X509Certificate array used for trust managers is not a chain #4973

harbulot · 2016-03-13T18:40:43Z

In multiple parts of the code related to SSLContext, the X509Certificate[] array is called trustCertChain and is referred to as a "chain".

However, trust anchors are generally not a chain at all (unlike keyCertChain, which is a chain). There can indeed be multiple completely unrelated certificates in that collection of trusted certificates.

This simple PR comprises two commits, addressing two sub-problems on this topic:

The code in SSLContext was using the Subject DN (Subject Principal) of each certificate in this array as the alias to build the trust store. However, it is perfectly possible to have multiple distinct certificates with the same Subject DN in that array. Using the Subject DN effectively discards some of those certificates with overlapping names.
- The fix here is to use the SHA-1 fingerprint of the certificate (nothing to do with the signature). This identifier is used by a number of other tools, most notably, by the Windows API.
- If this was to become a performance problem, since the identifiers themselves don't matter, another way could be to use an incremented identifier.
The second commit is more cosmetic: it renames all occurrences of trustCertChain into trustCertCollection, and modifies the Javadoc accordingly. This is mostly to prevent confusion based on the name.

normanmaurer · 2016-03-13T19:24:52Z

Could you add a unit test ?

Also please sign our icla and adjust the commit message to match our template:
http://netty.io/s /icla
http://netty.io/wiki/writing-a-commit-message.html

netkins · 2016-03-13T20:07:39Z

handler/src/main/java/io/netty/handler/ssl/OpenSslClientContext.java

@@ -165,7 +166,7 @@ public OpenSslClientContext(File certChainFile, TrustManagerFactory trustManager
     * @deprecated use {@link SslContextBuilder}
     */
    @Deprecated
-    public OpenSslClientContext(File trustCertChainFile, TrustManagerFactory trustManagerFactory,
+    public OpenSslClientContext(File trustCertCollectionFile, TrustManagerFactory trustManagerFactory,


Constructor has 11 parameters, which is greater than 7 authorized.

harbulot · 2016-03-14T11:33:07Z

@normanmaurer

Sure, I can try to write unit tests (although there weren't any before anyway).

Could you tell me what's wrong with the commit messages I've submitted? I thought I'd followed the required format.

normanmaurer · 2016-03-14T13:20:29Z

@harbulot actually the commit message is correct, I somehow missed this.

It is correct that there were no unit tests for this before and this is why it had a bug ;) So I would say please submit a unit-test with this as well so we can be sure it is fixed and stay fixed.

Scottmitch · 2016-03-14T18:40:23Z

@harbulot - Thanks for the patch! Unit test would be much appreciated :)

normanmaurer · 2016-03-15T13:26:58Z

@harbulot let me know once you added the test-case an I will review :) Thanks!

harbulot · 2016-03-15T14:07:15Z

@normanmaurer Sure, I'll try to do it within the next few days, as I'll need to spend a bit more time on it to have good test cases (and create test certs).

normanmaurer · 2016-03-15T14:18:52Z

@harbulot awesome!

normanmaurer · 2016-03-22T08:54:52Z

@harbulot any news ?

harbulot · 2016-03-22T15:56:56Z

@normanmaurer Just updated (and rebased against the latest 4.1 branch):

I've actually changed the way to name each cert. Instead of computing the SHA-1 fingerprint for each certificate to use as alias name, I'm just using a counter (and converting it into a String). Since the names are never used to get access to the certificate later, there doesn't seem to be any point spending time calculating a digest.
I've added a test case, with two CAs with the same name and one CA with a different name. Each of these three CAs has a test server certificate.

normanmaurer · 2016-03-22T16:01:32Z

handler/src/test/java/io/netty/handler/ssl/SslContextTrustManagerTest.java

+        TrustManagerFactory tmf = SslContext.buildTrustManagerFactory(
+                certCollection, null);
+
+        X509TrustManager x509Tm = null;


remove redundant null assignment.

This one is not redundant (it won't compile without), unless we return the value directly in the loop (instead of breaking) and return null after the loop. Would that be the preferred style?

just return directly. and throw an exception if the loop not return early.

normanmaurer · 2016-03-22T16:44:13Z

@harbulot after addressing my comment can you also please squash your commits into one ?

Also please sign our ICLA: http://netty.io/s/icla

netkins · 2016-03-22T16:45:14Z

handler/src/main/java/io/netty/handler/ssl/JdkSslClientContext.java

@@ -239,13 +241,13 @@ public JdkSslClientContext(File trustCertChainFile, TrustManagerFactory trustMan
     * @deprecated use {@link SslContextBuilder}
     */
    @Deprecated
-    public JdkSslClientContext(File trustCertChainFile, TrustManagerFactory trustManagerFactory,
+    public JdkSslClientContext(File trustCertCollectionFile, TrustManagerFactory trustManagerFactory,


Constructor has 11 parameters, which is greater than 7 authorized.

renamed trustCertChain into trustCertCollection. Motivation: SSLContext.buildTrustManagerFactory(...) builds a KeyStore to initialize the TrustManagerFactory from an array of X509Certificates, assuming that array is a chain and that each certificate will have a unique Subject Distinguised Name. However, the collection of certificates used as trust anchors is generally not a chain (it is an unordered collection), and it is legitimate for it to contain multiple certificates with the same Subject DN. The existing code uses the Subject DN as the alias name when filling in the `KeyStore`, thereby overwriting other certificates with the same Subject DN in this collection, so some certificates may be discarded. In addition, the code related to building trust managers can take an array of X509Certificate instances to use as trust anchors. The variable name is usually trustCertChain, and the documentation refers to them as a "chain". However, while it makes sense to talk about a "chain" from a keymanager point of view, these certificates are just an unordered collection in a trust manager. (There is no chaining requirement, having the Subject DN matching its predecessor's Issuer DN.) This can create confusion to for users not used with PKI concepts. Modifications: SSLContext.buildTrustManagerFactory(...) now uses a distinct alias for each array (simply using a counter, since this name is never used for reference later). This patch also includes a unit test with CA certificates using the same Subject DN. Also renamed trustCertChain into trustCertCollection, and changed the references to "chain" in the Javadoc. Result: Each loaded certificate now has a unique identifier when loaded, so it is now possible to use multiple certificates with the same Subject DN as trust anchors. Hopefully, renaming the parameter should also reduce confusion around PKI concepts.

netkins · 2016-03-22T19:31:33Z

handler/src/main/java/io/netty/handler/ssl/JdkSslClientContext.java

@@ -239,13 +241,13 @@ public JdkSslClientContext(File trustCertChainFile, TrustManagerFactory trustMan
     * @deprecated use {@link SslContextBuilder}
     */
    @Deprecated
-    public JdkSslClientContext(File trustCertChainFile, TrustManagerFactory trustManagerFactory,
+    public JdkSslClientContext(File trustCertCollectionFile, TrustManagerFactory trustManagerFactory,


Constructor has 11 parameters, which is greater than 7 authorized.

normanmaurer · 2016-03-22T20:35:09Z

Cherry-picked into 4.0 (f5fab38) and 4.1 (9ebb4b7).

@harbulot thanks a lot!

normanmaurer self-assigned this Mar 13, 2016

netkins reviewed Mar 13, 2016
View reviewed changes

Scottmitch added the defect label Mar 14, 2016

Scottmitch added this to the 4.0.35.Final milestone Mar 14, 2016

normanmaurer modified the milestones: 4.0.35.Final, 4.0.36.Final Mar 21, 2016

harbulot force-pushed the ssl_trust_anchors branch 3 times, most recently from 823f41b to 5f85917 Compare March 22, 2016 15:53

normanmaurer reviewed Mar 22, 2016
View reviewed changes

netkins reviewed Mar 22, 2016
View reviewed changes

harbulot force-pushed the ssl_trust_anchors branch from 5f85917 to 9863fba Compare March 22, 2016 18:12

harbulot force-pushed the ssl_trust_anchors branch from 9863fba to 502644f Compare March 22, 2016 18:16

netkins reviewed Mar 22, 2016
View reviewed changes

normanmaurer closed this Mar 22, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

X509Certificate array used for trust managers is not a chain #4973

X509Certificate array used for trust managers is not a chain #4973

harbulot commented Mar 13, 2016

normanmaurer commented Mar 13, 2016

netkins Mar 13, 2016

harbulot commented Mar 14, 2016

normanmaurer commented Mar 14, 2016

Scottmitch commented Mar 14, 2016

normanmaurer commented Mar 15, 2016

harbulot commented Mar 15, 2016

normanmaurer commented Mar 15, 2016

normanmaurer commented Mar 22, 2016

harbulot commented Mar 22, 2016

normanmaurer Mar 22, 2016

harbulot Mar 22, 2016

normanmaurer Mar 22, 2016

normanmaurer commented Mar 22, 2016

netkins Mar 22, 2016

netkins Mar 22, 2016

normanmaurer commented Mar 22, 2016

X509Certificate array used for trust managers is not a chain #4973

X509Certificate array used for trust managers is not a chain #4973

Conversation

harbulot commented Mar 13, 2016

normanmaurer commented Mar 13, 2016

netkins Mar 13, 2016

Choose a reason for hiding this comment

harbulot commented Mar 14, 2016

normanmaurer commented Mar 14, 2016

Scottmitch commented Mar 14, 2016

normanmaurer commented Mar 15, 2016

harbulot commented Mar 15, 2016

normanmaurer commented Mar 15, 2016

normanmaurer commented Mar 22, 2016

harbulot commented Mar 22, 2016

normanmaurer Mar 22, 2016

Choose a reason for hiding this comment

harbulot Mar 22, 2016

Choose a reason for hiding this comment

normanmaurer Mar 22, 2016

Choose a reason for hiding this comment

normanmaurer commented Mar 22, 2016

netkins Mar 22, 2016

Choose a reason for hiding this comment

netkins Mar 22, 2016

Choose a reason for hiding this comment

normanmaurer commented Mar 22, 2016