Re: inetd losing sockets

On Mon, 31 Oct 2005, Neil Walker wrote:
>> If you look in /var/log/messages, you'll probably see it complaining that 
>> the service is looping and so will be terminated. This is down to the 
>> number of requests per minute. Up this by putting a number after the 
>> wait/nowait, e.g.:
> That was precisely the problem.
>
> Seems a bit odd, though. Here am I, a theoretical program running a denial of 
> service attack on your server. If I launch enough requests, inetd is going to 
> close the socket, thereby making the service permanently unavailable. Doesn't 
> that mean that I win?

Yes, but you have to weigh this up against DoS down to resources being 
exhausted and insulation from a brute-force password attack (for example), 
both of which could cause more major damage. Perhaps inetd should restart the service 
after a while (exactly as init does with gettys). Or perhaps it does, but 
is undocumented.

(CC: set to netbsd-users).

-- 
Stephen