RFR: 8279066: Still see private key entries without certificates in a PKCS12 keystore

[Michael StJohns]

On 12/21/2021 1:26 PM, Sean Mullan wrote:
> On Tue, 21 Dec 2021 16:31:57 GMT, Weijun Wang <weijun at openjdk.org> wrote:
>
>> Before password-less PKCS12 keystores are supported, certificates in a PKCS12 file are always encrypted. Therefore if one loads the keystore with a null pass, it contains `PrivateKeyEntry`s without certificates. This has always been awkward (and most likely useless) so when JDK-8076190 introduced the password-less feature I also added a line to remove such an entry.
>>
>> https://github.com/openjdk/jdk/blob/a729a70c0119ed071ff490b0dfd4e3e2cb1a5ae4/src/java.base/share/classes/sun/security/pkcs12/PKCS12KeyStore.java#L2272
>>
>> Unfortunately, the line is not coded correctly, it should have been `remove(key)` but here it's `remove(value)`.
>>
>> This code change correctly removes the entry.
>>
>> That said, this behavior, although weird, has been there from the beginning since PKCS12 keystore was introduced. If you can find out a usage of a private key entry without any certificate and think it's worth kept that way, I can simply remove the `remove` call and leave the entry there.
> I still think it's useful even if I can't see the certificate chain. I'd rather see the entry if it actually exists in the keystore and I think removing it is odd because it still exists in the keystore. Also, sometimes I use keytool without a storepass just to see what is in it, and then if I see the certificates are not showing up, I can try again with the password.
>
> -------------
>
> PR: https://git.openjdk.java.net/jdk/pull/6910

I got curious and took a look at P11KeyStore.java - 
https://github.com/openjdk/jdk/blob/master/src/jdk.crypto.cryptoki/share/classes/sun/security/pkcs11/P11KeyStore.java 
- and it appears to do exactly the opposite of what you've got here.  
E.g. if there's no certificate that matches up with the private key, 
then it returns a null - that's in engineGetEntry() around line 963.

Speaking personally, I've always found it a bit annoying that a 
KeyStore.PrivateKeyEntry required a certificate(chain).  It actually 
made dealing with PKCS11 painful as the Java conventions didn't always 
conform to the actual structure of the PKCS11 contents.

Not suggesting that a change necessarily needs to be made, but perhaps 
you might want to have a consistent behavior between the two?


Mike