[389-users] StartTLS with F14, 389-console and system-config-authentication

[Bob McKay]

I'm trying to do set up authentication between fedora 14 client and server
using ldap passwords. I seem to be having a problem getting the server to
accept startTLS. It has all been set up with 389-console and
system-config-authentication.

In more detail, I set up the client using system-config-authentication, to
use LDAP for the user account authentication, TLS to encrypt connections,
and LDAP passwords. I set the LDAP server to ldap://<server name>. All this
seems to be working OK, on the server (wireshark) I can see the initial ldap
handshake on port 389, as expected. The problem seems to start at the next
stage. I see a tcp packet to port 389 from the client, protocol LDAP, Info
extendedReq(1) LDAP_START_TLS_OID, which I assume is the request packet from
the client to start the TLS handshake. I see a TCP ack from the server, but
this is immediately followed by a protocol LDAP message from the server to
the client, extendedResp(1) (unsupported extended operation). I assume this
is a failure message. However the client replies with an SSLv2 Client Hello.
There is never an SSL response,

On the client, I see in the logs
ssd[be[LDAP]]: Could not start TLS encryption. unsupported extended
operation
which I assume confirms that the attempt to start up TLS failed.

In the server  dirsrv access log, I see:

[08/Apr/2011:16:40:46 +0900] conn=12 fd=64 slot=64 connection from
192.168.1.7 to 192.168.1.192
[08/Apr/2011:16:40:46 +0900] conn=12 op=0 EXT oid="1.3.6.1.4.1.1466.20037"
[08/Apr/2011:16:40:46 +0900] conn=12 op=0 RESULT err=2 tag=120 nentries=0
etime=0
[08/Apr/2011:16:40:46 +0900] conn=12 op=-1 fd=64 closed error 71 (Protocol
error) - B1

and in the error log (logging verbosity turned up a bit) I see

[08/Apr/2011:17:13:22 +0900] - new connection on 64
[08/Apr/2011:17:13:22 +0900] - activity on 64r
[08/Apr/2011:17:13:22 +0900] - read activity on 64
[08/Apr/2011:17:13:22 +0900] - conn 16 activity level = 0
[08/Apr/2011:17:13:22 +0900] - listener got signaled
[08/Apr/2011:17:13:22 +0900] - flush_ber() wrote 44 bytes to socket 64
[08/Apr/2011:17:13:22 +0900] - activity on 64r
[08/Apr/2011:17:13:22 +0900] - read activity on 64
[08/Apr/2011:17:13:22 +0900] - conn=16 received a non-LDAP message (tag
0x80, expected 0x30)
[08/Apr/2011:17:13:22 +0900] - conn 16 leaving turbo mode due to 3
[08/Apr/2011:17:13:22 +0900] - listener got signaled


All this seems to suggest that either I don't have startTLS built into my
installation of 389, or I'm somehow failing to enable it. However I can't
see any plugin in yum that would help, and the only place I can see (at
least in 389-console) that looks like it could affect this is the Domain
name "Secure connection" checkbox. However I've tried toggling this and I
see exactly the same behaviour either way (I'm guessing that the checkbox
actually enables SSL support on port 636, but the documentation isn't too
clear on this).

One possibility, that I'm not too sure how to check, is that the
(self-signed) CA certificate or the server certificate might be bad. However
they look fine on dumping, and both appear to be installed in the server.
And hopefully, I would have got rather more informative log messages in this
case. I'm not too sure how to test this locally on the server, any pointers
on how to do this would be great.

Any help anyone can give on this would be _hugely_ appreciated - I am now
three days in on what looked like it was supposed to be a straightforward
install, with no real progress, and getting further behind on other urgent
stuff. Urrgggg.

System:
Client and server:
F14, 2.6.35.11-83.fc14.x86_64 kernel

Server:
slapd 2.4.23

Client:
sssd 64-bit 1.5.4-1.fc14


      Best Wishes
      Bob
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.fedoraproject.org/pipermail/389-users/attachments/20110408/ebd41ebd/attachment.html>