Bug#406465: [bind backend] TXT record parsing overflow with special characters

2007-02-16 Thread Jeroen van Wolffelaar
On Sat, Feb 10, 2007 at 11:13:11AM +0100, Jeroen van Wolffelaar wrote:
 An option, therefore, is to have a pdns uploaded without the bind
 backend, and a NEWS.Debian stating that sorry, no bind backend
 available, because it's not of release quality or something.
 
 Since other than our brief attempt at using pdns-with-bind-backend, I'm
 not having any experience with pdns, I don't feel comfortable making
 this change (and decision) myself, it's also pretty invasive so not
 typically something to do in a NMU.

Maintainers, what's the status? As it stands now, powerdns runs the risk
of being removed from testing and that way not making it into etch.

If you'd give your opinion on whether or not removing the bind backend
would be an acceptable solution, someone could make an upload of it.

--Jeroen

-- 
Jeroen van Wolffelaar
[EMAIL PROTECTED] (also for Jabber  MSN; ICQ: 33944357)
http://Jeroen.A-Eskwadraat.nl


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#406465: [bind backend] TXT record parsing overflow with special characters

2007-02-16 Thread Christoph Haas
On Friday 16 February 2007 13:57, Jeroen van Wolffelaar wrote:
 On Sat, Feb 10, 2007 at 11:13:11AM +0100, Jeroen van Wolffelaar wrote:
  An option, therefore, is to have a pdns uploaded without the bind
  backend, and a NEWS.Debian stating that sorry, no bind backend
  available, because it's not of release quality or something.
 
  Since other than our brief attempt at using pdns-with-bind-backend,
  I'm not having any experience with pdns, I don't feel comfortable
  making this change (and decision) myself, it's also pretty invasive so
  not typically something to do in a NMU.

 Maintainers, what's the status? As it stands now, powerdns runs the risk
 of being removed from testing and that way not making it into etch.

Apologies. I'll contact the upstream about this bug report now. 
Unfortunately Matthijs is currently very busy at work so he didn't handle 
it yet. And I had no internet connection for a while that kept me from 
working on it. So the package has actually been badly maintained for a 
while. I will try to improve that.

 If you'd give your opinion on whether or not removing the bind backend
 would be an acceptable solution, someone could make an upload of it.

Let us first see what the upstream thinks of it. If I don't get a timely 
answer we can still consider removing the bind backend.

Help is definitely welcome in maintaining the package. But I'll get on my 
backlog today anyway.

 Christoph


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#406465: [bind backend] TXT record parsing overflow with special characters

2007-02-16 Thread Christoph Haas
Update: upstream says it's not a serious security issue in his opinion.
He intends to release a fix this weekend anyway.

 Christoph


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#406465: [bind backend] TXT record parsing overflow with special characters

2007-02-10 Thread Jeroen van Wolffelaar
On Sat, Feb 10, 2007 at 01:09:19AM +0100, Moritz Muehlenhoff wrote:
 Jeroen van Wolffelaar wrote:
  Package: pdns-server
  Version: 2.9.20-7
  Severity: serious
  Tags: security
  
  (serious because what I see looks like a buffer overflow, however, I
  didn't look into the code yet, so I make no claims as to whether this is
  exploitable)
 
 Despite having a team in the maintainer field and being RC this bug log
 shows no visible reaction since a month. If it's unmaintained we shouldn't
 include it in Etch.

One note, for all the bugs that I reported, it seemed limited to the
bind backend, and one not-yet-reported bug about it is that it actually
often 'forgets' zones too (not reported yet because I wanted to try to
reproduce on etch instead of on sarge, I will report it this weekend).

Reportedly, the other backends work just fine, and also, those are the
backend configuration in which pdns sees wide deployment (the usecase
for pdns with bind backend is pretty limited).

An option, therefore, is to have a pdns uploaded without the bind
backend, and a NEWS.Debian stating that sorry, no bind backend
available, because it's not of release quality or something.

Since other than our brief attempt at using pdns-with-bind-backend, I'm
not having any experience with pdns, I don't feel comfortable making
this change (and decision) myself, it's also pretty invasive so not
typically something to do in a NMU.

--Jeroen

-- 
Jeroen van Wolffelaar
[EMAIL PROTECTED] (also for Jabber  MSN; ICQ: 33944357)
http://Jeroen.A-Eskwadraat.nl


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#406465: [bind backend] TXT record parsing overflow with special characters

2007-02-09 Thread Moritz Muehlenhoff
Jeroen van Wolffelaar wrote:
 Package: pdns-server
 Version: 2.9.20-7
 Severity: serious
 Tags: security
 
 (serious because what I see looks like a buffer overflow, however, I
 didn't look into the code yet, so I make no claims as to whether this is
 exploitable)

Despite having a team in the maintainer field and being RC this bug log
shows no visible reaction since a month. If it's unmaintained we shouldn't
include it in Etch.

Cheers,
Moritz


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#406465: [bind backend] TXT record parsing overflow with special characters

2007-01-11 Thread Jeroen van Wolffelaar
Package: pdns-server
Version: 2.9.20-7
Severity: serious
Tags: security

(serious because what I see looks like a buffer overflow, however, I
didn't look into the code yet, so I make no claims as to whether this is
exploitable)

Having a TXT record in a bind-backend zone file that contains a
parentizes ( character, causes all kinds of weird things.

Firstly, the zone fails to serve. Syslog says:
Jan 11 11:40:47 foo pdns[29515]: Zone 'a-eskwadraat.nl' 
(/etc/powerdns/zonefiles/db.nl.a-eskwadraat) reloaded

but all queries including zone transfers result in servfail:
Jan 11 11:40:47 foo pdns[29515]: Not authoritative for 'foo.a-eskwadraat.nl', 
sending servfail to 127.0.0.1 (recursion was desired)

After replacing

foo TXT (

with

foo TXT paren-open

and reloading, I get the following:

| foo:/etc/powerdns# dig  foo.a-eskwadraat.nl TXT @localhost
| 
| ;  DiG 9.3.3  foo.a-eskwadraat.nl TXT @localhost
| ; (1 server found)
| ;; global options:  printcmd
| ;; Got answer:
| ;; -HEADER- opcode: QUERY, status: NOERROR, id: 8804
| ;; flags: qr aa rd; QUERY: 1, ANSWER: 2, AUTHORITY: 0, ADDITIONAL: 0
| 
| ;; QUESTION SECTION:
| ;foo.a-eskwadraat.nl.   IN  TXT
| 
| ;; ANSWER SECTION:
| foo.a-eskwadraat.nl.3600IN  TXT paren-open
| foo.a-eskwadraat.nl.3600IN  TXT foo a 1.2.3.4\010@ ns
| ns1.xel.nl. ns ns3.xel.nl.\010$ttl 1d@ in soa ns.a-eskwadraat.nl.
| sysop.a-eskwadraat.nl. ( 2006110910 6h 30m 4w 1d

This is interesting, because the data listed here comes from the *old*
zonefile (afaics). Also, of course the TXT record shouldn't suddenly
contain literal zonefile data like this.

Powerdns should really treat such TXT record strings as opaque strings,
and not treat characters in them specially.

--Jeroen

-- System Information:
Debian Release: 4.0
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: i386 (i686)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.18-3-686
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)

-- 
Jeroen van Wolffelaar
[EMAIL PROTECTED] (also for Jabber  MSN; ICQ: 33944357)
http://Jeroen.A-Eskwadraat.nl


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]