Network unreachable - SLS 1.03 - NET-2

Network unreachable - SLS 1.03 - NET-2

Post by Sait Um » Sat, 14 Aug 1993 23:45:32



I went through the latest version of SLS 1.03 (including the fixes put
in AUg. 12 at 10 am). This version installed flowlessly and went through
the menu driven network setup part. I configured all the files I could
see, host, hosts, HOSTNAME, resolv.conf, rc.net ... Everthing was working
before under SLS 1.02. When the boot goes through rc.net every external
route command gives SIOOO: Network unreachable error. I have the net
round before gateway... Can someone ship me configuration files that
work. I went through the NET-2 FAQ as well.

Thanks,
--
-------------------------------------------------------------------------


Tel: (615) 322-2459                     Vanderbilt University
Fax: (615) 343-7263                     Nashville, TN 37235
-------------------------------------------------------------------------

 
 
 

1. NET-2 / SLS 1.03 network routing problems: partial diagnosis

I've been having trouble getting IP routing to work properly.  I've seen
several symptoms:

[1] If I issue up a "route add default eth0 gw router", and then a
    "netstat -r" command, the default route shows up without the
    router's IP address, and the G (gateway) flag is not set.

[2] Sometimes, when I issue this command sequence, the default route
    appears with the "lo" loopback device specified, rather than eth0.

[3] If the route is set up with eth0 as the interface, and if I try to
    telnet to a machine which is accessible via the gateway, then I
    observe Linux issuing ARP packets for the destination machine's IP
    address, not for the gateway's (this is consistent with the G flag
    not being set properly).

These symptoms appear to have a common cause.  In looking through the
NET-2 version of the route.c command, I've noticed that this command is
completely ignoring the argument which contains the interface device
name!  It is not filling in the rt_dev field in the struct route at all.
This causes two problems... the rt_dev is garbage down in the kernel
(causing indeterminate effects) and (because the interface argument
isn't processed) the "gw" option to set the gateway status of this route
isn't parsed and is silently ignored.

There needs to be some way for the route command to pass the interface
name into the rt_add and rt_del services... either by pointer, or by
string-name (with a lookup in the rt_add or rt_del services).

I suspect that this bug in the route command may have been causing some
of the system hangs that people have been reporting.  If you set up
routes in the wrong order, then the default route may be aimed at the
loopback device... which would cause bad things to happen, or at least
would cause many addresses to be unreachable.
--
Dave Platt                                                VOICE: (415) 813-8917

 USNAIL: New Technologies Group Inc. 2470 Embarcardero Way, Palo Alto CA 94303

2. Urgent!! group descriptors corrupted!!

3. network unreachable in SLS 1.03

4. kernel failure when rerunning lilo

5. 3Com 3C509 & SLS 1.03/Net-2

6. Large files > 2GB

7. SLS 1.03 Set-up for NET-2

8. handwriting2ascii

9. SLS 1.03 and NET-2 ?

10. SLS 1.03 and NET-2

11. SLS 1.03 and NET-2 help please!

12. net-2 ALL in SLS 1.03??

13. Is Net 2 & SLIP in SLS 1.03 on sunsite.unc.edu