically assign a 4-address subnet to the link
(which would then, logically, have a subnet mask 255.255.255.252),
and assign the two available addresses from it to the two end points,
but even then I still think it is better to use a netmask of
255.255.255.255 for the actual config commands (if only to avoid
confusion).
On the contrary, I consider this practice to be less confusing, as the
netmask serves as a bit of documentation. Given the local IP address and
the netmask, the correct remote IP address is obvious; this helps to
recognize and avoid misconfiguration. I didn't even think that Linux used
the netmask value of a point-to-point interface for any purpose, though now
that I look, it does.

This is a common practice anyway, because some popular router vendors do not
have the same "point to point" interface concept as many UNIX derivatives
do, and use netmasks universally to calculate routing information, which is
where I (at least) developed this habit. And in this case, .248 would serve
just as well, though 255.255.255.255 is arguably more correct.
When you assign two addresses from an existing larger subnet range,
then it is definitely *wrong* to use the subnet mask of the existing
subnet for the PtP link, because the guest end of the link *cannot*
directly reach the other addresses in the subnet. It must route
those to the host end of the link. Thus, both ends of the link
need to have a host route, rather than a net route for the PtP link
to make it work. (having part of this wrong may actually happen to
work, but that doesn't make it right... ;-))
Yes, but this is not at all what I was doing of course (since there are no
other generally usable host addresses on a .252 subnet).
In the case of Peter's config, he could, if he wanted, reuse the
192.168.1.100 address of his host system for the host-end of the
CTC-TUN PtP link, and use only a single, lone IP address for the
guest system, e.g. 192.168.10.1 or 10.1.1.1, or whatever.
As long as he uses a netmask of 255.255.255.255, the routing should
come out just right.
He could in fact do this, but I would prefer to ship an example
configuration file which uses separate addresses for the point to point
interface, to avoid having to predict what address the user is using for one
of their non-loopback interfaces in order to have a working example. :-)
--
- mdz

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Willem Konynenberg
2003-01-13 16:04:53 UTC
Permalink
Post by Matt Zimmerman
Post by Willem Konynenberg
but even then I still think it is better to use a netmask of
255.255.255.255 for the actual config commands (if only to avoid
confusion).
On the contrary, I consider this practice to be less confusing, as the
netmask serves as a bit of documentation. Given the local IP address and
the netmask, the correct remote IP address is obvious; this helps to
recognize and avoid misconfiguration. I didn't even think that Linux used
the netmask value of a point-to-point interface for any purpose, though now
that I look, it does.
That's one reason I got cautious about this: the device itself will
happily ignore the netmask as it knows it is simply Point-to-Point,
but the standard configuration utilities will use the same netmask
to set up routing, and here anything other than 255 (which generally
works for any set of IP addresses) or 252 (which works for two IP
addresses from a 252 subnet) may cause routing problems.
Post by Matt Zimmerman
Post by Willem Konynenberg
In the case of Peter's config, he could, if he wanted, reuse the
192.168.1.100 address of his host system for the host-end of the
CTC-TUN PtP link, and use only a single, lone IP address for the
guest system, e.g. 192.168.10.1 or 10.1.1.1, or whatever.
As long as he uses a netmask of 255.255.255.255, the routing should
come out just right.
He could in fact do this, but I would prefer to ship an example
configuration file which uses separate addresses for the point to point
interface, to avoid having to predict what address the user is using for one
of their non-loopback interfaces in order to have a working example. :-)
Yes, naturally. I also ran into situations where I found it more
convenient to give each end of the link its own separate address,
for example so I could move a configuration between hosts easily.

The key point I am trying to get across is that for Point-to-Point
links, such as the CTC-TUN link is, the common LAN Subnet rules of
thumb for IP configuration decisions do not always work.
You have more flexibility in your choice of IP addresses, but you
also need to be extra careful not to mix up your netmasks, or you
will end up having routing problems.
PtP and LAN devices are different, but the distinction may be
hidden by some implementations that handle PtP devices "as if"
they were LAN devices. Linux does make the distinction, and
as a result is perhaps more sensitive to the mixups that can
result if you aren't aware of this.
--
Willem Konynenberg <w.f.konynenberg-/NLkJaSkS4VmR6Xm/***@public.gmane.org>
Konynenberg Software Engineering

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Marian Gasparovic
2003-01-14 16:35:27 UTC
Permalink
Hi all,
could anyone send me working routing configuration for
LCS Hercules device ? I am able to ping LCS device
from Linux where Hercules runs but I am not able to
ping anything from guest OS.
How do working routing tables look like on both host
and guest OS's ?
Thank you

=====
===================
Marian Gasparovic
===================
"The mere thought hadn't even begun to speculate about the merest possibility of crossing my mind."



__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Marian Gasparovic
2003-01-17 14:31:51 UTC
Permalink
Post by Peter J Farley III
Hi all,
could anyone send me working routing configuration
for
LCS Hercules device ? I am able to ping LCS device
from Linux where Hercules runs but I am not able to
ping anything from guest OS.
How do working routing tables look like on both host
and guest OS's ?
Nobody ? And is there please anybody who could tell me
what is wrong with my config ?
I run current cvs Hercules version on Mandrake 9
Linux.

hercules.cnf:
0602 LCS --dev /dev/net/tun --oat oatfile

oatfile:
0602 IP 00 PRI 192.168.102.2
0604 IP 01 SEC 192.168.102.3
HWADD 00 11:11:11:11:11:11
HWADD 01 12:12:12:12:12:12

after I start Hercules, devlist tells me
0602 3088 LCS Port 00 IP (tap0) open busy
0603 3088 LCS Port 00 IP (tap0) open
0604 3088 LCS Port 01 IP (tap1) open busy
0605 3088 LCS Port 01 IP (tap1) open

TCPIP PROFILE:
DEVICE CTC1 LCS 602
LINK CTC602 ETHERNET 0 CTC1
DEVICE CTC2 LCS 604
LINK CTC604 ETHERNET 1 CTC2
(yes, I named them ctc, ignore it)
HOME
192.168.102.2 CTC602
192.168.102.3 CTC604
BEGINRoutes

ROUTE 192.168.4.0 255.255.252.0 = CTC602 mtu 1500
ROUTE 192.168.5.50 HOST = CTC602 mtu 1500
ROUTE Default 192.168.5.50 CTC602 mtu 1500
ENDRoutes

d tcpip,,n,route gives

DEFAULT 192.168.5.50 UG 000000 CTC602
127.0.0.1 0.0.0.0 UH 000002 LOOPBACK
192.168.4.0 0.0.0.0 U 000000 CTC602
192.168.5.50 0.0.0.0 UH 000000 CTC602
192.168.102.2 0.0.0.0 UH 000000 CTC602
192.168.102.3 0.0.0.0 UH 000000 CTC604


So I have 5.50 which is Linux adapter to our network
and 102.2,102.3 which are two adapters for MVS. When I
start TCPIP I get no errors and I can ping both 102.2
and 102.3 from Linux. From MVS I can ping itself
(102.3 and 102.3) and 5.50. Anything else timeouts me.
And real fun is when I try TRACERT from MVS - it works
for 102.x but if FAILS for 5.50 although ping works.
What can be wrong ? Thank you


=====
===================
Marian Gasparovic
===================
"The mere thought hadn't even begun to speculate about the merest possibility of crossing my mind."



__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs Online-No late fees! Try Netflix for FREE!
http://us.click.yahoo.com/OdoN7C/oEZFAA/46VHAA/W4wwlB/TM
---------------------------------------------------------------------~->
w_hecox
2003-01-17 16:44:19 UTC
Permalink
--- In hercules-390-***@public.gmane.org, Marian Gasparovic <***@y...>
wrote:

I always thought the default route had to be in the same subnet
as the home statement. You could try changing your home
statements to the same subnet as your real ethernet device.
You will then need to do proxy arp so that other devices
will be able to reach your os host.
I run the following script to enable proxy-arp:
---------------------------------------------------
arp -i eth0 -s 192.168.1.109 -D eth0
arp
printf 1 >/proc/sys/net/ipv4/conf/eth0/proxy_arp
cat /proc/sys/net/ipv4/conf/eth0/proxy_arp
printf 1 >/proc/sys/net/ipv4/ip_forward
cat /proc/sys/net/ipv4/ip_forward
netstat -r
----------------------------------------------------
my linux eth0 address is 192.168.1.101 mask 255.255.255.0
my host profile is:
HOME
192.168.1.109 ETH1
;
GATEWAY
192.168.1 = ETH1 1500 0
192.168.1.101 = ETH1 1500 HOST
;
defaultnet 192.168.1.101 ETH1 1500 0

Regards,
Bill
Post by Marian Gasparovic
Post by Peter J Farley III
Hi all,
could anyone send me working routing configuration
for
LCS Hercules device ? I am able to ping LCS device
from Linux where Hercules runs but I am not able to
ping anything from guest OS.
How do working routing tables look like on both host
and guest OS's ?
Nobody ? And is there please anybody who could tell me
what is wrong with my config ?
I run current cvs Hercules version on Mandrake 9
Linux.
0602 LCS --dev /dev/net/tun --oat oatfile
0602 IP 00 PRI 192.168.102.2
0604 IP 01 SEC 192.168.102.3
HWADD 00 11:11:11:11:11:11
HWADD 01 12:12:12:12:12:12
after I start Hercules, devlist tells me
0602 3088 LCS Port 00 IP (tap0) open busy
0603 3088 LCS Port 00 IP (tap0) open
0604 3088 LCS Port 01 IP (tap1) open busy
0605 3088 LCS Port 01 IP (tap1) open
DEVICE CTC1 LCS 602
LINK CTC602 ETHERNET 0 CTC1
DEVICE CTC2 LCS 604
LINK CTC604 ETHERNET 1 CTC2
(yes, I named them ctc, ignore it)
HOME
192.168.102.2 CTC602
192.168.102.3 CTC604
BEGINRoutes
ROUTE 192.168.4.0 255.255.252.0 = CTC602 mtu 1500
ROUTE 192.168.5.50 HOST = CTC602 mtu 1500
ROUTE Default 192.168.5.50 CTC602 mtu 1500
ENDRoutes
d tcpip,,n,route gives
DEFAULT 192.168.5.50 UG 000000 CTC602
127.0.0.1 0.0.0.0 UH 000002 LOOPBACK
192.168.4.0 0.0.0.0 U 000000 CTC602
192.168.5.50 0.0.0.0 UH 000000 CTC602
192.168.102.2 0.0.0.0 UH 000000 CTC602
192.168.102.3 0.0.0.0 UH 000000 CTC604
So I have 5.50 which is Linux adapter to our network
and 102.2,102.3 which are two adapters for MVS. When I
start TCPIP I get no errors and I can ping both 102.2
and 102.3 from Linux. From MVS I can ping itself
(102.3 and 102.3) and 5.50. Anything else timeouts me.
And real fun is when I try TRACERT from MVS - it works
for 102.x but if FAILS for 5.50 although ping works.
What can be wrong ? Thank you
=====
===================
Marian Gasparovic
===================
"The mere thought hadn't even begun to speculate about the
merest
possibility of crossing my mind."
Post by Marian Gasparovic
__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Marian Gasparovic
2003-01-20 16:30:46 UTC
Permalink
Tried that too, it didn't help. We use supernetting
not subnetting, could that cause a problem ? net
address is 4.0 with netmask 255.255.252.0
BTW if I don't use LCS but CTCI, it works nicely, so
some routing here works :) And when I use
point-to-point I can imagine how the network looks
like but in LCS config I have problems to imagine how
all layers (our network - host os - Hercules - guest
os) interact with each other. I am in no way a network
person.
Is there any help for me or am I totaly lost case ?
Marian
Post by w_hecox
I always thought the default route had to be in the
same subnet
as the home statement. You could try changing your
home
statements to the same subnet as your real ethernet
device.
You will then need to do proxy arp so that other
devices
will be able to reach your os host.
---------------------------------------------------
arp -i eth0 -s 192.168.1.109 -D eth0
arp
printf 1 >/proc/sys/net/ipv4/conf/eth0/proxy_arp
cat /proc/sys/net/ipv4/conf/eth0/proxy_arp
printf 1 >/proc/sys/net/ipv4/ip_forward
cat /proc/sys/net/ipv4/ip_forward
netstat -r
----------------------------------------------------
my linux eth0 address is 192.168.1.101 mask
255.255.255.0
HOME
192.168.1.109 ETH1
;
GATEWAY
192.168.1 = ETH1 1500 0
192.168.1.101 = ETH1 1500 HOST
;
defaultnet 192.168.1.101 ETH1 1500 0
Regards,
Bill
Post by Marian Gasparovic
Post by Peter J Farley III
Hi all,
could anyone send me working routing
configuration
Post by Marian Gasparovic
Post by Peter J Farley III
for
LCS Hercules device ? I am able to ping LCS
device
Post by Marian Gasparovic
Post by Peter J Farley III
from Linux where Hercules runs but I am not able
to
Post by Marian Gasparovic
Post by Peter J Farley III
ping anything from guest OS.
How do working routing tables look like on both
host
Post by Marian Gasparovic
Post by Peter J Farley III
and guest OS's ?
Nobody ? And is there please anybody who could
tell me
Post by Marian Gasparovic
what is wrong with my config ?
I run current cvs Hercules version on Mandrake 9
Linux.
0602 LCS --dev /dev/net/tun --oat oatfile
0602 IP 00 PRI 192.168.102.2
0604 IP 01 SEC 192.168.102.3
HWADD 00 11:11:11:11:11:11
HWADD 01 12:12:12:12:12:12
after I start Hercules, devlist tells me
0602 3088 LCS Port 00 IP (tap0) open busy
0603 3088 LCS Port 00 IP (tap0) open
0604 3088 LCS Port 01 IP (tap1) open busy
0605 3088 LCS Port 01 IP (tap1) open
DEVICE CTC1 LCS 602
LINK CTC602 ETHERNET 0 CTC1
DEVICE CTC2 LCS 604
LINK CTC604 ETHERNET 1 CTC2
(yes, I named them ctc, ignore it)
HOME
192.168.102.2 CTC602
192.168.102.3 CTC604
BEGINRoutes
ROUTE 192.168.4.0 255.255.252.0 = CTC602 mtu 1500
ROUTE 192.168.5.50 HOST = CTC602 mtu 1500
ROUTE Default 192.168.5.50 CTC602 mtu 1500
ENDRoutes
d tcpip,,n,route gives
DEFAULT 192.168.5.50 UG 000000 CTC602
127.0.0.1 0.0.0.0 UH 000002
LOOPBACK
Post by Marian Gasparovic
192.168.4.0 0.0.0.0 U 000000 CTC602
192.168.5.50 0.0.0.0 UH 000000 CTC602
192.168.102.2 0.0.0.0 UH 000000 CTC602
192.168.102.3 0.0.0.0 UH 000000 CTC604
So I have 5.50 which is Linux adapter to our
network
Post by Marian Gasparovic
and 102.2,102.3 which are two adapters for MVS.
When I
Post by Marian Gasparovic
start TCPIP I get no errors and I can ping both
102.2
Post by Marian Gasparovic
and 102.3 from Linux. From MVS I can ping itself
(102.3 and 102.3) and 5.50. Anything else timeouts
me.
Post by Marian Gasparovic
And real fun is when I try TRACERT from MVS - it
works
Post by Marian Gasparovic
for 102.x but if FAILS for 5.50 although ping
works.
Post by Marian Gasparovic
What can be wrong ? Thank you
=====
===================
Marian Gasparovic
===================
"The mere thought hadn't even begun to
speculate about the
merest
possibility of crossing my mind."
Post by Marian Gasparovic
__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up
now.
Post by Marian Gasparovic
http://mailplus.yahoo.com
__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Peter J Farley III
2003-01-13 18:16:03 UTC
Permalink
On Mon, Jan 13, 2003 at 04:08:54AM -0000, Peter J Farley III
Post by Peter J Farley III
<Snipped>
Post by Matt Zimmerman
#0A00 3088 CTCI /dev/net/tun 2000 192.168.10.2
192.168.10.1
Post by Peter J Farley III
255.255.255.252
Post by Matt Zimmerman
#0A01 3088 CTCI /dev/net/tun 2000 192.168.10.2
192.168.10.1
Post by Peter J Farley III
255.255.255.252
<Snipped>
Because the system console cannot be used to configure the base
system, Debian/390 installs with telnetd up and running by default
(running from inetd). If you telnet to the system after it
finishes booting, the base system configuration sequence will
begin immediately, and you will be asked a series of questions,
enabling you to set the root password, create a user
account, install additional software, etc.
OK, so to keep this absolutely clear for me, you're saying I should
be able to (from my RH base system) telnet to the Debian image with
the following command:

telnet 192.168.10.2

I.E., telent to the leftmost address in the CTC definitions (that's
the Debian end, right?).

The reason I question this is that when I do login from the herc
console as root and execute "ps -A", I *DO NOT* see a telnetd or sshd
process running. I do see inetd, so you are saying that telnetd and
sshd are running under inetd as part of the inetd process?

In any case, I'll try this out and report back.

Peter


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Matt Zimmerman
2003-01-13 19:22:07 UTC
Permalink
Post by Peter J Farley III
OK, so to keep this absolutely clear for me, you're saying I should
be able to (from my RH base system) telnet to the Debian image with
telnet 192.168.10.2
I.E., telent to the leftmost address in the CTC definitions (that's
the Debian end, right?).
That is correct. If you had tried it, you would have answered your own
question. :-)
Post by Peter J Farley III
The reason I question this is that when I do login from the herc
console as root and execute "ps -A", I *DO NOT* see a telnetd or sshd
process running. I do see inetd, so you are saying that telnetd and
sshd are running under inetd as part of the inetd process?
Correct, inetd is a service for managing other network daemons. See "man
inetd" and "man inetd.conf".
--
- mdz

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Peter J Farley III
2003-01-13 20:01:43 UTC
Permalink
On Mon, Jan 13, 2003 at 06:16:03PM -0000, Peter J Farley III
Post by Peter J Farley III
OK, so to keep this absolutely clear for me, you're saying I
should be able to (from my RH base system) telnet to the Debian
telnet 192.168.10.2
I.E., telent to the leftmost address in the CTC definitions
(that's the Debian end, right?).
That is correct. If you had tried it, you would have answered your
own question. :-)
Sorry, I'm not at my RH system today, or I would have tried it first.
Post by Peter J Farley III
The reason I question this is that when I do login from the herc
console as root and execute "ps -A", I *DO NOT* see a telnetd or
sshd process running. I do see inetd, so you are saying that
telnetd and sshd are running under inetd as part of the inetd
process?
Correct, inetd is a service for managing other network daemons.
See "man inetd" and "man inetd.conf".
OK, will do. Thanks for your patient help.

Peter


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Peter J Farley III
2003-01-15 06:04:29 UTC
Permalink
<Snipped>
Post by Matt Zimmerman
That is correct. If you had tried it, you would have answered
your own question. :-)
OK, mea culpa. It just works, like you said. I'm getting paranoid
in my advancing years, I suppose.

I got as far as specifying the method of download. I tried http,
that didn't work too well. Tried ftp, couldn't resolve any of the
names. (Can't give you real messages here, not on linux at the
moment, but it just did not work.)

Aha! I says to meself. Well, just ping from the base system and
enter the actual internet address returned for the ftp site. Not
ideal, I know, but workable. At least, it did work, to a point.
Later, one of the further stages started using names again, instead
of the numbers I had entered, so I stopped the process. (Besides, my
spouse needed the machine.) Again, apologies for not having actual
messages, but I think they were all due to lack of an available
nameserver.

Wouldn't it work better to first set up a resolv.conf file with my
ISP's DNS servers specified? A couple of echo commands would take
care of that, no vi required. Could be done from the herc console
even.

What's the "right" way to handle this? I *can* ping my ISP's DNS
servers from a telnet login to L/390, so I *think* I have the
connectivity licked (though I wouldn't bet money on it).

Peter


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/sXBHAA/W4wwlB/TM
---------------------------------------------------------------------~->
Jim Pierson (Hercules)
2003-01-13 16:49:24 UTC
Permalink
When it comes to the technical absolutes of this discussion,
Willem is correct. However there is a side-effect when using
TUN on Linux that make using a netmask that is not
255.255.255.255 "sometimes" desirable.

The TUN code in Linux currently makes a ioctl call adding a
route to the routing table, and uses the netmask for the
device. As Willem has pointed out, a netmask has no meaning
to a point-to-point device, but this route that gets added
automagically by TUN (Linux code, not mine) can be used to
generate a subnet route to multiple stacks behind the guest
OS running on Hercules.

Try this:

CA02 3088 ctci /dev/net/tun 1500 172.21.1.244 172.21.4.34
255.255.255.0
CA03 3088 ctci /dev/net/tun 1500 172.21.1.244 172.21.4.34
255.255.255.0

you will see something like this in your routes:

Destination Gateway Genmask Flags Metric Ref Use
Iface
172.21.4.0 * 255.255.255.0 U 0 0 0
tun0
172.21.0.0 * 255.255.0.0 U 0 0 0
eth0
127.0.0.0 * 255.0.0.0 U 0 0 0
lo
default 172.21.1.1 0.0.0.0 UG 0 0 0
eth0

I dont know what happens on other *nix's when you try this, but
it world under Linux. BSD's implementation of TUN/TAP is a little
different and I've not had the opportunity to see it's behavior.

The new format if the CTCI configuration statement (2.17) does not
have netmask as a required parameter anymore. I do however allow
it as an optional parameter so those using this side-effect can
still continue to do so (or just continue to use the existing style
config statement as it is still supported).

Jim

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Willem Konynenberg
2003-01-13 17:23:15 UTC
Permalink
Post by Jim Pierson (Hercules)
When it comes to the technical absolutes of this discussion,
Willem is correct. However there is a side-effect when using
TUN on Linux that make using a netmask that is not
255.255.255.255 "sometimes" desirable.
The TUN code in Linux currently makes a ioctl call adding a
route to the routing table, and uses the netmask for the
device. As Willem has pointed out, a netmask has no meaning
to a point-to-point device, but this route that gets added
automagically by TUN (Linux code, not mine) can be used to
generate a subnet route to multiple stacks behind the guest
OS running on Hercules.
CA02 3088 ctci /dev/net/tun 1500 172.21.1.244 172.21.4.34
255.255.255.0
CA03 3088 ctci /dev/net/tun 1500 172.21.1.244 172.21.4.34
255.255.255.0
Destination Gateway Genmask Flags Metric Ref Use
Iface
172.21.4.0 * 255.255.255.0 U 0 0 0
tun0
172.21.0.0 * 255.255.0.0 U 0 0 0
eth0
127.0.0.0 * 255.0.0.0 U 0 0 0
lo
default 172.21.1.1 0.0.0.0 UG 0 0 0
eth0
Basically, you are abusing the device route for "general routing
purposes", and it just so happens to work accidentally.
It would probably be better to create two separate routes in your
routing table, like so:
172.21.4.244 * 255.255.255.255 UH 0 0 0 tun0
172.21.4.0 172.21.4.244 255.255.255.0 UG 0 0 0 tun0

Note that the direct device route now only claims direct access to
the remote end of the link (as is technically the case), and the
subnet is routed through the remote end as the gateway router,
just as one would otherwise do when a subnet is reachable through
a router.
--
Willem Konynenberg <w.f.konynenberg-/NLkJaSkS4VmR6Xm/***@public.gmane.org>
Konynenberg Software Engineering

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Jim Pierson (Hercules)
2003-01-13 17:40:13 UTC
Permalink
I Agree.... I was just pointing out a side-effect that I know
people are making use of. It's not right, but it works (for now).

Technically, netmask should never have been a parameter for CTCI,
but it was and it's been used "creatively" by some. Those that
are exploiting the side-effect just need to realize that a later
release of Linux could very well reject this usage and they will
have to deal with it accordingly.

Jim
-----Original Message-----
Sent: Monday, January 13, 2003 12:23 PM
Subject: Re: [hercules-390] Re: CTC errors ipl'ing Debian Linux/390
Basically, you are abusing the device route for "general routing
purposes", and it just so happens to work accidentally.
It would probably be better to create two separate routes in your
172.21.4.244 * 255.255.255.255 UH 0
0 0 tun0
172.21.4.0 172.21.4.244 255.255.255.0 UG 0
0 0 tun0
Note that the direct device route now only claims direct access to
the remote end of the link (as is technically the case), and the
subnet is routed through the remote end as the gateway router,
just as one would otherwise do when a subnet is reachable through
a router.
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Turn flat surfaces into speakers with the Soundbug.
http://us.click.yahoo.com/QWAVSC/onCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Willem Konynenberg
2003-01-13 17:55:21 UTC
Permalink
Post by Jim Pierson (Hercules)
Technically, netmask should never have been a parameter for CTCI,
but it was and it's been used "creatively" by some.
You can blame me for that, as I was the one that put that parameter
in there for VMNET originally, and it just stuck.
Little did I know... ;-)
--
Willem Konynenberg <w.f.konynenberg-/NLkJaSkS4VmR6Xm/***@public.gmane.org>
Konynenberg Software Engineering

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Flexible Keyboard is the ideal accessory for PDA users that are on the move.
http://us.click.yahoo.com/dCBVZC/WnCFAA/xGHJAA/W4wwlB/TM
---------------------------------------------------------------------~->
Continue reading on narkive:
Loading...