Satori 4.0 bug? (ip_finish_output2: No header cache...)

Discussion in 'Sveasoft Firmware' started by sune, Jul 26, 2004.

  1. sune

    sune Network Guru Member

    I use Satori 4.0 stable firmware, but my WRT54G v1.1 crashes about every 24-48 hrs. with the following syslog'ed entry;

    klogd: ip_finish_output2: No header cache and no neighbour!

    I get this error three times in a row then the router stops responding and
    must be restarted.

    Anyone knows what to do about this?

    I have a simple config using dchp on the WAN interface, most administrative stuff is disabled except for telnet, and syslog, also wireless is completely disabled. I do forward some udp and tcp ports however.

    Hope anyone can help.

  2. Anonymous

    Anonymous Guest

    This means you have no ARP entries on your WRT and it can't find where to send the packet.

    I would suspect this is caused either by a wired network outage or a wireless connection drop.

    You may want to enable the AP Watchdog Timer in Satori and have it restart the wireless interface once a day (86400 seconds).
  3. sune

    sune Network Guru Member

    Thanks for your quick reply,

    It must be related to the ethernet somehow since I have wireless disabled completely.

    I have some P2P traffic forwarded (udp and tcp) but I don't think the traffic load is very heavy as my bandwith is limited (512/128). If the ip_conntrack table is full, could this cause this error?

    edit: I use static ip's on my lan so this should not be a dhcp issue.

  4. Anonymous

    Anonymous Guest

    If somehow available memory was exhausted or if there is a memory leak in the connection tracking this could be one explanation of the cause of this. Hard to say for sure.
  5. tim-atl

    tim-atl Network Guru Member

    If you've increased the number of allowed incoming connections in order to get more sources for your P2P app then that is probably your problem. I did the same but it was with a Netgear.

  6. sune

    sune Network Guru Member


    cat /proc/sys/net/ipv4/ip_conntrack_max = 1024

    is this the default value?

    edit: I've now decreased the default 5 day tcp "established" timeout to 1 hour only, as I guess this could easily fill ip_conntrack with dead P2P connections.

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice