On 06/05/2018 11:43 AM, Daniel P. Berrangé wrote: > On Tue, Jun 05, 2018 at 10:45:55AM +0200, Michal Privoznik wrote: >> We are building with GnuTLS everywhere because GnuTLS is widely >> available. In addition after recent patches Libvirt relies on >> GnuTLS' PRNG. > > This second sentance isn't true AFAIK - we still have fallback > to /dev/urandom - GNUTLS is merely the first choice. Okay. But after Peter's patches we do rely on GnuTLS more than ever ;-) I'll reword and resend though. Michal > > None the less I think its desirable to make GNUTLS mandatory > since it is on all the platforms we care about and I prefer > that we can assume a good crypto impl all the time. This mostly > frees us from worrying about fallback impls which have higher > risk of security problems. Unfortunately not. Both suid and nss libs build with virhash.c which requires virRandom*(). But this is a bogus dependency and hash tables are not really used (at least in NSS module, did not bother to check for suid lib). So we need a stub for virRandom*(). Michal -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
