Re: [PATCH v12 1/3] x86/tdx: Add TDX Guest attestation interface driver

From: Greg Kroah-Hartman
Date: Fri Sep 09 2022 - 01:26:27 EST


On Thu, Sep 08, 2022 at 01:36:00PM -0700, Dave Hansen wrote:
> On 9/8/22 12:07, Sathyanarayanan Kuppuswamy wrote:
> > On 9/7/22 10:31 PM, Greg Kroah-Hartman wrote:
> >> On Wed, Sep 07, 2022 at 05:27:20PM -0700, Kuppuswamy Sathyanarayanan wrote:
> >>> + /*
> >>> + * Per TDX Module 1.0 specification, section titled
> >>> + * "TDG.MR.REPORT", REPORTDATA length is fixed as
> >>> + * TDX_REPORTDATA_LEN, TDREPORT length is fixed as
> >>> + * TDX_REPORT_LEN, and TDREPORT subtype is fixed as
> >>> + * 0. Also check for valid user pointers.
> >>> + */
> >>> + if (!req.reportdata || !req.tdreport || req.subtype ||
> >>> + req.rpd_len != TDX_REPORTDATA_LEN ||
> >>> + req.tdr_len != TDX_REPORT_LEN)
> >>> + return -EINVAL;
> >> You never verify that your reserved[7] fields are actually set to 0,
> >> which means you can never use them in the future :(
> > Currently, we don't use those fields in our code. Why do we have to
> > make sure they are set to zero?
>
> Yes.
>
> > Can't we add checks when we really use them in future?
>
> No.
>
> This has been a hard learned lesson both by people writing software and
> designing hardware interfaces: if you _let_ folks pass garbage you have
> to _keep_ letting them pass garbage forever. It becomes part of the ABI.
>
> I'm sorry you missed the memo on this one. But, this is one million
> percent a best practice across the industry. Please do it.

And it's documented in the Documentation/ directory as a requirement to
do as well, the memo shouldn't have been missed :(