On Mon, Dec 09, 2019 at 12:31:53PM -0800, James Bottomley wrote: > On Mon, 2019-12-09 at 21:47 +0200, Jarkko Sakkinen wrote: > > On Mon, Dec 02, 2019 at 10:55:32AM -0800, James Bottomley wrote: > > > blob but it looks like we need to fix the API. I suppose the good > > > news is given this failure that we have the opportunity to rewrite > > > the API since no-one else can have used it for anything because of > > > this. The > > > > I did successfully run this test when I wrote it 5 years ago: > > > > https://github.com/jsakkine-intel/tpm2-scripts/blob/master/keyctl-smo > > ke.sh > > > > Given that there is API a way must be found that backwards > > compatibility > > is not broken. New format is fine but it must co-exist. > > The old API is unsupportable in the combination of policy + auth as I > already explained. The kernel doesn't have access to the nonces to > generate the HMAC because the session was created by the user and the > API has no way to pass them in (plus passing them in would be a huge > security failure if we tried). Given that Shirley appears to be the > first person ever to try this, I don't think the old API has grown any > policy users so its safe to remove it. If we get a complaint, we can > discuss adding it back. It works within limits so it can be definitely be maintained for backwards compatibility. Also, you are making a claim of the users that we cannot verify. Finally, the new feature neither handles sessions. You claim that it could be added later. I have to deny that because until session handling is there we have no ways to be sure about that. I see your point but this needs more consideration. It does not make sense to rush. /Jarkko
