[Dovecot] Bug in driver-mysql.c + fix

[Alain Williams]

On Wed, Mar 10, 2010 at 04:18:19PM +0200, Timo Sirainen wrote:

> On Fri, 2010-03-05 at 23:44 +0000, Alain Williams wrote:
...
> > This failed with the message:
> > 	User query failed: PROCEDURE imap.user_info can't return a result set in the given context
> 
> I thought it was possible to avoid that error by implementing the MySQL
> procedure in a specific way?..

I have tried ... see another thread.

See:
	http://dev.mysql.com/doc/refman/5.1/en/mysql-real-connect.html

Just under the table near the top it says:

	If your program uses CALL statements to execute stored procedures, the CLIENT_MULTI_RESULTS flag must be enabled.

> > The root of this problem is that mysql_real_connect() needs to be called with option CLIENT_MULTI_RESULTS
> 
> The problem with doing that is that 1) it's not normally necessary and
> more importantly 2) doing that makes any potential SQL injection
> security holes a lot easier to exploit. So I'm not all that eager in
> adding such code, especially if it can be worked around another way..

CLIENT_MULTI_STATEMENTS allows multiple statements in one call (you separate by ',').
CLIENT_MULTI_RESULTS does not imply CLIENT_MULTI_STATEMENTS.
Is this what you were concerned about ?

Look at the URL above for details.

> > I am looking to use mysql procedures, there are some interesting things that can be done.
> > Two suggestions that I have will help with this:
> > 
> > 1) There be variable (say) %o - this be the obscured password, ie what password_query returns.
> > 
> > 2) that dovecot look for either ''password_query'' as above, or ''password_check''.
> >    password_check would NOT return a password, but would be given %o and determine
> >    itself if the password is correct.
> >    It would return the other values (user, userdb_home, ...) and auth_result
> >    that would encode success/retry/fail (0, 1, 2 - or maybe more structured [**])
> >    and auth_reason some human readable reason.
> >    The ''nologin'' value encodes some of this.
> > 
> > The motivation for this is that my stored procedure will record the number of successive
> > login failures and lock the account after 3 of them. It would also be possible to
> > do time based restrictions & the such.
> > 
> > Also: by passing %o the password is not sent in plain to the database server - which
> > will increase security.
> 
> So what kind of obscuring where you thinking about? You could already
> use "%Mw", which gives you MD5 of the password.

That is not how I store passwords - I keep them as DIGEST-MD5, this is:
	md5('username:domain:password')
So I want %o to be that value. Squirrelmail should be able to deduce that from
the line in the dovecot-sql.conf:
	default_pass_scheme="DIGEST-MD5"

> And password_query can already do basically what you were thinking about
> with password_check:
> 
> password_query = select NULL as password, 'Y' as nopassword FROM users
> WHERE username = '%u' and password = '%w'
> 
> or something. The main problem with that is that it can't differentiate
> between "wrong password" and "user doesn't exist", so it logs all
> password failures as "user doesn't exist".

I am trying to find a definition of the API to plugins, ...
if the SQL stored procedure can return arbitrary variables that can then be used
by PHP plugins then I can do things like issuing a warning about the password
about to expire, number of failed login attempts since the last success, ...
Ie all sorts of things that the authentication stored procedures could store
and manage.

-- 
Alain Williams
Linux/GNU Consultant - Mail systems, Web sites, Networking, Programmer, IT Lecturer.
+44 (0) 787 668 0256  http://www.phcomp.co.uk/
Parliament Hill Computers Ltd. Registration Information: http://www.phcomp.co.uk/contact.php
Past chairman of UKUUG: http://www.ukuug.org/
#include <std_disclaimer.h>