Hmm, what spl*() is required to prevent soft clock (callout_*()) from
interrupting running code?  But that is only a problem when expiration
is racing using the token next...the "+ 2" in the line:

ipt->ipt_die = fr_ticks + 2;

might need to be larger, but increasing the 2 only masks another real
problem elsewhere.

Darren

> Geoff Wing <mason@primenet.com.au> typed around 3rd May:
> : Somewhere something in the last three days is causing
> : 	Data modified on freelist: word 3 of object 0xc2c2ad80 size 36 \
> : 	previous type bar (0xce2bf63c != 0xdeadbeef)
> : Kernel from May  2 07:09 (UTC) shows it.  Apr 30 01:35 did not.
> : Any likely suspects?
> 
> And the winner is "ipfstat -i".
> 
> "ipfstat -o" doesn't cause it but I don't have many rules in there (and
> no group rules).
> 
> Regards,
> Geoff