[PATCH] netlink: report why a packet is bad

Ian Jackson ijackson at chiark.greenend.org.uk
Mon Dec 26 17:57:59 GMT 2011


Ian Jackson writes ("[PATCH] netlink: report why a packet is bad"):
> Replace
>  <netlink>: bad IP packet from <source>
> with
>  <netlink>: bad IP packet from <source>: reason

I should report the motivation for this patch, really.

At some point on Friday the 23rd, secnet on chiark stopped working.
Looking at the log in ~secnet/live/log showed this:

 22:27:09 sgo/chiark/chiark<->chiark/relativity/zealot: new
                   key activated [this is entirely normal]
 23:18:13 nl-ipif: bad IP packet from host [repeated dozens of times]
 23:41:31 nl-ipif: bad IP packet from host [only once]
 [and that was the end of the log]

I was online and noticed that something was amiss and investigated.  
I found these messages in the log.  There was no secnet running, but
there was a set of userv ipif processes.

I killed the userv ipif and restarted secnet:

 Dec 23 23:44:33 nl-ipif: about to invoke: userv root ipif [blah blah]
 Dec 23 23:44:33 nl-ipif: userv-ipif pid is 3367
 Dec 23 23:44:34 secnet 0.2.1 [3364]: starting
 [and then everything seems to go well]

Since then we had several recurrences of the "bad IP packet from host"
messages.  But there were fewer in each case.  secnet survived.

I have looked for a core dump in ~secnet/live but there wasn't one.
At some point I will make sure that secnet's startup arrangements
arrange that core dumps are properly enabled.

Ian.



More information about the sgo-software-discuss mailing list