[Openswan Users] Pluto Page Allocation Failure

[Paul Wouters]

On Tue, 24 Aug 2004, Jason Sigurdur wrote:

> I am using FC2 with openswan version 2.1.2. With 16 tunnels. Every few days
> Pluto consumes 50+ percent of cpu ussage and I get the following error:
>
> kernel: pluto: page allocation failure. order:4, mode:0xd0
>
> I have to restart ipsec to get things working again.
>
> Any suggestions explanations would be appreciated

Edit programs/pluto/Makefile

uncomment -DLEAK_DETECTIVE

when it gives you that failure, please do:
tail -f /var/log/secure > pluto-mem.log &
service ipsec restart
(then kill tail)

leak detective will only log the memory leaks it finds when restarting. So it
won't log it during normal operations. It is important to let it shut down
nicely to obtain the logs.

With those logs, we can see about locating this particular memory leak.
Though I find it strange. It might actually be the "ip compression" kernel bug
where the native 2.6 stack allocates a lot of memory for ip compression and
you end up with fragmented memory problems. Perhaps a newer 2.6 kernel would
fix this problem.

Paul