On Wed, Jan 11, 2017 at 2:03 PM, Shrikant Bobade <bobadeshrikant@gmail.com> wrote:
>
> Hi Daniel, Good Morning,
>
>
> On Wed, Jan 11, 2017 at 1:46 PM, Daniel Wagner <wagi@monom.org> wrote:
> >
> > Good Morning Shrikant,
> >
> > On 01/11/2017 07:11 AM, Shrikant Bobade wrote:
> > >> if you start ConnMan with 'GNUTLS_NO_EXPLICIT_INIT=1 ./connmand -n -d'
> > > it should work.
> > >>
> > >> Now I cross my fingers :D
> > >
> > > yes, with zero entropy and GNUTLS_NO_EXPLICIT_INIT=1
> >
> > Yay!
> >
> > [...]
> >
> > > got successful ip assignment.
> > >
> > > do I need to post complete log some where to share.. or its fine..
> >
> > No there is no need for it anymore. We know now what it happening.
> >
> > > as we are getting now expected ip. please advice..
> >
> > So this is a configuration issue. I think the best thing we can do here
> > is update your README file and let people know what to look out for.
>
> sounds fine..
> >
> > Additionally we should also change urandom users in our code base to
> > random. Will send patches for both.
> >
> > Obviously, with GNUTLS_NO_EXPLICIT_INIT=1 we just do lazy initialization
> > and we still might hang later. Not really sure what the best approach
> > is. It depends on the use cases etc. On my desktop system I never have
> > seen this problem. The main reason is, that my system does on default
> > lazy initialization and I don't have a portal in front of my.
>
> ok, I will run some regressions to confirm it.
> >
> > > So will move ahead with explicit GNUTLS_NO_EXPLICIT_INIT=1 & remove the
> > > rngd dependency(added earlier for ~3k entropy available), will it be  fine ?
> >
> > If you don't need wipsr really early on, you should be fine without
> > rngd. But testing might reveal that I am wrong. Hard to say :)
>
> yes, agree..
>
> I am running the regressions with GNUTLS_NO_EXPLICIT_INIT=1,
> will revert back with the observations.
>

observed during regression, there was no further hang using GNUTLS_NO_EXPLICIT_INIT=1,
using it resolves the issue.

thanks
shrikant

>
> >
> > cheers,
> > daniel
>
> Thanks for your continuous help.
>
> Thanks
> Shrikant
>