On 7/13/21 11:25 AM, Peter Krempa wrote: > On Tue, Jul 13, 2021 at 08:57:30 +0200, Michal Privoznik wrote: >> In not so distant past (v6.5.0~3) I've updated the private key we >> use for virnettls* tests. Back then I was driven by Fedora 33 >> change which deprecated RSA-1024 which we used back then. I >> generated an EC-384 key which was fine as it was considered >> strong enough until RHEL-9 came along. RHEL-9 no longer considers >> any of EC keys strong enough (for key exchange) and thus we're >> back to RSA, but this time with 2048 bits. Generated by this cmd >> line: > > I'd go for 4096 bits to stay ahead a bit. > >> >> openssl genpkey -algorithm RSA -out key.pem -pkeyopt rsa_keygen_bits:2048 >> >> Signed-off-by: Michal Privoznik <mprivozn@xxxxxxxxxx> >> --- > > My quick google search yielded just some JDK changes for improving the > implementation of EC algorithms: > > https://bugs.openjdk.java.net/browse/JDK-8208698 > > but nothing that would state it's no longer secure or anything. > > Either way. > > Reviewed-by: Peter Krempa <pkrempa@xxxxxxxxxx> > Alright, so after more debugging this turned out to be a bug in crypto-policies package in RHEL-9. It's fixed by the following commit: https://gitlab.com/redhat-crypto/fedora-crypto-policies/-/commit/a5e64bb9a4afcd67965218ba41e28a6839aa9a12 And I can confirm that with that commit the virnettlssessiontest passes again. Thus I think this patch can be discarded. Michal
