Re: [HACKERS] GUC --- prevent non-super user changes - Mailing list pgsql-patches

From Bruce Momjian
Subject Re: [HACKERS] GUC --- prevent non-super user changes
Date
Msg-id 200306112229.h5BMTXP21826@candle.pha.pa.us
Whole thread Raw
List pgsql-patches
Robert Treat wrote:
> On Wed, 2003-06-11 at 01:01, Bruce Momjian wrote:
> > Here is a proposed fix for the new SUSET of various variables.  The
> > solution is to create a new GUC context called PGC_USERLIMIT, which
> > limits changes by non-super users.  For example, non-super users can
> > turn on logging, but can't turn it off, and log_min_* logging can have
> > added output, but not less output.
> >
>
> Is there a danger here that users can crank logging up to the max and
> either crash a server due to i/o load, or perhaps create enough "noise"
> to cover tracks of something malicious?

There perhaps is, but if they can connect to the database, I would think
there would be other worse things they can do.

The base problem is that we don't know if the person is a super user
until we connect to pg_shadow, and this is much later than when we
process the flags and PGOPTIONS packet, so the patch seemed like the
cleanest way to go.  We could have delayed the setting of those
variables, but there are some variables that have to be set _before_ we
connect to pg_shadow, so it would get tricky.

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

pgsql-patches by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: [HACKERS] "Adding missing from clause" (replacement)
Next
From: Bruce Momjian
Date:
Subject: Re: Simple LRU for clog and (later) subtrans