Well the first thing "if i might chip in here" is the nettools version, of
course that should not cause linux to lock up but it will have nasty
effects, of which have been discussed here on this list before.
nettools-1.33 patched with net-tools-1.33.rose.tjd.diff will help in many
ways.
>
> >
> >
> > following lines in rc.local:
> > /sbin/arp -H netrom -s 44.144.90.5 ON4AAK -i nr0
>
> Is this necessary? (I'm not using NETROM, so I can't tell).
Very adviseable.
>
> > /sbin/route add -host 44.144.90.5 dev nr0
> > /sbin/route add -net 44.144.9.0 netmask 255.255.255.0 gw 44.144.90.5
>
> Ehh, wait... Your gateway is not part of the subnetwork? I've never tried
> this... Isn't 255.255.0.0 an option as netmask? Or even 255.0.0.0? Then your
> gateway is on the subnet as well....Perhaps a look at the routing table might
> help here...
>
This could be where nettools-1.33 will do beter, i for one have had a
problem (maybe it has nothing to do with Guy's problem) with ICMP redirect
packets, caused by having the hidden transmitter syndroom, by this i mean
3 stn's on the same freq, who could not work each other directly, i was
using a netmask which was unaplickable, 255.0.0.0 so there was many unwanted
packets flying around, i now use netmask 255.255.255.255 which solves this
problem. Thanks to Eddy PE1CIG who pointed that one out to me.
> > 44.144.9.13 can ping, telnet,ftp,... to 44.144.0.2
>
> Directly? Or is this going through the gateway 44.144.90.5?
>
> >
>
> > but as soons as 44.144.9.13 goes to for example 44.144.0.1 .... on4ufa
>
> > crashes complete!!
> > a ping to 44.144.0.1 works but as soons as there is a bit of data.... all
> > locks up.
>
> What kind of data? Ping sends data as well...
>
> > The only thing you can do at that moment is switch off the computer...!!!
>
> As I said, I'm not into NETROM, but why would the traffic from
> 44.144.0.1->44.144.9.13 have anything to do with on4ufa?? I can't think of any
> special reason why this would even bother on4ufa. Have I missed something
> here? What about their IP routing tables (the 0.1 and 9.13 nodes...)? Are they
> using on4ufa as their gateway? Apart from this, on4ufa doesn't seem to have a
> route to the 44.144.0.0 network, but that might have been created
> automagically somewhere. Again, a look at IP routing tables would help here.
> If there is no route to 44.144.0.0, then that certainly seems an error to me.
>
As far as i see it no way should anything mentioned here cause kernel(s)
2.0.29<>33 to crash, i have and still use them, and have no problems like
Guy's.
> >
> >
> > So the crashes only occur when there is data comming in/out via netrom and
> > going out via encap on the ethernet!!
What i do have a problem with is interfaces just dissapiering, ie, half
dying, i can see packet activity using listen, but call just does not do
anything, i have mentioned this problem here before, and i know that PE0MAR
has the same problem on his ipip interface.
> >
> > Has anyone seen something like this?
> > Could the mss value play any part in this??
>
> MaxSegmentSize only applies to TCP connections (telnet, ftp); ping uses ICMP.
> The one thing that binds them is the MaxTransferUnit on the interface, since
> that applies to all IP traffic.
> You might need to tweak that one with ifconfig...
> On the other hand, you said the problems started with transport of 'data'. If
> you mean using telnet/ftp instead of ping, it could indeed indicate TCP
> problems. What happens when you ping large frames between the machines (ping
> -s 512 or something like that...)?
I have also experimented with this sort of thing, i can get linux to do all
sorts of things ie, fragmentaion, etc, but i cannot get it to crash.
>
> >
> >
> > btw. optimize as router is switched on in the kernel. and always defragment
> > is switched on.
> >
I have used that option also, to be honest i have never seen any differance.
> > I already tried 2.0.31/32 but they had the same problem...
>
> (Sick) You might try 2.0.33...
I cannot get 2.0.33 to crash when doing what Guy explanes.
>
> >
> >
> > Any assistance would be much appreciated !!!!
> >
Well this mail is probably more confusing than assisting, but maybe it
contains some hints in it somewhere.
Another point, when PI1VLI was linked to Gent, i used in PI1GOE ip routing
via netrom, i never had any problems as i can recall, that was with kernel
2.0.29.
> > 73's the Guy (cgfei.dmkbdngjtb@kerailya.tunkki.fi)
>
> Good luck (btw, great job on your site),
> Jan, pa3gyf
>
Same here, and we do notice here in Zeeland when on4ufa is not presant.
-- Regards Richard. rvdlla@roskapostia.tunk.net