[tulip] Re: 192.168.1.0 network weird?

[John Sutton]

OK, fixed it now.  My omission was /etc/nsswitch.conf.  Why this omission
should result in the bizarre behaviour described below is surely
symptomatic of a bug somewhere?  If not, it's one helluvah "feature"!

On Wed, 04 Sep 2002, you wrote:
> Hi there
> 
> Sorry this is way OT but I'm desperate...
> 
> I'm building a cut down system as a rescue system to boot off DiskOnChip
> and I've hit a really extraordinary gotcha...
> 
> I started from a RH7.1 rescue image but running with a 2.4.18 kernel on a
> box with two tulip cards.
> 
> I've cut the system right down to try to isolate the problem so I'm doing
> name resolution using /etc/hosts only (so host.conf contains only the
> single line "order hosts") and /etc/hosts itself contains only the single
> line:
> 
> 127.0.0.1   localhost
> 
> After booting the kernel and bringing up the lo interface:
> 
> ping localhost     - works
> 
> ifconfig eth0 192.168.0.1
> 
> ping localhost     - works
> 
> ifconfig eth0 192.168.1.1
> 
> ping localhost
> ping: unknown host localhost    - name resolution has failed!
> 
> ifconfig eth0 192.168.2.1
> 
> ping localhost     - works
> 
> 
> In summary, I can configure the interface with any IP 192.168.X.Y
> and everything is fine *except* with an IP from the .1.0 network!  If I
> choose one of these IP's, name resolution fails.  Note that the interface
> is actually up and so I *can* ping 127.0.0.1 and 192.168.1.1 but not using
> a name - /etc/hosts is completely ignored until I either bring the
> interface down or reconfigure it with another IP *not* in the .1.0 network.
> 
> I can't really believe what I've just written and would expect any sane
> person to just call me a nutter and ignore this post ;-(  But it's true!
> 
> I've tried it on 2 different machines with the same result.  And to add
> insult to injury, if I boot this *same* kernel but give it a root device
> containing a full blown RH 7.1 system, then the problem disappears.
> 
> So it appears that I have cut something out and in so doing exposed some
> "special significance" of the 192.168.1.0 network?  But where is it?  My
> first thought is that it has to be in some library code that I've removed
> from the rescue system, so I reverted the rescue system to use the original
> /lib and /usr/lib but the problem still persists.
> 
> I'm completely baffled!  Any ideas?
> 
> TIA
> 
> ***************************************************
> John Sutton
> SCL Internet
> URL http://www.scl.co.uk/
> Tel. +44 (0) 1239 711 888
> ***************************************************
-- 

***************************************************
John Sutton
SCL Internet
URL http://www.scl.co.uk/
Tel. +44 (0) 1239 711 888
***************************************************