From: David Howells Subject: Re: [RFC][PATCH 00/16] Crypto keys and module signing [ver #2] Date: Mon, 05 Dec 2011 11:43:39 +0000 Message-ID: <776.1323085419@redhat.com> References: <4EDCABBD.9020401@intel.com> <20111129234258.13625.21153.stgit@warthog.procyon.org.uk> Cc: dhowells@redhat.com, keyrings@linux-nfs.org, linux-crypto@vger.kernel.org, linux-security-module@vger.kernel.org, linux-kernel@vger.kernel.org, dmitry.kasatkin@intel.com, zohar@linux.vnet.ibm.com, arjan.van.de.ven@intel.com, alan.cox@intel.com To: "H. Peter Anvin" Return-path: In-Reply-To: <4EDCABBD.9020401@intel.com> Sender: linux-security-module-owner@vger.kernel.org List-Id: linux-crypto.vger.kernel.org H. Peter Anvin wrote: > Do we really need the complexity of a full OpenPGP parser? Parsers are > notorious security problems. Actually, I don't have a full PGP parser. I only handle the minimum I need. It can parse the packet stream, public key packets and signature packets. That's it. I cannot guarantee that GPG is always going to put packets in the stream in the same order, and dealing with the possible variation should be simple enough. Note that it might be possible to share the parser with other things like eCryptFS. > Furthermore, using DSA in anything but a hard legacy application is not > something you want to encourage, so why support DSA? Because DSA is what Red Hat currently uses in its module signing. So the first thing to do was to get that working again and then work on getting RSA working. If I just jumped straight into the RSA parser, I would have a lot more stuff that might be wrong; doing DSA first at least validated the common stuff. However, we don't have to include the DSA stuff in the kernel; I can always discard that patch from the upstream-aimed patchset. David