Code review request, CR 7180038 regression test failure, SSLEngineBadBufferArrayAccess.java (original) (raw)

Weijun Wang weijun.wang at oracle.com
Mon Jul 2 08:35:06 UTC 2012

I take a look at the test output. When the last handshake starts:

================ server unwrap: OK/NEED_TASK, 230/0 bytes running delegated task... new HandshakeStatus: NEED_WRAP

server wrap: OK/NEED_WRAP, 0/86 bytes

Here the first wrap only generates 86 bytes, I guess that's the ServerHello message? It keeps the state at NEED_WRAP but then never really generates the Certificate message. What might be the problem?

Thanks Max

On 07/02/2012 10:39 AM, Xuelei Fan wrote:

Hi Weijun,

Would you please review the test update for CR 7180038? http://cr.openjdk.java.net./~xuelei/7180038/webrev.00/ We cannot reproduce the issue. However, from the test log, there is two possible issues exposed by this CR. 1. the improper test case senarios of un/wrap() In the test case, the scenarios is unwrap()->wrap()->serverOut.remaining()->"serverIn.remaining() != clientMsg.length". After the wrap(), the next operation may need to be unwrap() to get more incoming data before comparing serverIn buffer with the expected client message. This fix is trying to do the comparing after the engine has closed. 2. From the log, the engine status and handshaking status move from CLOSED/NOTHANDSHAKING to OK/FINISHED. FINISHED means the TLS handshaking just finished. As the handshaking should have completed for a while, it does not sound like a correct status change. However, I did not find why this happens. Need more info. So I added a line of log (suggested by Brad Wetmore) to collect the next failure: IF THIS FAILS, PLEASE REPORT THIS TO THE SECURITY TEAM. WE HAVE BEEN UNABLE TO RELIABLY DUPLICATE.

Thanks, Xuelei

More information about the security-dev mailing list