Well you helped me getting things to work, so if you need someone to test these things with the Belgian EID, let me know. On 11/08/2013 12:16 AM, David Woodhouse wrote: > On Thu, 2013-11-07 at 17:08 +0100, Nikos Mavrogiannopoulos wrote: >> On Tue, Nov 5, 2013 at 4:14 PM, David Woodhouse <dwmw2 at infradead.org> wrote: >> >>> If either of them are responsible for signing your personal cert, then >>> OpenConnect will include them in its SSL negotiation, and that can often >>> 'help' the server to realise that it actually *does* trust the cert in >>> question. >>> If that's the issue, then perhaps OpenConnect needs to be taught to go >>> looking for these 'supporting' certs in the PKCS#11 store, as well as >>> the --cafile. But then again, perhaps GnuTLS ought to do that for >>> itself. >>> Nikos? >> Indeed, that's a nice feature and not too difficult to be implemented >> as PKCS #11 allows searching stored certificates using a DN. It is on >> my todo-list for quite some time but never found the time for that. >> Patches are (of course) more than welcome! > Ick, please not by name! It should be done by key ID. > > If we have to do it by DN, is there at least a way to say "that wasn't > the one I wanted. Please give me the *next* cert with the same name"? > > This code would slot in at about line 1642 of OpenConnect's gnutls.c, if > gnutls_certificate_get_issuer() fails to find an appropriate issuer in > the system CA file. (Hey, if we want to make > gnutls_certificate_get_issuer() just *work* in this situation and find > it in the PKCS#11 token, that's great too!) > > For the benefit of the peanut gallery and the the archives, this *is* > the solution to Christof's problem. The supporting CA 'Citizen CA' is > also present on the token, and the client needs to explicitly include > that in the SSL negotiations on the wire because the server can't find > it for itself. > > > > _______________________________________________ > openconnect-devel mailing list > openconnect-devel at lists.infradead.org > http://lists.infradead.org/mailman/listinfo/openconnect-devel