On 10/10/2011 12:45 PM, Mario Ivankovits wrote:
Hi! back again ;-)
To my surprise Digest will be chosen first (this is documented in
sun.net.www.protocol.http.AuthenticationHeader, but wrong I think, shouldn't it be read
like this: negotiate -> kerberos -> ntlm -> digest -> basic) - also no
other method will
be tried yet.
This is not a surprise. In order for Negotiate to happen, you need some
Kerberos settings and without them it will not go very long. If I remember
correctly, Digest is preferred to NTLM so it's the natural fallback.
Just the part, that digest is preferred over NTLM was the surprising part. If
you've configured an Swing-Authenticator this will prevent the application from
transparently authenticate against the server, no?
And then, also that the system did not try to NTLM, after Digest failed,
surprises me.
I don't think there is a fallback mechanism from Digest to NTLM. If
digest fails, then probably the whole auth just fails. You can specify
the system property "http.auth.preference" to "NTLM" if you always want
it to be used.
That seems to be the only approach. But I still hope it is a server
configuration error.
I've checked this against 3 different Exchange Server (2007+2010), all show the
same behavior. Two of them are installed by me and one is a server from an
exchange server hoster (http://itsolutionsnow.com/)
And since my iPhone works against all these 3 server and my Java 1.6 App works
too, I don't think this is a server issue. We are just digging in undocumented
areas here ...
Instead of passing "dontUseNegotiate" to the AuthenticationHeader, we will pass
in a Set of already tried authentication methods.
I'm not sure of that. The code logic here is quite fragile and it has underdone
several rounds of update and tweaking, and I surely don't like to see anything
broken.
I see that the code is ... well ... evolved and highly complicated. And I
perfectly understand that you would not like to change much here as it seems
VERY hard to test.
A rather ugly hack is to choose NTLM as long as it has extra parameter(s).
Hopefully that's safe. We know NTLM has defined 3 messages and that means the
final answer from server is simply 200 OK without any 4th confirmation...
Yep, that will fix my main issue, but will not help in trying all the possible
authentication methods announced by the server.
Why not? If it's the first time NTLM is proposed as a scheme, it should
not have any parameter. Therefore, I don't see a possibility that other
schemes will be ignored.
I haven't touched these codes for some time, and forget if the parsing
of WWW-Authenticate header(s) have any difference when it (or they) is
presented the first time and during the handshake.
I couldn't resist, sorry, attached you will find a diff and the source files I
had to change to patch both issues (in my environment at least). I active them
in an bootclasspath/p environment. Don't be too curious about
AuthenticationHeader2, I just had to rename it to avoid a conflict with javas
original class as I changed the constructor.
As you can see, the patch is not that big. Which does not mean it doesn't
breaks something ... I'll go with them in my test environment for some days.
I would be more than happy if you could give it an hour or two to have a look
at it. Unhappily, I am sure you might need more time to test all use-cases, but
I am still hoping ... ;-)
Also, if we'd like to keep Digest before NTLM, I'd vote for adding a property to change
the ordering of the authentication methods. But this means have you accept my patch
first, else the fallback does not work "as expected" anyway.
That's "http.auth.preference" again, we now only accept one value for
this property, so you can make it "NTLM" but not "NTLM,Digest".
I can take a look at the patch, but honestly, I'm so afraid of touching
the HttpURLConnection class and its siblings.
Thanks
Max
Well, probably we can (if required, and that is for sure) bugfix my patch and
have fixed two issues with it then. ;-)
Thanks for your time!
Ciao,
Mario
This also should make it possible to avoid the use of inNegotiate.
What do you think?
I think I am going to play a bit with this idea ... :-)
Ciao,
Mario
[1] http://www.ietf.org/rfc/rfc2616.txt
[2] http://msdn.microsoft.com/en-us/library/aa479391.aspx
-----Ursprüngliche Nachricht-----
Von: Weijun Wang [mailto:weijun.w...@oracle.com]
Gesendet: Montag, 10. Oktober 2011 17:33
An: Chris Hegarty
Cc: Mario Ivankovits; net-dev@openjdk.java.net
Betreff: Re: ntlm with ms exchange server not working since java 1.7
During an NTLM handshake, I've never seen a server mentioning another scheme.
As seen in message #4, the NTLM header still contains data, so there should not
be WWW-Authenticate: Negotiate header.
That said, this is only my experience. I tried to find any words on this from
an RFC but failed.
-Max
On Oct 10, 2011, at 7:48 AM, Chris Hegarty<chris.hega...@oracle.com> wrote:
Max [to'ed],
Does this look familiar? Is it wrong for the server to be returning
"WWW-Authenticate: Negotiate" during NTLM handshake?
-Chris.
On 08/10/2011 14:41, Mario Ivankovits wrote:
Hi net-devs,
I hope you do not mind that I post to this list, but I hope I can
provide enough in-depth information about the problem to justify the
post here.
Accessing a “normal” ntlm protected resource – a simple index.html
in an protected directory on an IIS 7.5 server - the ntlm
authentication works fine.
However, trying to access the Microsoft Exchange 2010 webservice
failes with “401 Unauthorized”.
I used this few lines to debug the connection/authentication process
URL url = new URL("https://exchange/ews/Services.wsdl";);
byte[] buf = new byte[10240];
int read = url.openStream().read(buf);
System.err.println(new String(buf, 0, read));
This snipped works fine in java 1.6, but failes with an IOException
(http status 401) in java 1.7.
I found an interesting difference when accessing the “normal”
web-page and the exchange webservice.
When accessing the web-page, the server answers “WWW-Authenticate:
Negotiate” just after the first 401 response which triggers the
authentication process then. In contrast, when accessing the
Exchange webservice the “WWW-Authenticate: Negotiate” is sent during
the negotiation process too, which then triggers the inNegotiate
flag in sun.net.www.protocol.http.HttpURLConnection in
getInputStream and let the negotiation process fail.
If I hack the response values and change any subsequent Negotiate to
e.g. NegotiateXX, then the inNegotiate flag will not change and the
authentication process will finish and authentication finally works.
Here is the request/response cycle which fail then:
#1: {GET /ews/Services.wsdl HTTP/1.1: null}{User-Agent:
Java/1.7.0_02-ea}{Host: exchange }{Accept: text/html, image/gif,
image/jpeg, *; q=.2, */*; q=.2}{Connection: keep-alive}
#2: {null: HTTP/1.1 401 Unauthorized}{Server:
Microsoft-IIS/7.5}{WWW-Authenticate: Negotiate}{WWW-Authenticate:
NTLM}{X-Powered-By: ASP.NET}{Date: Sat, 08 Oct 2011 13:17:39
GMT}{Content-Length: 0}
#3: {GET /ews/Services.wsdl HTTP/1.1: null}{User-Agent:
Java/1.7.0_02-ea}{Host: exchange }{Accept: text/html, image/gif,
image/jpeg, *; q=.2, */*; q=.2}{Connection: keep-alive}{Authorization:
NTLM MY_NTLM_DATA}
#4: {null: HTTP/1.1 401 Unauthorized}{Server:
Microsoft-IIS/7.5}{WWW-Authenticate: NTLM
SERVER_NTLM_DATA}{WWW-Authenticate: Negotiate}{X-Powered-By:
ASP.NET}{Date: Sat, 08 Oct 2011 13:17:39 GMT}{Content-Length: 0}
Exception in thread "main" java.io.IOException: Server returned HTTP
response code: 401 for URL: https://exchange/ews/Services.wsdl
at
sun.net.www.protocol.http.HttpURLConnection.getInputStream(HttpURLCo
n
nection.java:1612)
at
sun.net.www.protocol.https.HttpsURLConnectionImpl.getInputStream(Htt
p
sURLConnectionImpl.java:254)
at java.net.URL.openStream(URL.java:1035)
Does this make sense to you?
It seems to me the “inNegotiate” handling needs a review as it does
not work in all cases.
I hope my informations are of any help to fix this issue.
Ciao,
Mario
