On Thu, 16 Dec 2004, Antoine Martin spake:
> Looks to me like this is a routing problem, check your arp tables and
> your routes. It may be that something is occasionally causing a packet
> to come out of the wrong interface/wrong mac, which confuses the lower
> network layers.

Alas, I don't think it's that simple :(

My arp tables are simple; virtually everything has the MAC of the ADSL
router.

Every packet going out of the wrong interface gets syslogged, and I see
no such syslog messages, just heaps of NFS timeout messages.

(See <http://www.esperi.demon.co.uk/nix/iptables> for my firewall rules;
a few sensitive external IP addresses have been elided. Look at the log
rules therein.)

ip addr:

1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue 
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 brd 127.255.255.255 scope host lo
2: teql0: <NOARP> mtu 1500 qdisc noop qlen 100
    link/void 
3: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop 
    link/ether 00:00:00:00:00:00 brd ff:ff:ff:ff:ff:ff
4: gordianet: <BROADCAST,MULTICAST,ALLMULTI,UP> mtu 1458 qdisc pfifo_fast qlen 
100
    link/ether 00:60:97:79:e2:c1 brd ff:ff:ff:ff:ff:ff
    inet 192.168.14.1/24 brd 192.168.14.255 scope global gordianet
5: adsl: <BROADCAST,MULTICAST,ALLMULTI,UP> mtu 1458 qdisc cbq qlen 10
    link/ether 00:60:97:79:e2:c1 brd ff:ff:ff:ff:ff:ff
    inet 194.247.41.52/24 brd 255.255.255.255 scope global adsl
    inet 192.168.14.159/32 scope global adsl


ip route:

192.168.14.160 dev adsl  proto static  scope link  src 192.168.14.159 
194.247.41.0/24 dev adsl  proto kernel  scope link  src 194.247.41.52 
192.168.14.0/24 dev gordianet  proto kernel  scope link  src 192.168.14.1 
224.0.0.0/4 dev gordianet  scope link 
default via 194.247.41.52 dev adsl 


(note an arcaneness: the ADSL router is on 192.168.14.160, the same
subnet as all the stuff on the other interface: it's accessed via a
different source address (to avoid being masqueraded).


... hm. Some more info. This is definitely an interface-specific thing.
The blockages appear to affect different interfaces at different times:
i.e., one interface freezes, in both directions --- and it's not just
one destination, it's everyone except for myself. e.g., pinging the
broadcast address on the local net:

64 bytes from 192.168.14.1: icmp_seq=343 ttl=64 time=9.7 ms
64 bytes from 192.168.14.16: icmp_seq=342 ttl=64 time=5010.4 ms (DUP!)
64 bytes from 192.168.14.14: icmp_seq=342 ttl=64 time=5011.0 ms (DUP!)
64 bytes from 192.168.14.18: icmp_seq=342 ttl=64 time=5011.5 ms (DUP!)

I'm starting to suspect the host's TUN/TAP here, y'know: the fact that
it's a simultaneous bidirectional freeze, and that the packet to myself
(automatically sent directly to myself by the kernel) is *not* delayed
is possibly significant.

(But there were no TUN/TAP changes between 2.4.26 and 2.4.28. Damn.
Maybe it's TCP Vegas or something? *clutches at straws in
ChangeLog-2.4.2[78]*)

-- 
`The sword we forged has turned upon us
 Only now, at the end of all things do we see
 The lamp-bearer dies; only the lamp burns on.'


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user

Reply via email to