[
https://issues.apache.org/jira/browse/QPID-4636?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13601802#comment-13601802
]
Robbie Gemmell commented on QPID-4636:
--------------------------------------
Hi Michal,
I took a look and the patches, and ran the default profile of the test suite
against them while I was doing so. Unfortunately, a large number of the
existing system tests which use SSL failed or threw errors. I didn't actually
spot anything obvious wrong with the patches while I was looking them over, but
its been a looong week so far (which is why im doing this after 11pm :)) so I
might have missed something. It's also possible the tests and/or previous
implementation themselves are at issue, but eihter way it appears that it needs
a further look into.
Were you able to run the tests successfully? You can run them all by using 'ant
test' in the java folder, and optionally specifying a particular test profile
to run and a particular test name( or pattern of names/packages) to run. The
default test profile is java-mms.0-10, which uses AMQP 0-10 and the Memory
message store. You can see a full list of the profiles in the
java/test-profiles folder. Typically its a good idea to run an AMQP 0-10 and
0-9-1 profile since their behaviour can differ fairly substantially, though I
wouldnt really expect them to in the case of SSL. If you were to use any of the
bdb profiles (which use the persistent bdb message store instead of the memory
message store) you would also have to add -Doptional=true to allow the optional
bdbstore module (and any other optional modules) to build. E.g:
ant build test [[[-Dprofile=java-mms.0-9-1] -Dtest=SSLTest] -Doptional=true]
Although I didn't spot anything with the functionality itself, I do think the
tests could use a bit more work to ensure robustness. In
testPeerStoreClientAuthentication() and
testMultipleStoreClientAuthentication(), while posibly unlikely it seems as if
the test could actually skip the important calls on the trust managers and
still pass (due to the for and if statements, e.g empty list or unexpected key
manager types) and we should possibly verify that hasnt occurred. Whilst
everything looks ok just now, I think it would still be useful to add some more
tests to ensure the functionality isn't unwittingly broker in future. For
example, checking that the QpidPeersOnlyTrustManager never returns trusted CAs,
and the QpidMultipleTrustManager does indeed throw a CertificateException if
presented with a cert that neither the QpidPeersOnlyTrustManager or a regular
keystore TrustManager trusted. It doesn't seem like SSLContextFactory is
actually used in the unit tests, so it might be best to move the tests to their
own file in the package. Beyond that, adding even a single system test (e.g to
ExternalAuthenticationTest) that shows the peerStore with the broker in
operation still seems wise, since the unit tests dont strictly guarantee the
peer store is actually hooked up properly within the broker.
Robbie
> SSL Client Authentication in Java broker
> ----------------------------------------
>
> Key: QPID-4636
> URL: https://issues.apache.org/jira/browse/QPID-4636
> Project: Qpid
> Issue Type: New Feature
> Components: Java Broker, Java Common
> Affects Versions: 0.21
> Reporter: Michal Zerola
> Assignee: Robbie Gemmell
> Priority: Minor
> Attachments: java_broker_peerstore.jks, ssl_new_trustmanager.patch,
> ssl_new_trustmanager_test.patch
>
>
> *Motivation:*
> The C++ broker implementation is based on the NSS library from Mozilla. The
> user creates a certificate database and configures the broker to load the
> database at start-up. The NSS certificate database can store the private
> keys used by the broker as well as the public keys related to the connecting
> clients. The public keys can be divided into several groups - the keys of
> trusted CAs and the keys of trusted peers. The difference between the trusted
> CA and trusted peer is that the trusted CA allows logging in even to clients
> who have a certificate signed by the CA, while the peers allow logging in
> only to clients who have the certificate exactly matching the certificate
> loaded in the certificate database.
> The SSL Client Authentication in the Java broker is based on the Java JSSE
> implementation. The certificates are stored in the JKS format. The JKS format
> doesn't distinguish between trusted peers and trusted CA. Therefore all
> public keys behave as trusted CAs. As a result, the current implementation
> cannot be used together with self signed certificates. As we found out, there
> seems to be no support for the trusted peers in the JKS truststores.
> *Proposed solution:*
> The current configuration for the SSL Client Authentication supports only one
> truststore . We can add a second configuration entry which would allow to
> specify "peerstore" . When creating the SSL context, the existing truststore
> would be handled as it is handled today. If the "peerstore" is specified, the
> new TrustManager will be added to the SSL context. The custom TrustManager
> will use the peerstore to verify the peers only as peers. The client will
> pass the authentication if it is authenticated either with the original
> Trustmanager against the keystore or by the custom trust manager against the
> peer store.
> *Configuration and patch explanation:*
> The current configuration model for the broker (trunk) is based on JSON. We
> have added two optional attributes (peerStorePath, peerStorePassword):
> {quote}
> …
> "keyStorePath": "...",
> "keyStorePassword": "123456",
> "trustStorePath": "...",
> "trustStorePassword": "123456",
> {color:red}
> "peerStorePath": "...",
> "peerStorePassword": "123456",
> {color}
> …
> {quote}
> Internally, the broker is prepared to handle multiple truststores, since it
> can store them in the collection. If the above new attributes were specified,
> the additional truststore is added into the collection (BrokerAdapter.java).
> A new truststore will have optional flag TrustStore.PEERS_ONLY set to True.
> The SSLContextFactory was extended for collection configuration. The Broker
> creates the SSLContext using the collection of truststores (either only
> single truststore or with a new peerstore). The SSLFactory parses the
> collection and depending on the TrustStore.PEERS_ONLY flag creates either
> regular JSSE trustmanager or uses a newly introduced one
> QpidPeersOnlyTrustManager.
> The new QpidPeersOnlyTrustManager works as a wrapper around standard JSSE
> trustmanager. When client connects, the check is delegated to the underlying
> JSSE verification and if it succeeds, the additional check is done, whether
> the peer’s certificate (in the chain index 0) is present in the keystore
> file. Only then the client is authenticated, otherwise the
> CertificateException is thrown.
> Since SSLContext.init method from the array of trustmanagers uses only the
> first one which is an instance of the X509TrustManager class, we have created
> also the extension. Otherwise, it would not be possible to use simultaneously
> trustmanager (JSSE implementation) and peermanager (our new implementation)
> because both implements X509TrustManager and only the first from the array
> would be considered. Therefore, we have introduced the
> QpidMultipleTrustManager which is doing nothing else but delegating the check
> to its underlying X509TrustManagers and only if all fails, the check itself
> fails. If some underlying manager succeeds, the check itself succeeds as
> well. This QpidMultipleTrustManager is loaded with truststore and peerstore
> manager and added into the array which is further passed to the
> SSLContext.init method.
> The implementation attached in the patch seems to be working fine and adds
> the above requirements for peers only truststore. It is also backwards
> compatible- anyone without interest for peerstores will not see any change.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@qpid.apache.org
For additional commands, e-mail: dev-h...@qpid.apache.org
