On Sat, Jun 14, 2014 at 04:42:19PM +0000, Viktor Dukhovni wrote:
> On Sat, Jun 14, 2014 at 04:23:13PM +0200, Kurt Roeckx via RT wrote:
> 
> > > Yes.  As far as I can see all reports are about 0.9.8o sending
> > > large amounts of data to 1.0.1e.
> > 
> > So I can reproduce it.  But I can only seem to be reproducing it
> > when using postgres having a 1.0.1 talk to a 0.9.8.  For me it
> > happens at exactly the same place in the dump file each time,
> > after 480 MB has been transfered.  Other are reporting it after a
> > different amount.
> 
> Is it perhaps a renegotiation with resumption.  Can you arrange to
> export the session master key in wireshark-compatible form, and
> decrypt the second handshake?
> 
> Which is the client, which is the server, and which one reports
> the "early ccs"?  Have you run the party that complains under a
> debugger with a breakpoint at the line where the problem is reported?
> What is the stack trace and what are the values of the fields of the
> connection's SSL structure?

postgresql has an option ssl_renegotiation_limit.  Lowering that
makes the error appear faster.

So it's 0.9.8o (+patches) (server, sending data) talking to
OpenSSL_1_0_1-stable (client).  After some data transfer I see:
s->c: Hello Request
c->s: Client Hello
s->c: Server Hello, Certificate, Server Hello Done
c->s: Client Key Exchange, Change Cipher Spec, Finished
s->c: Change Cipher Spec, Finished
c->s: Alert (Fatal, Unexpected Message)


kurt


______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       [email protected]
Automated List Manager                           [email protected]

Reply via email to