Re: [stable] openssl rfc5746 / renegotiation support

2010-11-20 Thread Peter Palfrader
On Fri, 19 Nov 2010, Philipp Kern wrote:

  When we throw out tor 0.2.0.x from lenny, are we replacing it with the
  version currently in lenny-volatile?
 
 I'd be ok with that.

Ok, I'll upload it later today.  I'll include the no-coredump patch too,
unless somebody shouts.

Patches attached.
-- 
   |  .''`.  ** Debian GNU/Linux **
  Peter Palfrader  | : :' :  The  universal
 http://www.palfrader.org/ | `. `'  Operating System
   |   `-http://www.debian.org/
From 0856f24277cd984231418420fc7a9dbf7e32bf86 Mon Sep 17 00:00:00 2001
From: Peter Palfrader pe...@palfrader.org
Date: Sat, 20 Nov 2010 11:30:04 +0100
Subject: [PATCH 1/2] Changelog for upload to lenny as 0.2.1.26-1~lenny+1

---
 debian/changelog |   19 +++
 1 files changed, 19 insertions(+), 0 deletions(-)

diff --git a/debian/changelog b/debian/changelog
index 62639b7..047276b 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -1,3 +1,22 @@
+tor (0.2.1.26-1~lenny+1) stable; urgency=low
+
+  * Upload to stable, since a potential future security upload
+for openssl will break Tor 0.2.0.35, the version currently in
+stable.
+.
+This libssl update will change how openssl does renegotiation,
+and Tor 0.2.0.x relies on the old behaviour but does not yet
+know how to tell openssl to stick to it.  The 0.2.1.x tree
+has special support for newer openssls and sets the proper, new
+flags so libssl continues to provide the old renegotiation
+features.
+.
+The old-style renegotiation feature is potentially dangerous
+in some cases, but it is believed that Tor does everything
+correctly and thus can use it safely.
+
+ -- Peter Palfrader wea...@debian.org  Sat, 20 Nov 2010 11:24:32 +0100
+
 tor (0.2.1.26-1~lennyvolatile1) lenny-volatile; urgency=low
 
   * Upload to lenny-volatile.
-- 
1.7.1

From 3b9ecd3001478b15a0c418f4e825d4fea98de426 Mon Sep 17 00:00:00 2001
From: Peter Palfrader pe...@palfrader.org
Date: Sat, 20 Nov 2010 11:32:36 +0100
Subject: [PATCH 2/2] disable coredumps by default

Up until now the init script (or actually /etc/default/tor) raised
the ulimit for coredumps to unlimited, so that Tor would produce
coredumps on assert errors or segfaults.  Coredumps however can
leak sensitive information, like cryptographic session keys and
clients' data should the core files get into the wrong hands.  As
such it seems prudent to only enable coredumps if the user or
operator explicitly asks for them, and knows what to do with them.

Also include a cron.weekly job that removes old coredumps from /var/lib/tor.
This action can be disabled in etc/default/tor.

Cherry-picked from debian-0.2.1.  Original commit is
270ed52406acd28eea57114acf04ba7d63cedad5.

Conflicts:

	debian/changelog
---
 debian/changelog   |   10 ++
 debian/rules   |1 +
 debian/tor.cron.weekly |   14 ++
 debian/tor.default |   20 ++--
 4 files changed, 43 insertions(+), 2 deletions(-)
 create mode 100755 debian/tor.cron.weekly

diff --git a/debian/changelog b/debian/changelog
index 047276b..1ce1cd6 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -14,6 +14,16 @@ tor (0.2.1.26-1~lenny+1) stable; urgency=low
 The old-style renegotiation feature is potentially dangerous
 in some cases, but it is believed that Tor does everything
 correctly and thus can use it safely.
+  * No longer set ulimit -c to unlimited:
+Up until now the init script (or actually /etc/default/tor) raised
+the ulimit for coredumps to unlimited, so that Tor would produce
+coredumps on assert errors or segfaults.  Coredumps however can
+leak sensitive information, like cryptographic session keys and
+clients' data should the core files get into the wrong hands.  As
+such it seems prudent to only enable coredumps if the user or
+operator explicitly asks for them, and knows what to do with them.
+  * Also include a cron.weekly job that removes old coredumps from
+/var/lib/tor.  This action can be disabled in /etc/default/tor.
 
  -- Peter Palfrader wea...@debian.org  Sat, 20 Nov 2010 11:24:32 +0100
 
diff --git a/debian/rules b/debian/rules
index f5aaa95..5c6a302 100755
--- a/debian/rules
+++ b/debian/rules
@@ -187,6 +187,7 @@ binary-common:
 	dh_installexamples
 	dh_installlogrotate
 	dh_installinit
+	dh_installcron
 	dh_installman
 	dh_link
 	dh_strip --dbg-package=tor-dbg
diff --git a/debian/tor.cron.weekly b/debian/tor.cron.weekly
new file mode 100755
index 000..ae4469d
--- /dev/null
+++ b/debian/tor.cron.weekly
@@ -0,0 +1,14 @@
+#!/bin/sh
+
+set -e
+set -u
+
+DEFAULTSFILE=/etc/default/tor
+
+if [ -f $DEFAULTSFILE ] ; then
+	. $DEFAULTSFILE
+fi
+
+if [ ${CLEANUP_OLD_COREFILES:-} = y ] ; then
+	find /var/lib/tor -mindepth 1 -maxdepth 1 -type f -mtime +21 -user debian-tor -regex '.*/core\(\.[0-9]+\)?' -exec rm '{}' +
+fi
diff --git a/debian/tor.default b/debian/tor.default
index 

Re: [stable] openssl rfc5746 / renegotiation support

2010-11-20 Thread Peter Palfrader
On Sat, 20 Nov 2010, Peter Palfrader wrote:

   When we throw out tor 0.2.0.x from lenny, are we replacing it with the
   version currently in lenny-volatile?
  
  I'd be ok with that.
 
 Ok, I'll upload it later today.  I'll include the no-coredump patch too,
 unless somebody shouts.

Uploaded just the volatile version, with no further changes as per your
request.

interdiff and debdiff attached.

Thanks,
weasel
-- 
   |  .''`.  ** Debian GNU/Linux **
  Peter Palfrader  | : :' :  The  universal
 http://www.palfrader.org/ | `. `'  Operating System
   |   `-http://www.debian.org/
diff -u tor-0.2.1.26/debian/changelog tor-0.2.1.26/debian/changelog
--- tor-0.2.1.26/debian/changelog
+++ tor-0.2.1.26/debian/changelog
@@ -1,3 +1,22 @@
+tor (0.2.1.26-1~lenny+1) stable; urgency=low
+
+  * Upload to stable, since a potential future security upload
+for openssl will break Tor 0.2.0.35, the version currently in
+stable.
+.
+This libssl update will change how openssl does renegotiation,
+and Tor 0.2.0.x relies on the old behaviour but does not yet
+know how to tell openssl to stick to it.  The 0.2.1.x tree
+has special support for newer openssls and sets the proper, new
+flags so libssl continues to provide the old renegotiation
+features.
+.
+The old-style renegotiation feature is potentially dangerous
+in some cases, but it is believed that Tor does everything
+correctly and thus can use it safely.
+
+ -- Peter Palfrader wea...@debian.org  Sat, 20 Nov 2010 11:24:32 +0100
+
 tor (0.2.1.26-1~lennyvolatile1) lenny-volatile; urgency=low
 
   * Upload to lenny-volatile.
diff -u tor-0.2.1.26/debian/micro-revision.i 
tor-0.2.1.26/debian/micro-revision.i
--- tor-0.2.1.26/debian/micro-revision.i
+++ tor-0.2.1.26/debian/micro-revision.i
@@ -1 +1 @@
-59d3dbb874f0deb3
+0856f24277cd9842
File lists identical (after any substitutions)

Control files of package tor: lines which differ (wdiff format)
---
Version: [-0.2.1.26-1~lennyvolatile1-] {+0.2.1.26-1~lenny+1+}

Control files of package tor-dbg: lines which differ (wdiff format)
---
Depends: tor (= [-0.2.1.26-1~lennyvolatile1)-] {+0.2.1.26-1~lenny+1)+}
Version: [-0.2.1.26-1~lennyvolatile1-] {+0.2.1.26-1~lenny+1+}

Control files of package tor-geoipdb: lines which differ (wdiff format)
---
Depends: tor (= [-0.2.1.26-1~lennyvolatile1)-] {+0.2.1.26-1~lenny+1)+}
Version: [-0.2.1.26-1~lennyvolatile1-] {+0.2.1.26-1~lenny+1+}


Re: [stable] openssl rfc5746 / renegotiation support

2010-11-19 Thread Peter Palfrader
Hey release folks,

On Tue, 16 Nov 2010, Kurt Roeckx wrote:

 I would like to add rfc5746 support to openssl in stable, so that
 CVE-2009-3555 can be fixed.  But adding that support means that
 the old renegotiation doesn't work anymore unless you set an
 option.


 There are atleast 2 packages that have an issue with this that I'm
 currently aware of:

 - tor: It should always disable the new renegotiation.  Running it
   as a server doesn't work.  Newer versions than in stable, like
   the version in volatile, do work properly with any version of
   openssl.  The maintainer and upstream favour dropping the
   version currently in stable.

What's the verdict here?

Are we going to update openssl in stable (either via -security or with
the next point release), and if yes, what are we doing with tor?

When we throw out tor 0.2.0.x from lenny, are we replacing it with the
version currently in lenny-volatile?

Cheers,
-- 
   |  .''`.  ** Debian GNU/Linux **
  Peter Palfrader  | : :' :  The  universal
 http://www.palfrader.org/ | `. `'  Operating System
   |   `-http://www.debian.org/


-- 
To UNSUBSCRIBE, email to debian-release-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20101119103548.gs10...@anguilla.noreply.org



Re: [stable] openssl rfc5746 / renegotiation support

2010-11-19 Thread Philipp Kern
On Fri, Nov 19, 2010 at 11:35:48AM +0100, Peter Palfrader wrote:
 On Tue, 16 Nov 2010, Kurt Roeckx wrote:
  I would like to add rfc5746 support to openssl in stable, so that
  CVE-2009-3555 can be fixed.  But adding that support means that
  the old renegotiation doesn't work anymore unless you set an
  option.
  There are atleast 2 packages that have an issue with this that I'm
  currently aware of:
 
  - tor: It should always disable the new renegotiation.  Running it
as a server doesn't work.  Newer versions than in stable, like
the version in volatile, do work properly with any version of
openssl.  The maintainer and upstream favour dropping the
version currently in stable.
 What's the verdict here?
 
 Are we going to update openssl in stable (either via -security or with
 the next point release), and if yes, what are we doing with tor?

That's up to security. 

 When we throw out tor 0.2.0.x from lenny, are we replacing it with the
 version currently in lenny-volatile?

I'd be ok with that.

Kind regards
Philipp kern 


signature.asc
Description: Digital signature


Re: [stable] openssl rfc5746 / renegotiation support

2010-11-19 Thread Michael Gilbert
On Fri, 19 Nov 2010 11:35:48 +0100, Peter Palfrader wrote:
 What's the verdict here?
 
 Are we going to update openssl in stable (either via -security or with
 the next point release), and if yes, what are we doing with tor?

The decision was made a while back to fix this in an SPU.  Kurt put out
packages a few weeks ago and called for testing, but I've only seen one
response so far.  So I think that what needs to happen is a little more
testing?

Mike


-- 
To UNSUBSCRIBE, email to debian-release-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/20101119102808.9a4b8a48.michael.s.gilb...@gmail.com



Re: [stable] openssl rfc5746 / renegotiation support

2010-11-19 Thread Philipp Kern
On Fri, Nov 19, 2010 at 10:28:08AM -0500, Michael Gilbert wrote:
 On Fri, 19 Nov 2010 11:35:48 +0100, Peter Palfrader wrote:
  What's the verdict here?
  
  Are we going to update openssl in stable (either via -security or with
  the next point release), and if yes, what are we doing with tor?
 
 The decision was made a while back to fix this in an SPU.  Kurt put out
 packages a few weeks ago and called for testing, but I've only seen one
 response so far.  So I think that what needs to happen is a little more
 testing?

You might want to point us to that decision.

We'd prefer to have this handled by security.

Thanks
Philipp Kern 


signature.asc
Description: Digital signature


openssl rfc5746 / renegotiation support

2010-11-16 Thread Kurt Roeckx
Hi,

I would like to add rfc5746 support to openssl in stable, so that
CVE-2009-3555 can be fixed.  But adding that support means that
the old renegotiation doesn't work anymore unless you set an
option.  This has the potentional to break both client and server
applications making use of openssl.  See the SSL_CTX_set_options
manpage for the behaviour and the options you can set.

There are atleast 2 packages that have an issue with this that I'm
currently aware of:
- apache2: It would need an option an admin can turn on to allow
  insecure renegotiation.
- tor: It should always disable the new renegotiation.  Running it
  as a server doesn't work.  Newer versions than in stable, like
  the version in volatile, do work properly with any version of
  openssl.  The maintainer and upstream favour dropping the
  version currently in stable.

Other packages still need to be checked.

I think at this point we're not going to be able to check the
various packages that might be affected by this before the next
point release.  So I wonder when the next point release is
going to happen, or that we should try and update this via the
security archive.


Kurt


-- 
To UNSUBSCRIBE, email to debian-release-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20101116204806.ga14...@roeckx.be