Re: [PATCH v8 06/15] x86: Add early SHA support for Secure Launch early measurements

[Andy Lutomirski]

On Fri, Feb 23, 2024, at 10:30 AM, Eric Biggers wrote:
> On Fri, Feb 23, 2024 at 06:20:27PM +0000, Andrew Cooper wrote:
>> On 23/02/2024 5:54 pm, Eric Biggers wrote:
>> > On Fri, Feb 23, 2024 at 04:42:11PM +0000, Andrew Cooper wrote:
>> >> Yes, and I agree.  We're not looking to try and force this in with
>> >> underhand tactics.
>> >>
>> >> But a blind "nack to any SHA-1" is similarly damaging in the opposite
>> >> direction.
>> >>
>> > Well, reviewers have said they'd prefer that SHA-1 not be included and given
>> > some thoughtful reasons for that.  But also they've given suggestions on how to
>> > make the SHA-1 support more palatable, such as splitting it into a separate
>> > patch and giving it a proper justification.
>> >
>> > All suggestions have been ignored.
>> 
>> The public record demonstrates otherwise.
>> 
>> But are you saying that you'd be happy if the commit message read
>> something more like:
>> 
>> ---8<---
>> For better or worse, Secure Launch needs SHA-1 and SHA-256.
>> 
>> The choice of hashes used lie with the platform firmware, not with
>> software, and is often outside of the users control.
>> 
>> Even if we'd prefer to use SHA-256-only, if firmware elected to start us
>> with the SHA-1 and SHA-256 backs active, we still need SHA-1 to parse
>> the TPM event log thus far, and deliberately cap the SHA-1 PCRs in order
>> to safely use SHA-256 for everything else.
>> ---
>
> Please take some time to read through the comments that reviewers have left on
> previous versions of the patchset.

So I went and read through the old comments, and I'm lost.  In brief summary:

If the hardware+firmware only supports SHA-1, then some reviewers would prefer Linux not to support DRTM.  I personally think this is a bit silly, but it's not entirely unreasonable.  Maybe it should be a config option?

If the hardware+firmware does support SHA-256, then it sounds (to me, reading this -- I haven't dug into the right spec pages) that, for optimal security, something still needs to effectively turn SHA-1 *off* at runtime by capping the event log properly.  And that requires computing a SHA-1 hash.  And, to be clear, (a) this is only on systems that already support SHA-256 and that we should support and (b) *not* doing so leaves us potentially more vulnerable to SHA-1 attacks than doing so.  And no SHA-256-supporting tooling will actually be compromised by a SHA-1 compromise if we cap the event log.

So is there a way forward?  Just saying "read through the comments" seems like a dead end.

Thanks,
Andy