Re: [PATCH] x86/efi-bgrt: Switch pr_err() to pr_debug() for invalid BGRT

[josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org]

On Mon, Jun 29, 2015 at 04:17:24PM +0200, Borislav Petkov wrote:
> On Mon, Jun 29, 2015 at 07:00:22AM -0700, Josh Triplett wrote:
> > Definitely not FW_BUG.  The field is reserved *now*; it would be
> > legitimate for a new version of the BGRT spec to define one of those
> > bits for something else.
> 
> Which would mean that booting old kernels on new FW which defines those
> reserved bits would cause that warning to fire erroneously.

Not erroneously; those bits could potentially indicate some status
condition we don't know about, so we need to assume we can't handle the
table if a bit we don't understand is set.

(Specs like this should really do what ext4 does, and define whether a
given set of currently undefined bits are optional or mandatory; as in,
if you don't understand them, can you proceed or should you stop?)

> So then we probably don't need it at all or we need to check implemented
> BGRT version of the FW running to know which bits are defined by the
> spec and which are reserved...
> 
> Also, does the spec really say that reserved bits must be zero? Or it
> doesn't specify their value?

The spec says those bits of the status field are reserved and must be
zero.

- Josh Triplett