ssl with client authentication between cxf and IIS

Question

So, I've been wrestling with a two way ssl setup between Tomcat6/Camel/Cxf to IIS6 the last few days.

A quick overview: I have a cxf consumer running in Camel in Tomcat, which calls a webservice hosted with a third party on IIS. The third party is using a 'valid' certificate, whereas we are using a self-signed certificate. IIS is configured to require client authentication.

I've gotten to the point where cxf is configured correctly, and mutual authentication is being performed.

Some gotcha's I needed to fix to get there:

• allowUnsafeRenegotiation=true
• https.protocols=TLSv1 (else java would send a SSLv2 client hello in the middle of the handshake)
• remove the default ciphersuite filter that I copied from most CXF examples, it was too strict

All certificates, firewall settings etc. are correct, as I'm able to do a call to the third party using curl and I get a valid SOAP response back.

The problem: I get a timeout exception when sending a SOAP POST message from my application.

Looking into the ssl debug log there are some strange errors (I tried to remove all unnecessary stuff):

trigger seeding of SecureRandom
done seeding SecureRandom
Allow unsafe renegotiation: true
Allow legacy hello messages: true
Is initial handshake: true
Is secure renegotiation: false
http-9203-1, setSoTimeout(120000) called
%% No cached client session
*** ClientHello, TLSv1
http-9203-1, WRITE: TLSv1 Handshake, length = 111
http-9203-1, READ: TLSv1 Handshake, length = 2454
*** ServerHello, TLSv1
RandomCookie:  GMT: 1354816328 bytes = { 63, 194, 181, 182, 56, 251, 234, 198, 234, 92, 162, 175, 243, 127, 133, 182, 0, 237, 102, 9, 37, 133, 141, 3, 175, 120, 34, 92 }
Session ID:  {36, 10, 0, 0, 166, 131, 36, 81, 251, 181, 83, 154, 76, 240, 235, 82, 39, 135, 239, 235, 231, 195, 17, 225, 220, 59, 105, 207, 116, 185, 114, 172}
Cipher Suite: SSL_RSA_WITH_RC4_128_MD5
Compression Method: 0
***
Warning: No renegotiation indication extension in ServerHello
%% Created:  [Session-1, SSL_RSA_WITH_RC4_128_MD5]
** SSL_RSA_WITH_RC4_128_MD5
*** Certificate chain
***
Found trusted certificate:
*** ServerHelloDone
*** ClientKeyExchange, RSA PreMasterSecret, TLSv1
http-9203-1, WRITE: TLSv1 Handshake, length = 262
http-9203-1, WRITE: TLSv1 Change Cipher Spec, length = 1
*** Finished
verify_data:  { 124, 172, 175, 239, 103, 187, 134, 164, 206, 241, 145, 41 }
***
http-9203-1, WRITE: TLSv1 Handshake, length = 32
http-9203-1, READ: TLSv1 Change Cipher Spec, length = 1
http-9203-1, READ: TLSv1 Handshake, length = 32
*** Finished
verify_data:  { 84, 207, 104, 255, 19, 151, 27, 11, 159, 84, 124, 216 }
***
%% Cached client session: [Session-1, SSL_RSA_WITH_RC4_128_MD5]
[read] MD5 and SHA1 hashes:  len = 16
0000: 14 00 00 0C 54 CF 68 FF   13 97 1B 0B 9F 54 7C D8  ....T.h......T..
Padded plaintext before ENCRYPTION:  len = 365
0000: 50 4F 53 54 20 2F 69 73   61 2E 64 6C 6C 2F 41 43  POST /isa.dll/AC
http-9203-1, WRITE: TLSv1 Application Data, length = 365
Padded plaintext before ENCRYPTION:  len = 1200
0000: 3C 73 6F 61 70 3A 45 6E   76 65 6C 6F 70 65 20 78  <soap:Envelope x
http-9203-1, WRITE: TLSv1 Application Data, length = 1200
default-workqueue-1, READ: TLSv1 Handshake, length = 20

This is the first handshake without client authentication (as IIS is using re-negotiation to get the client certificate). Note that a first attempt to post the SOAP message is already been made (is that normal?).

Allow unsafe renegotiation: true
Allow legacy hello messages: true
Is initial handshake: false
Is secure renegotiation: false
*** HelloRequest (empty)
Warning: continue with insecure renegotiation
%% Client cached [Session-1, SSL_RSA_WITH_RC4_128_MD5]
%% Try resuming [Session-1, SSL_RSA_WITH_RC4_128_MD5] from port 58460
*** ClientHello, TLSv1
default-workqueue-1, WRITE: TLSv1 Handshake, length = 159
default-workqueue-1, READ: TLSv1 Handshake, length = 5416
*** ServerHello, TLSv1
%% Created:  [Session-2, SSL_RSA_WITH_RC4_128_MD5]
** SSL_RSA_WITH_RC4_128_MD5
*** Certificate chain
***
Found trusted certificate:
*** CertificateRequest
*** ServerHelloDone
matching alias: servercert
*** Certificate chain
*** ClientKeyExchange, RSA PreMasterSecret, TLSv1
default-workqueue-1, WRITE: TLSv1 Handshake, length = 909
*** CertificateVerify
default-workqueue-1, WRITE: TLSv1 Handshake, length = 150
[Raw write]: length = 155
default-workqueue-1, WRITE: TLSv1 Change Cipher Spec, length = 17
*** Finished
verify_data:  { 242, 160, 100, 95, 172, 14, 166, 71, 158, 148, 220, 42 }
***
default-workqueue-1, WRITE: TLSv1 Handshake, length = 32
default-workqueue-1, READ: TLSv1 Change Cipher Spec, length = 17
default-workqueue-1, READ: TLSv1 Handshake, length = 32
*** Finished
verify_data:  { 115, 32, 183, 236, 113, 63, 144, 117, 126, 132, 150, 67 }
***
%% Cached client session: [Session-2, SSL_RSA_WITH_RC4_128_MD5]
default-workqueue-1, handling exception: java.net.SocketException: Connection reset
%% Invalidated:  [Session-2, SSL_RSA_WITH_RC4_128_MD5]
default-workqueue-1, SEND TLSv1 ALERT:  fatal, description = unexpected_message
default-workqueue-1, WRITE: TLSv1 Alert, length = 18
default-workqueue-1, Exception sending alert: java.net.SocketException: Broken pipe
default-workqueue-1, called closeSocket()
default-workqueue-1, called close()
default-workqueue-1, called closeInternal(true)

The second handshake sends the client certificate to the server, I removed the detailed logging but the certificate was found and ok-ed. As you can see this handshake already ends with an error (without any delay in time).

Allow unsafe renegotiation: true
Allow legacy hello messages: true
Is initial handshake: true
Is secure renegotiation: false
default-workqueue-1, setSoTimeout(120000) called
%% No cached client session
*** ClientHello, TLSv1
default-workqueue-1, WRITE: TLSv1 Handshake, length = 111                                      
default-workqueue-1, READ: TLSv1 Handshake, length = 2454
*** ServerHello, TLSv1
Warning: No renegotiation indication extension in ServerHello
%% Created:  [Session-3, SSL_RSA_WITH_RC4_128_MD5]
** SSL_RSA_WITH_RC4_128_MD5
***
Found trusted certificate:
*** ServerHelloDone
*** ClientKeyExchange, RSA PreMasterSecret, TLSv1
default-workqueue-1, WRITE: TLSv1 Handshake, length = 262
... no IV used for this cipher
default-workqueue-1, WRITE: TLSv1 Change Cipher Spec, length = 1
*** Finished
verify_data:  { 130, 65, 225, 31, 41, 14, 21, 42, 102, 176, 151, 245 }
***
default-workqueue-1, WRITE: TLSv1 Handshake, length = 32
default-workqueue-1, READ: TLSv1 Change Cipher Spec, length = 1
default-workqueue-1, READ: TLSv1 Handshake, length = 32
*** Finished
verify_data:  { 125, 78, 165, 103, 115, 60, 111, 169, 180, 174, 0, 169 }
***
%% Cached client session: [Session-3, SSL_RSA_WITH_RC4_128_MD5]
default-workqueue-1, WRITE: TLSv1 Application Data, length = 365
default-workqueue-1, handling exception: java.net.SocketTimeoutException: Read timed out
default-workqueue-1, called close()
default-workqueue-1, called closeInternal(true)
default-workqueue-1, SEND TLSv1 ALERT:  warning, description = close_notify
default-workqueue-1, WRITE: TLSv1 Alert, length = 18

Then a third handshake is performed (normal?), and again a POST of the SOAP message is done (the write of the application date). After that POST the logging waits until the socket timeout expires (I've put it on 2 minutes or 10 secs, doesn't make a difference), after which I get a SocketTimeoutException. I tried switching to SSLv3 or 2, but that gives exactly the same result.

Any ideas what it causing this behaviour?

Answer 1

Should anyone ever encounter this same problem, this solved it eventually for me. I asked our third party to set this flag to true, since we only communicate with each other that shouldn't be a problem. Setting the flag to true disables renegotation, thus skipping the part in the ssl setup where it went wrong.