Re: [QUESTIONS] Re Trivial mSQL/MySQL DoS method

At 7:06 AM -0800 3/27/98, Cary B. O'Brien wrote:
>makrhr(at)SHMOOZE(dot)NET wrote
>> It seems that if one wants to bring a website that relies heavily on mSQL or
>> MySQL to it's knees, simply telnet to the port the server listens on (1112
>> for mSQL or 3333 for MySQL) and then just sit there, forget about it.

>Observations:
>
>2) Sounds like postgresql would be a bit better off, since postmaster
> forks backend processes, but I haven't checked it. Actually
> wouldn't tight host-based authentication prevent this?

Don't know about hba, but I concur that Postgres is structurally immune to
this attack. It's an artifact of mSQL being a single-threaded process:
great for fast response under light load, terrible under heavy load. It
does mean they don't have to worry about concurrent transactions though:
there is no concurrency.
>
>3) How hard would it be to create a postmaster/postgresql process
> that could be started from inetd under tcp_wrappers. That would
> provide authentication/logging/monitoring of what happens

I like the idea of allowing postgres to run from inetd since it makes for
less overhead if it's little-used. However I think there is some
coordination among the children that is managed by the postmaster which
would suffer, and would require duplicate reimplementation to work. Not a
good use of our programmers time IMHO.

On the other hand there is a library version of tcp_wrappers (libwrap under
NetBSD) which could, and probably should, be linked into the postmaster to
provide the same functionality.

I'm cc'ing the hackers list on this note in the hope that someone will take
an interest.

>4) Has anyone tried the 'send garbage and see what happens' test?
>
Not I, but it's a good idea.

Signature failed Preliminary Design Review.
Feasibility of a new signature is currently being evaluated.
h(dot)b(dot)hotz(at)jpl(dot)nasa(dot)gov, or hbhotz(at)oxy(dot)edu