[MBT] new entry in pkg oidentd "Buffer overflow..."

Marcin Krol krol at sigma.wsb-nlu.edu.pl
Mon Oct 21 16:51:20 CEST 2002


On 20 Oct 2002 bugs at pld.org.pl wrote:

> indent like all other indent damons do not fork().
> multiply indent instances can be foud in case when you are uses enteries in hosts.{allow,deny) like:
>
> <service>: All at ALL
>
> and on other sides also are used the same entries.
> Useing for ecample in hosts.allow:
>
> telnetd: ALL@<host>
>
> will allow additional indent lookup from tcpd. If on host someonn will have
>
> indent: ALL at ALL
>
> it will cause next indent from <host> .. classic indent ping-pong.
> How to avoid this ? Simple for do not use indent specyfic ruler for indent rntrues in hosts.{allow,deny}.
> Summarize: this not bug .. this feacture :)
>
>
>
> *** State changed to 'ignored'

OK. I know this. But... If there are no rules for ident in
hosts.{allow,deny}, and single ident request from some host/service
causes, lets say 7000 oidentd processes, this is not a classic ident
ping-pong. More than that, even when request is served, oidentd is still
multiplying with response for the same request infinitely, until machine
hangs due to "out of memory" problem. This is happening only on oidentd
2.x. I've never noticed that kind of behaviour on any other identd
software, including oidentd 1.x. And after upgrading to 2.x, this happens
at least one  time per week. Its irritating because if I'm away from
server when oidentd starts eating up memory, machine will hang and will
not respond until I'll reboot it. BTW: oidentd ping-pong will not eat
memory with speed >= 50 megabytes per second.

Marcin Król



More information about the pld-devel-en mailing list