[Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Philipp Hagemeister]

[-- Attachment #1: Type: text/plain, Size: 1002 bytes --]

On modern systems, topology subnet should always be set, but it's
missing in the configuration file.
Add it with a short explanation.

Signed-off-by: Philipp Hagemeister <phihag@...1921...>
---
 sample/sample-config-files/server.conf | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/sample/sample-config-files/server.conf
b/sample/sample-config-files/server.conf
index f483b6b..6a2b4c2 100644
--- a/sample/sample-config-files/server.conf
+++ b/sample/sample-config-files/server.conf
@@ -86,6 +86,13 @@ key server.key  # This file should be kept secret
 # 2048 bit keys.
 dh dh1024.pem

+# Network topology
+# Should be subnet (addressing via IP)
+# unless Windows clients v2.0.9 and lower have to
+# be supported (then net30, i.e. a /30 per client)
+# Defaults to net30 (not recommended)
+;topology subnet
+
 # Configure server mode and supply a VPN subnet
 # for OpenVPN to draw client addresses from.
 # The server will take 10.8.0.1 for itself,
-- 
2.0.1



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

[-- Attachment #1: Type: text/plain, Size: 1880 bytes --]

On 11/07/14 10:00, Philipp Hagemeister wrote:
> On modern systems, topology subnet should always be set, but it's
> missing in the configuration file.
> Add it with a short explanation.
NACK
There are a few annoying issues with
   topology subnet
esp when using server side things like
   route <remote subnet> <mask>

that will *fail* with 'topology subnet' ; until this is fixed this 
should NOT be considered the default.

JJK


> Signed-off-by: Philipp Hagemeister <phihag@...1921...>
> ---
>   sample/sample-config-files/server.conf | 7 +++++++
>   1 file changed, 7 insertions(+)
>
> diff --git a/sample/sample-config-files/server.conf
> b/sample/sample-config-files/server.conf
> index f483b6b..6a2b4c2 100644
> --- a/sample/sample-config-files/server.conf
> +++ b/sample/sample-config-files/server.conf
> @@ -86,6 +86,13 @@ key server.key  # This file should be kept secret
>   # 2048 bit keys.
>   dh dh1024.pem
>
> +# Network topology
> +# Should be subnet (addressing via IP)
> +# unless Windows clients v2.0.9 and lower have to
> +# be supported (then net30, i.e. a /30 per client)
> +# Defaults to net30 (not recommended)
> +;topology subnet
> +
>   # Configure server mode and supply a VPN subnet
>   # for OpenVPN to draw client addresses from.
>   # The server will take 10.8.0.1 for itself,
>
>
> ------------------------------------------------------------------------------
> Open source business process management suite built on Java and Eclipse
> Turn processes into business applications with Bonita BPM Community Edition
> Quickly connect people, data, and systems into organized workflows
> Winner of BOSSIE, CODIE, OW2 and Gartner awards
> http://p.sf.net/sfu/Bonitasoft
>
>
> _______________________________________________
> Openvpn-devel mailing list
> Openvpn-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/openvpn-devel


[-- Attachment #2: Type: text/html, Size: 2913 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Arne Schwabe]

[-- Attachment #1: Type: text/plain, Size: 701 bytes --]

Am 11.07.14 10:51, schrieb Jan Just Keijser:
> On 11/07/14 10:00, Philipp Hagemeister wrote:
>> On modern systems, topology subnet should always be set, but it's
>> missing in the configuration file.
>> Add it with a short explanation.
> NACK
> There are a few annoying issues with
>   topology subnet
> esp when using server side things like
>   route <remote subnet> <mask>
>
> that will *fail* with 'topology subnet' ; until this is fixed this
> should NOT be considered the default.

I am confused here, route <remote subnet> <mask> would expand to

route <remote subnet> <mask> remote_gw in p2p, net30 and subnet toplogy.
And that should work in all topology modes. What am I missing here?

Arne

[-- Attachment #2: Type: text/html, Size: 1433 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

Hi Arne,

Arne Schwabe wrote:
> Am 11.07.14 10:51, schrieb Jan Just Keijser:
>> On 11/07/14 10:00, Philipp Hagemeister wrote:
>>> On modern systems, topology subnet should always be set, but it's
>>> missing in the configuration file.
>>> Add it with a short explanation.
>>>       
>> NACK
>> There are a few annoying issues with
>>   topology subnet
>> esp when using server side things like
>>   route <remote subnet> <mask>
>>
>> that will *fail* with 'topology subnet' ; until this is fixed this 
>> should NOT be considered the default.
>
> I am confused here, route <remote subnet> <mask> would expand to
>
> route <remote subnet> <mask> remote_gw in p2p, net30 and subnet 
> toplogy. And that should work in all topology modes. What am I missing 
> here?
>

this is a long standing bug, check out the recent thread on the 
openvpn-users mailing list
  Re: [Openvpn-users] ping tests mostly work.  all OK,    except 
Server->Client private IP.  *almost* there!

On Mon, Jul 07, 2014 at 11:41:31AM -0700, pg0706@...1922... wrote:

> > Switching back to `verb 3`, @Server log on server START
> > 
> > (not SUPPOSED to be wrapping, at least it doesn't in the web editor --
> > just in case: http://pastebin.com/KmChijWX )
>   

"Wrapped like hell".  Is there no way to do attachments in the web thingie?

Anyway:


> > there's a suspect problem
> > 
> > 	Mon Jul  7 10:52:42 2014 OpenVPN ROUTE: OpenVPN needs a gateway
> > 	parameter for a --route option and no default was specified by
> > 	either --route-gateway or --ifconfig options
> > 	Mon Jul  7 10:52:42 2014 OpenVPN ROUTE: failed to parse/resolve
> > 	route for host/network: 192.168.1.0
> > 
> > which looks like a possible culprit.
>   

... indeed.  This is, uh, stupid code, and I think we already have a trac
ticket for it...


> > If yes, what's missing in OpenVPN config, and where?
>   

--route 192.168.1.0 255.255.255.0 10.0.0.8

("just any IP address that's inside your tun subnet, and not the first").

This is really not *necessary*, but the way the code does tun & tap and
net30 & subnet, it gets confused about things.  Needs fixing, sorry for
that.



so in some cases a server-side statement
  route 192.168.1.0 255.255.255.0
works fine with 'topology net30' but NOT with 'topology subnet'.

This is a long standing bug that needs to be fixed before we can forcefeed everybody 'topology subnet'.

JJK

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Arne Schwabe]

>
> This is really not *necessary*, but the way the code does tun & tap and
> net30 & subnet, it gets confused about things.  Needs fixing, sorry for
> that.
>
>
>
> so in some cases a server-side statement
>  route 192.168.1.0 255.255.255.0
> works fine with 'topology net30' but NOT with 'topology subnet'.
>
> This is a long standing bug that needs to be fixed before we can
> forcefeed everybody 'topology subnet'.

Okay. I see where the confusion is coming from. if you have net30, the
second parameter for the ifconfig is always the gateway. In subnet you
explicity need to set the gateway by setting/pushing route-gateway.

In normal setups the --server macro automatically does that for you.

I am not really seeing a bug here but a misconfiguration of OpenVPN.
Sure the error messages could be improved and I am all ear to this.

Arne

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

Hi Arne,

Arne Schwabe wrote:
>> This is really not *necessary*, but the way the code does tun & tap and
>> net30 & subnet, it gets confused about things.  Needs fixing, sorry for
>> that.
>>
>>
>>
>> so in some cases a server-side statement
>>  route 192.168.1.0 255.255.255.0
>> works fine with 'topology net30' but NOT with 'topology subnet'.
>>
>> This is a long standing bug that needs to be fixed before we can
>> forcefeed everybody 'topology subnet'.
>>     
>
> Okay. I see where the confusion is coming from. if you have net30, the
> second parameter for the ifconfig is always the gateway. In subnet you
> explicity need to set the gateway by setting/pushing route-gateway.
>
> In normal setups the --server macro automatically does that for you.
>   
I wish this were true. Consider the following server config:

proto udp
port 1194
dev tun
server 192.168.200.0 255.255.255.0
ca       /etc/openvpn/cookbook/ca.crt
cert     /etc/openvpn/cookbook/server.crt
key      /etc/openvpn/cookbook/server.key
dh       /etc/openvpn/cookbook/dh1024.pem
tls-auth /etc/openvpn/cookbook/ta.key 0
persist-key
persist-tun
keepalive 10 60
push "route 10.198.0.0 255.255.0.0"
user  nobody
group nobody
client-config-dir /etc/openvpn/cookbook/clients
route 192.168.4.0 255.255.255.0

(example from my cookbook)
this example works fine 'as is' but if you add
  topology subnet
you get

Fri Jul 11 11:31:28 2014 OpenVPN ROUTE: OpenVPN needs a gateway 
parameter for a --route option and no default was specified by either 
--route-gateway or --ifconfig options
Fri Jul 11 11:31:28 2014 OpenVPN ROUTE: failed to parse/resolve route 
for host/network: 192.168.4.0
Fri Jul 11 11:31:28 2014 TUN/TAP device tun0 opened


so the 'server' macro does not seem to properly publish this in all cases.

JJK

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Arne Schwabe]

>
> Fri Jul 11 11:31:28 2014 OpenVPN ROUTE: OpenVPN needs a gateway
> parameter for a --route option and no default was specified by either
> --route-gateway or --ifconfig options
> Fri Jul 11 11:31:28 2014 OpenVPN ROUTE: failed to parse/resolve route
> for host/network: 192.168.4.0
> Fri Jul 11 11:31:28 2014 TUN/TAP device tun0 opened
>
>
> so the 'server' macro does not seem to properly publish this in all
> cases.
>
I sent a patch to the list which should fix that use case. Can you test
that patch and if it works for you too?

Arne

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Philipp Hagemeister]

[-- Attachment #1: Type: text/plain, Size: 2317 bytes --]

After applying Arne's patch, would you ACK this?

In any case, the default is untouched by this patch. Can you enlist the
"few" critical bugs? I'll document them then on
https://community.openvpn.net/openvpn/wiki/Topology .

Best,

Philipp

On 07/11/2014 10:51 AM, Jan Just Keijser wrote:
> On 11/07/14 10:00, Philipp Hagemeister wrote:
>> On modern systems, topology subnet should always be set, but it's
>> missing in the configuration file.
>> Add it with a short explanation.
> NACK
> There are a few annoying issues with
>   topology subnet
> esp when using server side things like
>   route <remote subnet> <mask>
> 
> that will *fail* with 'topology subnet' ; until this is fixed this
> should NOT be considered the default.
> 
> JJK
> 
> 
>> Signed-off-by: Philipp Hagemeister <phihag@...1921...>
>> ---
>>   sample/sample-config-files/server.conf | 7 +++++++
>>   1 file changed, 7 insertions(+)
>>
>> diff --git a/sample/sample-config-files/server.conf
>> b/sample/sample-config-files/server.conf
>> index f483b6b..6a2b4c2 100644
>> --- a/sample/sample-config-files/server.conf
>> +++ b/sample/sample-config-files/server.conf
>> @@ -86,6 +86,13 @@ key server.key  # This file should be kept secret
>>   # 2048 bit keys.
>>   dh dh1024.pem
>>
>> +# Network topology
>> +# Should be subnet (addressing via IP)
>> +# unless Windows clients v2.0.9 and lower have to
>> +# be supported (then net30, i.e. a /30 per client)
>> +# Defaults to net30 (not recommended)
>> +;topology subnet
>> +
>>   # Configure server mode and supply a VPN subnet
>>   # for OpenVPN to draw client addresses from.
>>   # The server will take 10.8.0.1 for itself,
>>
>>
>> ------------------------------------------------------------------------------
>>
>> Open source business process management suite built on Java and Eclipse
>> Turn processes into business applications with Bonita BPM Community
>> Edition
>> Quickly connect people, data, and systems into organized workflows
>> Winner of BOSSIE, CODIE, OW2 and Gartner awards
>> http://p.sf.net/sfu/Bonitasoft
>>
>>
>> _______________________________________________
>> Openvpn-devel mailing list
>> Openvpn-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/openvpn-devel
> 
> 



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Gert Doering]

[-- Attachment #1: Type: text/plain, Size: 1063 bytes --]

Hi,

On Fri, Jul 11, 2014 at 10:51:54AM +0200, Jan Just Keijser wrote:
> On 11/07/14 10:00, Philipp Hagemeister wrote:
> >On modern systems, topology subnet should always be set, but it's
> >missing in the configuration file.
> >Add it with a short explanation.
> NACK
> There are a few annoying issues with
>   topology subnet
> esp when using server side things like
>   route <remote subnet> <mask>
> 
> that will *fail* with 'topology subnet' ; until this is fixed this 
> should NOT be considered the default.

Well, it's on the roadmap, and Arne has already sent a patch.

So, what other issues exist?  This is the only one I'm aware of... 
(and the recent trac ticket about "topology subnet not working on
FreeBSD 10" turned out to be local confusion...)

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@...1296...
fax: +49-89-35655025                        gert@...1297...

[-- Attachment #2: Type: application/pgp-signature, Size: 291 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

Arne Schwabe wrote:
>> Fri Jul 11 11:31:28 2014 OpenVPN ROUTE: OpenVPN needs a gateway
>> parameter for a --route option and no default was specified by either
>> --route-gateway or --ifconfig options
>> Fri Jul 11 11:31:28 2014 OpenVPN ROUTE: failed to parse/resolve route
>> for host/network: 192.168.4.0
>> Fri Jul 11 11:31:28 2014 TUN/TAP device tun0 opened
>>
>>
>> so the 'server' macro does not seem to properly publish this in all
>> cases.
>>
>>     
> I sent a patch to the list which should fix that use case. Can you test
> that patch and if it works for you too?
>
>   

the patch works on the 2.3 branch.
the master branch (from openvpn-testing) currently does not build on 
either CentOS 5 and 6.

JJK

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

Hi,

Gert Doering wrote:
> On Fri, Jul 11, 2014 at 10:51:54AM +0200, Jan Just Keijser wrote:
>   
>> On 11/07/14 10:00, Philipp Hagemeister wrote:
>>     
>>> On modern systems, topology subnet should always be set, but it's
>>> missing in the configuration file.
>>> Add it with a short explanation.
>>>       
>> NACK
>> There are a few annoying issues with
>>   topology subnet
>> esp when using server side things like
>>   route <remote subnet> <mask>
>>
>> that will *fail* with 'topology subnet' ; until this is fixed this 
>> should NOT be considered the default.
>>     
>
> Well, it's on the roadmap, and Arne has already sent a patch.
>
> So, what other issues exist?  This is the only one I'm aware of... 
> (and the recent trac ticket about "topology subnet not working on
> FreeBSD 10" turned out to be local confusion...)
>
>
>   
hmm I thought I had a TRAC ticket on this but I cannot find it anymore...
I recall some  minor annoyances all having to do with needing to 
explictly add a gateway for the server-side routes. I'm 99.9% confident 
that Arne's one-liner will have fixed them.

With those annoyances gone I'd even be in favor of making 'topology 
subnet' the default for the 2.4 release :)

cheers,

JJK

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Gert Doering]

[-- Attachment #1: Type: text/plain, Size: 563 bytes --]

Hi,

On Fri, Jul 11, 2014 at 04:50:38PM +0200, Jan Just Keijser wrote:
> the master branch (from openvpn-testing) currently does not build on 
> either CentOS 5 and 6.

Install snappy(-dev) or run configure with --disable-snappy :-) - besides 
that, it should build just fine.

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@...1296...
fax: +49-89-35655025                        gert@...1297...

[-- Attachment #2: Type: application/pgp-signature, Size: 291 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

Hi,

On 11/07/14 20:07, Gert Doering wrote:
> Hi,
>
> On Fri, Jul 11, 2014 at 04:50:38PM +0200, Jan Just Keijser wrote:
>> the master branch (from openvpn-testing) currently does not build on
>> either CentOS 5 and 6.
> Install snappy(-dev) or run configure with --disable-snappy :-) - besides
> that, it should build just fine.
>
>
an
   autoreconf -i
did the trick on CentOS 6; on CentOS 5 I get

checking for SSL_OP_NO_TICKET flag in OpenSSL... no
configure: error: OpenVPN 2.4+ requires SSL_OP_NO_TICKET in OpenSSL

which is logical as the "stock" openssl lib on CentOS 5 is openssl 0.9.8 
; to me, this breaks CentOS 5 builds, as I have no option nor desire to 
build & install openssl 1.0 on them.

share and enjoy,

JJK

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Steffan Karger]

Hi,

On 11-07-14 20:17, Jan Just Keijser wrote:
> on CentOS 5 I get
> 
> checking for SSL_OP_NO_TICKET flag in OpenSSL... no
> configure: error: OpenVPN 2.4+ requires SSL_OP_NO_TICKET in OpenSSL
> 
> which is logical as the "stock" openssl lib on CentOS 5 is openssl 0.9.8 
> ; to me, this breaks CentOS 5 builds, as I have no option nor desire to 
> build & install openssl 1.0 on them.

Hmm, I actually checked that 0.9.8 (the minimum version supported by
2.4) has SSL_OP_NO_TICKET. From the OpenSSL changelog:

"Changes between 0.9.8e and 0.9.8f  [11 Oct 2007]
[...]
If a client or server wishes to disable RFC4507 support then the option
SSL_OP_NO_TICKET can be set."

I expected distributions had incorporated that by now. Not sure whether
to add extra code here, or just declare it ancient and let distros/users
either backport newer OpenSSL or use 2.3.

-Steffan

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Gert Doering]

[-- Attachment #1: Type: text/plain, Size: 1738 bytes --]

Hi,

On Fri, Jul 11, 2014 at 08:17:50PM +0200, Jan Just Keijser wrote:
> On 11/07/14 20:07, Gert Doering wrote:
> >On Fri, Jul 11, 2014 at 04:50:38PM +0200, Jan Just Keijser wrote:
> >>the master branch (from openvpn-testing) currently does not build on
> >>either CentOS 5 and 6.
> >Install snappy(-dev) or run configure with --disable-snappy :-) - besides
> >that, it should build just fine.
>
> an
>   autoreconf -i
> did the trick on CentOS 6; 

Yeah, you'll need that if you refresh the git tree and configure.ac 
or Makefile.am has changed.  Sorry, we should display it more prominently :)

> on CentOS 5 I get
> 
> checking for SSL_OP_NO_TICKET flag in OpenSSL... no
> configure: error: OpenVPN 2.4+ requires SSL_OP_NO_TICKET in OpenSSL
> 
> which is logical as the "stock" openssl lib on CentOS 5 is openssl 0.9.8 
> ; to me, this breaks CentOS 5 builds, as I have no option nor desire to 
> build & install openssl 1.0 on them.

Mmmmh.  Steffann has already answered this - it "should" work with recent
versions of 0.9.8, but some of the distribution vendors have the nasty
habit of not bumping their versions but just backporting security fixes -
so your 0.9.8 might be "0.9.8d + security fixes", not "0.9.8-current"
(which would be 0.9.8za).

You could build using polarssl :-) - but that will require installation
of a separate ssl library as well, so, not much better than "build &
install openssl 1.0".  Hrmph.

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@...1296...
fax: +49-89-35655025                        gert@...1297...

[-- Attachment #2: Type: application/pgp-signature, Size: 291 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Jan Just Keijser]

Hi,

On 11/07/14 20:35, Steffan Karger wrote:
> Hi,
>
> On 11-07-14 20:17, Jan Just Keijser wrote:
>> on CentOS 5 I get
>>
>> checking for SSL_OP_NO_TICKET flag in OpenSSL... no
>> configure: error: OpenVPN 2.4+ requires SSL_OP_NO_TICKET in OpenSSL
>>
>> which is logical as the "stock" openssl lib on CentOS 5 is openssl 0.9.8
>> ; to me, this breaks CentOS 5 builds, as I have no option nor desire to
>> build & install openssl 1.0 on them.
> Hmm, I actually checked that 0.9.8 (the minimum version supported by
> 2.4) has SSL_OP_NO_TICKET. From the OpenSSL changelog:
>
> "Changes between 0.9.8e and 0.9.8f  [11 Oct 2007]
> [...]
> If a client or server wishes to disable RFC4507 support then the option
> SSL_OP_NO_TICKET can be set."
>
> I expected distributions had incorporated that by now. Not sure whether
> to add extra code here, or just declare it ancient and let distros/users
> either backport newer OpenSSL or use 2.3.
>
>
RHEL5 (on which CentOS 5 is based) includes openssl 0.9.8e + bug fixes 
and apparently no SSL_OP_NO_TICKET ; this will not change for the 
duration of RHEL5 which is a couple of more years, IIRC.

my question would be : why does openvpn need SSL_OP_NO_TICKET? why not 
#ifdef the code, e.g.

   SSL_CTX_set_options(ctx, SSL_OP_NO_SSLv2 | SSL_OP_NO_SSLv3
   #ifdef SSL_OP_NO_TICKET
      | SSL_OP_NO_TICKET
   #endif
   );

(this is non-openvpn code) ?

JM2CW,

JJK

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[David Sommerseth]

----- Original Message -----
> From: "Jan Just Keijser" <janjust@...1147...>
> To: "Steffan Karger" <steffan@...1856...>, openvpn-devel@lists.sourceforge.net
> Sent: Saturday, 12 July, 2014 1:17:22 AM
> Subject: Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file
> 
> Hi,
> 
> On 11/07/14 20:35, Steffan Karger wrote:
> > Hi,
> >
> > On 11-07-14 20:17, Jan Just Keijser wrote:
> >> on CentOS 5 I get
> >>
> >> checking for SSL_OP_NO_TICKET flag in OpenSSL... no
> >> configure: error: OpenVPN 2.4+ requires SSL_OP_NO_TICKET in OpenSSL
> >>
> >> which is logical as the "stock" openssl lib on CentOS 5 is openssl 0.9.8
> >> ; to me, this breaks CentOS 5 builds, as I have no option nor desire to
> >> build & install openssl 1.0 on them.
> > Hmm, I actually checked that 0.9.8 (the minimum version supported by
> > 2.4) has SSL_OP_NO_TICKET. From the OpenSSL changelog:
> >
> > "Changes between 0.9.8e and 0.9.8f  [11 Oct 2007]
> > [...]
> > If a client or server wishes to disable RFC4507 support then the option
> > SSL_OP_NO_TICKET can be set."
> >
> > I expected distributions had incorporated that by now. Not sure whether
> > to add extra code here, or just declare it ancient and let distros/users
> > either backport newer OpenSSL or use 2.3.
> >
> >
> RHEL5 (on which CentOS 5 is based) includes openssl 0.9.8e + bug fixes
> and apparently no SSL_OP_NO_TICKET ; this will not change for the
> duration of RHEL5 which is a couple of more years, IIRC.
> 
> my question would be : why does openvpn need SSL_OP_NO_TICKET? why not
> #ifdef the code, e.g.
> 
>    SSL_CTX_set_options(ctx, SSL_OP_NO_SSLv2 | SSL_OP_NO_SSLv3
>    #ifdef SSL_OP_NO_TICKET
>       | SSL_OP_NO_TICKET
>    #endif
>    );
> 
> (this is non-openvpn code) ?

I agree, this seems to be the best approach.  I suspect the issues related to 
SSL_OP_NO_TICKET has been resolved by Red Hat on EL5, but the fix may not have
provided that flag.

In addition, EL5 is the oldest distro we support (after we convinced James to
ditch EL4 when that went officially EOL; not counting the extended support
some few customers may pay for).


--
kind regards,

David Sommerseth

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Gert Doering]

[-- Attachment #1: Type: text/plain, Size: 2073 bytes --]

Hi,

On Sat, Jul 12, 2014 at 11:08:46AM +0200, David Sommerseth wrote:
> > my question would be : why does openvpn need SSL_OP_NO_TICKET? why not
> > #ifdef the code, e.g.
> > 
> >    SSL_CTX_set_options(ctx, SSL_OP_NO_SSLv2 | SSL_OP_NO_SSLv3
> >    #ifdef SSL_OP_NO_TICKET
> >       | SSL_OP_NO_TICKET
> >    #endif
> >    );
> > 
> > (this is non-openvpn code) ?
> 
> I agree, this seems to be the best approach.  I suspect the issues related to 
> SSL_OP_NO_TICKET has been resolved by Red Hat on EL5, but the fix may not have
> provided that flag.

This is what we do in 2.3.  For 2.4 we decided (on the list, btw!) to 
require that this is there, instead of silently hoping that the SSL library
would get it right (which it does not, according to James).  See 
25f4d4b49bff342fd9dd54cd and e9b088b208479.

commit 25f4d4b49bff342fd9dd54cd22f14c9de49e9f8b
Author: James Yonan <james@...515...>
Date:   Sun Mar 16 18:49:36 2014 -0600

    Set SSL_OP_NO_TICKET flag in SSL context for OpenSSL builds, to disable TLS stateless session resumption.
    
    OpenVPN doesn't want or need SSL session renegotiation or
    resumption, as it handles renegotiation on its own.
    
    For this reason, OpenVPN always disables the SSL session cache:
    
    SSL_CTX_set_session_cache_mode (ctx, SSL_SESS_CACHE_OFF)
    
    However, even with the above code, stateless session resumption
    is still possible unless explicitly disabled with the
    SSL_OP_NO_TICKET flag.  This patch does this.


> In addition, EL5 is the oldest distro we support (after we convinced James to
> ditch EL4 when that went officially EOL; not counting the extended support
> some few customers may pay for).

Any way to convince RH to pick up this security-relevant fix?

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@...1296...
fax: +49-89-35655025                        gert@...1297...

[-- Attachment #2: Type: application/pgp-signature, Size: 291 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[David Sommerseth]

----- Original Message -----
> From: "Gert Doering" <gert@...1296...>
> To: "David Sommerseth" <openvpn.list@...1114...>
> Cc: "Jan Just Keijser" <janjust@...1147...>, openvpn-devel@lists.sourceforge.net
> Sent: Saturday, 12 July, 2014 11:41:30 AM
> Subject: Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file
>
> > In addition, EL5 is the oldest distro we support (after we convinced James
> > to
> > ditch EL4 when that went officially EOL; not counting the extended support
> > some few customers may pay for).
> 
> Any way to convince RH to pick up this security-relevant fix?

I haven't had time to check the changelog for openssl on EL5 (not close to /my/ 
computers).  But it wouldn't surprise me if they have already fixed this.  But as
SSL_OP_NO_TICKET was introduced in a newer base release than EL5 uses, they have
most likely solved it in another way.  Maybe by always adding this flag implicit?
(Which would have the benefit of not needing to recompile all OpenSSL
applications with this flaw).

IIRC, the guy overseeing the Secure Response Team in RH is Mark Cox, which again
is also an upstream OpenSSL maintainer.  So I'm quite sure all RH releases have
fixed this issue.


--
kind regards,

David Sommerseth

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Gert Doering]

[-- Attachment #1: Type: text/plain, Size: 684 bytes --]

Hi,

On Sat, Jul 12, 2014 at 12:41:14PM +0200, David Sommerseth wrote:
> IIRC, the guy overseeing the Secure Response Team in RH is Mark Cox, which again
> is also an upstream OpenSSL maintainer.  So I'm quite sure all RH releases have
> fixed this issue.

Well, OpenSSL considers this a "feature", not an "issue"... and being
able to turn off session resumption is also considered a "feature"...

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@...1296...
fax: +49-89-35655025                        gert@...1297...

[-- Attachment #2: Type: application/pgp-signature, Size: 291 bytes --]

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[David Sommerseth]

----- Original Message -----
> From: "Gert Doering" <gert@...1296...>
> To: "David Sommerseth" <openvpn.list@...1114...>
> Cc: "Gert Doering" <gert@...1296...>, "Jan Just Keijser" <janjust@...1147...>, openvpn-devel@lists.sourceforge.net
> Sent: Saturday, 12 July, 2014 1:31:09 PM
> Subject: Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file
> 
> Hi,
> 
> On Sat, Jul 12, 2014 at 12:41:14PM +0200, David Sommerseth wrote:
> > IIRC, the guy overseeing the Secure Response Team in RH is Mark Cox, which
> > again
> > is also an upstream OpenSSL maintainer.  So I'm quite sure all RH releases
> > have
> > fixed this issue.
> 
> Well, OpenSSL considers this a "feature", not an "issue"... and being
> able to turn off session resumption is also considered a "feature"...

Ahh, right!

I thought this was related to a CVE, but it seems not, according the
OpenSSL changelog I found on the net [1].  SSL_OP_NO_TICKET comes with the 
implementation of RFC4507, which was introduced in 0.9.8f.  I just double
checked EL5, and it uses 0.9.8e as the base version.  According to the 
RPM changelog, I don't see that RFC4507 has ever been backported.

From what I can grasp out of the OpenSSL changelog, RFC4507 is enabled by
default *if* compiled into OpenSSL.  And if it is compiled into OpenSSL, it can
be disabled by setting the SSL_OP_NO_TICKET flag.  Which would means 0.9.8e
should be safe in this aspect and EL5 is not vulnerable at all.

But it would be good if others can double this and see if I've understood this
correctly, just so I don't say anything wrong.

[1] <https://www.openssl.org/news/openssl-0.9.8-notes.html>


--
kind regards,

David Sommerseth

Re: [Openvpn-devel] [PATCH] Add topology in sample server configuration file

[Gert Doering]

[-- Attachment #1: Type: text/plain, Size: 1222 bytes --]

Hi,

On Sat, Jul 12, 2014 at 02:55:21PM +0200, David Sommerseth wrote:
> > Well, OpenSSL considers this a "feature", not an "issue"... and being
> > able to turn off session resumption is also considered a "feature"...
> 
> Ahh, right!
> 
> I thought this was related to a CVE, but it seems not, according the
> OpenSSL changelog I found on the net [1].  SSL_OP_NO_TICKET comes with the 
> implementation of RFC4507, which was introduced in 0.9.8f.  I just double
> checked EL5, and it uses 0.9.8e as the base version.  According to the 
> RPM changelog, I don't see that RFC4507 has ever been backported.

Ah.  So if that is correct, our simplistic implementation "if it is not
there, just #define SSL_OP_NO_TICKET 0" would be perfectly safe in this
regard, then.

> But it would be good if others can double this and see if I've understood this
> correctly, just so I don't say anything wrong.

Indeed :-) - Steffann?

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@...1296...
fax: +49-89-35655025                        gert@...1297...

[-- Attachment #2: Type: application/pgp-signature, Size: 291 bytes --]

Re: [Openvpn-devel] more SSL_OP_NO_TICKET fallout

[Steffan Karger]

[-- Attachment #1: Type: text/plain, Size: 2525 bytes --]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Hi,

Oh, joy, more SSL_OP_NO_TICKET fallout. Changing the subject to
separate this discussion from the original thread about topology subnet.

On 12-07-14 15:50, Gert Doering wrote:
> On Sat, Jul 12, 2014 at 02:55:21PM +0200, David Sommerseth wrote:
>>> Well, OpenSSL considers this a "feature", not an "issue"... and
>>> being able to turn off session resumption is also considered a
>>> "feature"...
>> 
>> Ahh, right!
>> 
>> I thought this was related to a CVE, but it seems not, according
>> the OpenSSL changelog I found on the net [1].  SSL_OP_NO_TICKET
>> comes with the implementation of RFC4507, which was introduced in
>> 0.9.8f.  I just double checked EL5, and it uses 0.9.8e as the
>> base version.  According to the RPM changelog, I don't see that
>> RFC4507 has ever been backported.
> 
> Ah.  So if that is correct, our simplistic implementation "if it is
> not there, just #define SSL_OP_NO_TICKET 0" would be perfectly safe
> in this regard, then.
> 
>> But it would be good if others can double this and see if I've
>> understood this correctly, just so I don't say anything wrong.
> 
> Indeed :-) - Steffann?

Yes, the SSL_OP_NO_TICKET flag was introduced together with the
feature it disables (stateless session resumption). Sane backporters
would also backport the flag when backporting the feature. So using
the same "#define SSL_OP_NO_TICKET 0"-construction as in 2.3 should be
fine.

I think this is the way to go for 2.4 if we want to support RHEL5
until their end-of-production. The alternatives would be to either
drop support for RHEL5 (seems unreasonable for such a silly fix) or
keep maintaining 2.3 until RHEL5 end-of-support (much more work).

Attached a patch that adds the 2.3 #ifdef construction to -master too.
I updated the message to reflect that this is in master too now, and
include the string '0.9.8', which should help with stripping this out
again if 0.9.8-support is ever dropped.

- -Steffan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBCAAGBQJTwlHPAAoJEJgCyj0AftKImwcH/0mCSnA2LNFDP0542/yRnFTc
mwufEafiZ936xY0eEYKERNMuISucAoHiOGrp+5EqSHSoGqX7TTkxZ3hIBVgQ4L37
p4b9pbkXNj5J4RelcFkH78lNKnNr9AnU5JHO1SJEO1qKAY0q5ypRclvTbytcN8o6
6UVnRzcjmDnyUOgPrSct+JK7Rm9BVCpDU2a7PK0fGASs3qS9NZOmBSDV9Eg2P5tn
IdvoZUrI9ZPNLDWCoDpMOkpsvQPl2BeMJ0t5Ib/LRSE8dbvfxg+GldYYQ3YA4uu8
xylgnvNTwi1GfFEicwX+9g8NEkqDKmNJd6k/OzclVyhWi3209E6prwk20NX7RDs=
=u8ld
-----END PGP SIGNATURE-----

[-- Attachment #2: 0001-Define-dummy-SSL_OP_NO_TICKET-flag-if-not-present-in.patch --]
[-- Type: text/x-patch, Size: 2280 bytes --]

From bbe56efb085138dca65dfe15b480eab47074b7e2 Mon Sep 17 00:00:00 2001
From: Steffan Karger <steffan@...1856...>
Date: Sun, 13 Jul 2014 11:26:32 +0200
Subject: [PATCH] Define dummy SSL_OP_NO_TICKET flag if not present in OpenSSL.

This restores support for pre-0.9.8f OpenSSL versions, which do not include
stateless session resumption, and the accompanying SSL_OP_NO_TICKET flag.

Signed-off-by: Steffan Karger <steffan@...1856...>
---
 configure.ac              | 19 -------------------
 src/openvpn/ssl_openssl.h | 11 +++++++++++
 2 files changed, 11 insertions(+), 19 deletions(-)

diff --git a/configure.ac b/configure.ac
index 117eaf6..0d0ab88 100644
--- a/configure.ac
+++ b/configure.ac
@@ -814,25 +814,6 @@ if test "${have_openssl_crypto}" = "yes"; then
 	LIBS="${saved_LIBS}"
 fi
 
-if test "${enable_ssl}" = "yes" && test "${with_crypto_library}" = "openssl";
-then
-    saved_CPPFLAGS="${CPPFLAGS}"
-    CPPFLAGS="${CPPFLAGS} ${OPENSSL_CRYPTO_CFLAGS}"
-    AC_MSG_CHECKING([for SSL_OP_NO_TICKET flag in OpenSSL])
-    AC_EGREP_CPP(have_ssl_op_no_ticket, [
-    #include <openssl/ssl.h>
-    #ifdef SSL_OP_NO_TICKET
-    have_ssl_op_no_ticket
-    #endif
-    ], [
-        AC_MSG_RESULT([yes])
-    ], [
-        AC_MSG_RESULT([no])
-        AC_ERROR([OpenVPN 2.4+ requires SSL_OP_NO_TICKET in OpenSSL])
-    ])
-    CPPFLAGS="${saved_CPPFLAGS}"
-fi
-
 AC_ARG_VAR([POLARSSL_CFLAGS], [C compiler flags for polarssl])
 AC_ARG_VAR([POLARSSL_LIBS], [linker flags for polarssl])
 have_polarssl_ssl="yes"
diff --git a/src/openvpn/ssl_openssl.h b/src/openvpn/ssl_openssl.h
index fc2052c..97dc742 100644
--- a/src/openvpn/ssl_openssl.h
+++ b/src/openvpn/ssl_openssl.h
@@ -33,6 +33,17 @@
 #include <openssl/ssl.h>
 
 /**
+ * SSL_OP_NO_TICKET tells OpenSSL to disable "stateless session resumption",
+ * as this is something we do not want nor need, but could potentially be
+ * used for a future attack.  For compatibility reasons we keep building if the
+ * OpenSSL version is too old (pre-0.9.8f) to support stateless session
+ * resumption (and the accompanying SSL_OP_NO_TICKET flag).
+ */
+#ifndef SSL_OP_NO_TICKET
+# define SSL_OP_NO_TICKET 0
+#endif
+
+/**
  * Structure that wraps the TLS context. Contents differ depending on the
  * SSL library used.
  */
-- 
1.9.1


[-- Attachment #3: 0001-Define-dummy-SSL_OP_NO_TICKET-flag-if-not-present-in.patch.sig --]
[-- Type: application/pgp-signature, Size: 287 bytes --]

[Openvpn-devel] [PATCH applied] Re: Define dummy SSL_OP_NO_TICKET flag if not present in OpenSSL.

[Gert Doering]

ACK.  RHEL5 is still supported, and with the additional background, this
is good enough.

Your patch has been applied to the master branch.

commit 97bd862ed5c22956cb4405eabae64cf55cabb0d3

Author: Steffan Karger
Date:   Sun Jul 13 11:26:32 2014 +0200

     Define dummy SSL_OP_NO_TICKET flag if not present in OpenSSL.

     Signed-off-by: Steffan Karger <steffan@...1856...>
     Acked-by: Gert Doering <gert@...1296...>
     Message-Id: <53C251E2.7050605@...1856...>
     URL: http://article.gmane.org/gmane.network.openvpn.devel/8902
     Signed-off-by: Gert Doering <gert@...1296...>


--
kind regards,

Gert Doering

[Openvpn-devel] [PATCH applied] Re: Add topology in sample server configuration file

[Gert Doering]

ACK.

Your patch has been applied to the master and release/2.3 branch.

commit c277757fcf7fb4c2713db154439f937d48cfae61 (master)
commit 6fa97d03de3c70efc59e9fa8a606de12e2135620 (release/2.3)

Author: Philipp Hagemeister
Date:   Fri Jul 11 10:00:24 2014 +0200

     Add topology in sample server configuration file

     Signed-off-by: Philipp Hagemeister <phihag@...1921...>
     Acked-by: Gert Doering <gert@...1296...>
     Message-Id: <53BF9998.5020906@...1921...>
     URL: http://article.gmane.org/gmane.network.openvpn.devel/8878
     Signed-off-by: Gert Doering <gert@...1296...>


--
kind regards,

Gert Doering