[11u] RFR: 8171279: Support X25519 and X448 in TLS

[Lindenmaier, Goetz]

Hi Martin, 

Thanks for your hints and the patch!
I got the test running and can reproduce the issue. And yes, 
your change fixes the test.  I understand the code in 11
needs to check whether there is a cryptoProvider preselected
in SunJSSE.  

I think you don't need to rush, the rampdown period is 4 weeks,
and it is dedicated to fix such issues. So I think it's completely 
fine if this goes to jdk11u. But it is good the issue is
detected and understood before rampdown, thanks again 
for your help!

Best regards,
  Goetz.


> -----Original Message-----
> From: Martin Balao <mbalao at redhat.com>
> Sent: Tuesday, December 1, 2020 4:37 AM
> To: Lindenmaier, Goetz <goetz.lindenmaier at sap.com>; jdk-updates-
> dev at openjdk.java.net
> Cc: 'Andrew Haley' <aph at redhat.com>; Severin Gehwolf
> <sgehwolf at redhat.com>
> Subject: Re: [11u] RFR: 8171279: Support X25519 and X448 in TLS
> 
> Quick update:
> 
> I'm not done yet but here you have a preview of my changes:
> http://people.redhat.com/mbalaoal/openjdk/workspace/sunjsse_experimen
> tal_fips_support_and_dh_jdk11u/test_experimental_fips_with_dh.jdk11u.v1.
> patch
> 
> That fix looks enough for the reproducer to pass, but I still need to
> track a few things to make sure only SunJSSE's FIPS provider (if one) is
> used. When done, I'll create a new bug and send a Webrev for review. If
> we can't meet the ramp-down deadline, I'll request a critical fix for
> maintainers to decide.
> 
> Thanks,
> Martin.-
> 
> 
> On 11/30/20 11:43 AM, Martin Balao wrote:
> > Hi Goetz,
> >
> > Thanks for having a look at this.
> >
> > On 11/30/20 7:06 AM, Lindenmaier, Goetz wrote:
> >>
> >> I have been looking at your test, but it is not yet working
> >> on my machine. It skips the test after initializing.
> >>
> >
> > Yes, NSS tests require some help from the environment so they might be
> > skipped. A Linux-based environment with the NSS library located in the
> > (major distros) standard path should make it. Let me know if I can help
> > with that.
> >
> >> Before backing out, we should consider whether
> >> not having the new EC curves introduced by 8171279
> >> in 11.0.10 is acceptable. This is an extension that is
> >> documented as CSR and might be expected by people.
> >> It is in 11.0.10-oracle, too.
> >>
> >
> > I should be able to come up with a fix later today. The fix looks
> > straight forward -it's essentially replacing KeyAgreement::getInstance
> > calls with the previous calls-, but I want to make sure that everything
> > else is fine.
> >
> >> To me, it seems more relevant than the FIPS feature broken,
> >> which never has been an official feature as I understand,
> >> and of which it has been communicated (inofficially) that it
> >> does not work any more since 9.
> >
> > FIPS support in SunJSSE works up to 13, and our users rely on that. The
> > comment about stopping to work in 9 is wrong -I'll try to have it fixed,
> > as it has caused enough confusion-. There is a public API to initialize
> > FIPS in SunJSSE, which is through the java.security configuration file
> > (when you pass an argument to the SunJSSE security provider line).
> >
> > Thanks,
> > Martin.-
> >