Date: Wed, 9 Jul 2014 14:23:47 +0100 Subject: Re: Apache Rampart test failure with wss4j 1.6.5 and later From: [email protected] To: [email protected] CC: [email protected]
Well the thing is that WSS4J 1.6.16 was just released, and so the next release won't happen for a couple of months probably. So if you want to see a Rampart release before then, you could just submit a patch to check that the Id isn't empty. Colm. On Wed, Jul 9, 2014 at 1:32 PM, <[email protected]> wrote: I can open a defect in Rampart, but I'm not sure what should be the proposed change there - I was thinking that it can check for empty id tag and skip the result, but if wss4j does not generate results with empty id anymore, this will not be required. Rampart uses the following code to identify the encryption key id from the request, for which a response in generated: for (WSSecurityEngineResult wsSecEngineResult : wsSecEngineResults) { Integer actInt = (Integer) wsSecEngineResult.get(WSSecurityEngineResult.TAG_ACTION); String encrKeyId = (String) wsSecEngineResult.get(WSSecurityEngineResult.TAG_ID); if (actInt == WSConstants.ENCR && encrKeyId != null) { return encrKeyId; } } If you think the above is improper or can be improved, just let me know and I will follow up with Rampart devs. Detelin MG>the rampart distro that accompanies Axis2-1.6.2 detected the missing tag 'id' for (WSSecurityEngineResult wsSecEngineResult : wsSecEngineResults) { Integer actInt = (Integer) wsSecEngineResult.get(WSSecurityEngineResult.TAG_ACTION); if (actInt == WSConstants.ENCR) { if (wsSecEngineResult.get(WSSecurityEngineResult.TAG_ID) != null && ((String) wsSecEngineResult.get(WSSecurityEngineResult.TAG_ID)).length() != 0) { try { String encryptedKeyID = (String) wsSecEngineResult.get(WSSecurityEngineResult.TAG_ID); Date created = new Date(); Date expires = new Date(); expires.setTime(System.currentTimeMillis() + 300000); EncryptedKeyToken tempTok = new EncryptedKeyToken(encryptedKeyID, created, expires); tempTok.setSecret((byte[]) wsSecEngineResult.get(WSSecurityEngineResult.TAG_SECRET)); tempTok.setSHA1(getSHA1((byte[]) wsSecEngineResult. get(WSSecurityEngineResult.TAG_ENCRYPTED_EPHEMERAL_KEY))); rmd.getTokenStorage().add(tempTok); return encryptedKeyID; } catch (TrustException e) { throw new RampartException("errorInAddingTokenIntoStore"); } MG>i would add a else condition to toss a message to the op else { throw new RampartException("security_engine_result_missing_id"); } MG>end else MG>errors.properties would contain new entry: security_engine_result_missing_id=Rampart Security Engine Result is missing 'id' token element MG>if you make the suggestion for correction I will follow thru and make sure it gets implemented MG>https://issues.apache.org/jira/browse MG>Thanks Detelin, MG>Martin On Wed, Jul 9, 2014 at 12:45 PM, Colm O hEigeartaigh <[email protected]> wrote: Thanks for the investigation. It turns out Maven 3.0.x is required to build Rampart. I've merged a "fix" for this issue in WSS4J, where we don't store the token Id if it is an empty String. IMO Rampart should also be fixed. Colm. On Tue, Jul 8, 2014 at 6:03 PM, <[email protected]> wrote: I have not seen these, probably it is the "copy-mars" execution in the integration module that is causing them. It could be some dependency resolution problem for "mar" artifacts, I'm using Maven 3.0.4 and did not experience such issues. I have some more input on the problem - I think that the introduction of an "id" tag for reference list results is confusing Rampart, specifically the first change here: http://svn.apache.org/viewvc/webservices/wss4j/trunk/src/main/java/org/apache/ws/security/processor/ReferenceListProcessor.java?r1=1294114&r2=1294113&pathrev=1294114 In the example request that I attached, there is a ReferenceList element that looks like this: <xenc:ReferenceList xmlns:xenc="http://www.w3.org/2001/04/xmlenc#";><xenc:DataReference URI="#ED-5"/></xenc:ReferenceList> When processing this, the ReferenceListProcessor with the mentioned change now creates a result instance, but with an empty "id" tag, since the ReferenceList element does not have "Id" attribute. The result object looks like this: {id=, data-ref-uris=[org.apache.ws.security.WSDataRef@3e9c6879], action=4, validated-token=false} When generating the response, Rampart's AssymetricBindingBuilder searches for the encrypted key by iterating over the results list and checking for a result with action=4 (ENCR) and a non-empty id tag, see AsymmetricBindingBuilder.setupEncryptedKey and RampartUtil.getRequestEncryptedKeyId methods: http://svn.apache.org/viewvc/axis/axis2/java/rampart/branches/1_6/modules/rampart-core/src/main/java/org/apache/rampart/builder/AsymmetricBindingBuilder.java?view=markup#l868 http://svn.apache.org/viewvc/axis/axis2/java/rampart/branches/1_6/modules/rampart-core/src/main/java/org/apache/rampart/util/RampartUtil.java?view=markup#l1442 Apparently, it now picks up the result of the ReferenceListProcessor since it has an "id" tag, but since it has empty value, the "AssymetricBindingBuilder.encryptedKeyId" field is also left out empty and this leads to missing token in response... Commenting out the line in the ReferenceListProcessor that adds the "id" tag fixes the issue - Rampart then properly finds the result of the DerivedKeyTokenProcessor and not the one of the ReferenceListProcessor. Now the question is whether this has to be fixed in Rampart or in WSS4J? Regards, Detelin On Tue, Jul 8, 2014 at 6:39 PM, Colm O hEigeartaigh <[email protected]> wrote: I keep getting these "Could not find file .../target/artifacts/addressing-1.6.3-SNAPSHOT.mar to copy" type errors on the 1.6.x branch. How do I work around this? Colm. On Tue, Jul 8, 2014 at 4:21 PM, <[email protected]> wrote: Hi Colm, What I did so far is to checkout Rampart (I have tried both trunk and 1.6 branches), increase the wss4j dependency to 1.6.5 and run "mvn clean package -Dtest=RampartTest". This fails on the "Testing WS-Sec: custom scenario 7" with the error I described. Switching the dependency back to 1.6.4 fixes this issue, but still there is one additional scenario (28) which is failing, however I presume it is not related with wss4j but probably with Axiom. I have checked out wss4j 1.6.x branch and build it locally, then switched Rampart to this version and re-executed the tests. The tests succeeded up until the point I switched to wss4j revision 1294114. With previous 1294094 revision, this scenario is working fine. I was thinking it might be related with changes of other dependencies, but I doubt this is the case, since this revision does not introduce dependency changes. I will continue with the investigation and let you know once I have more information. Thanks, Detelin On Tue, Jul 8, 2014 at 4:51 PM, Colm O hEigeartaigh <[email protected]> wrote: Are you sure that the commit you referenced above is causing the problem? Rampart trunk fails on that test for me with WSS4J 1.6.4. Rampart 1.6.x branch fails on something else... If you have time to look into it, you could try checking out that SNAPSHOT version of WSS4J (Before the commit) + check that it works + then apply each change and see what change causes the failure. Ultimately, it looks like Rampart might be at fault, as the response message is not composed properly Colm. On Tue, Jul 8, 2014 at 12:55 PM, <[email protected]> wrote: Hi everyone, Our team worked on new functionality that is to be released with upcoming wss4j 1.6.16 (WSS-500 & WSS-501). We have managed to integrate this functionality within Apache Rampart 1.6.2 and are willing to contribute the necessary pieces there as well. However, so far we have been using wss4j 1.6.4 + the corresponding patches and they seem to work fine with Rampart 1.6.2. Once I saw the vote for releasing wss4j 1.6.16, I decided to try to build Rampart 1.6.2 against it, just to make sure it can adopt this new version in near future. However, I stumbled upon a test failure in Rampart integration module, which I managed to track down to a specific commit in wss4j. The commit is quite old, it is released in wss4j 1.6.5 (latest Rampart uses 1.6.4). The change that causes trouble is the following: http://svn.apache.org/viewvc?view=revision&revision=1294114 Log message says "Only decrypt a Data Reference in the ReferenceListProcessor, if it hasn't already been decrypted by the EncryptedDataProcessor". The specific Rampart test that fails is "org.apache.rampart.RampartTest#testWithPolicy()" using the following security policy: http://svn.apache.org/repos/asf/axis/axis2/java/rampart/trunk/modules/rampart-integration/src/test/resources/rampart/policy/7.xml I'm attaching the SOAP request and response (request.xml and response.xml), the actual error message is on the client side, when processing the response from the service: java.lang.StringIndexOutOfBoundsException: String index out of range: 0 at java.lang.String.charAt(String.java:658) at org.apache.ws.security.WSDocInfo.getResult(WSDocInfo.java:225) at org.apache.ws.security.str.DerivedKeyTokenSTRParser.parseSecurityTokenReference(DerivedKeyTokenSTRParser.java:90) at org.apache.ws.security.processor.DerivedKeyTokenProcessor.handleToken(DerivedKeyTokenProcessor.java:53) at org.apache.ws.security.WSSecurityEngine.processSecurityHeader(WSSecurityEngine.java:398) at org.apache.ws.security.WSSecurityEngine.processSecurityHeader(WSSecurityEngine.java:304) at org.apache.ws.security.WSSecurityEngine.processSecurityHeader(WSSecurityEngine.java:249) at org.apache.rampart.RampartEngine.process(RampartEngine.java:147) The stack trace is generated using wss4j revision 1294114. It can be seen that the response contains invalid references (URI not correctly set): <wsse:SecurityTokenReference ... wsu:Id="STR-AA4ACE8415228CCC8E140481886870110"> <wsse:Reference URI="#" ValueType="http://docs.oasis-open.org/wss/oasis-wss-soap-message-security-1.1#EncryptedKey"; /> </wsse:SecurityTokenReference> I'm now trying to figure out what is the root cause of this and whether the problem is on the wss4j side or on Rampart's side, but I would be glad if anyone more experienced takes a look into this and provides some feedback. Thanks! Detelin --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected] -- Colm O hEigeartaigh Talend Community Coder http://coders.talend.com -- Colm O hEigeartaigh Talend Community Coder http://coders.talend.com -- Colm O hEigeartaigh Talend Community Coder http://coders.talend.com -- Colm O hEigeartaigh Talend Community Coder http://coders.talend.com
