Re: [dmidecode] [Patch v3] firmware: dmi-sysfs: add SMBIOS entry point area raw attribute

[Ard Biesheuvel]

On 8 March 2015 at 18:11, Jean Delvare <jdelvare@xxxxxxx> wrote:
> On Sun, 8 Mar 2015 14:53:04 +0100, Ard Biesheuvel wrote:
>> On 8 March 2015 at 12:31, Jean Delvare <jdelvare@xxxxxxx> wrote:
>> > On Sat, 07 Mar 2015 22:53:32 +0200, Ivan.khoronzhuk wrote:
>> >> The specification doesn't oblige firmware to provide two entry points.
>> >> An implementation may provide either the 32-bit entry point or the 64-bit
>> >> entry point, or both. For compatibility with existing SMBIOS parsers, an
>> >> implementation should provide the 32-bit entry point, but it's not required.
>> >
>> > I expect most implementations will do, as it's trivial to implement.
>>
>> Not quite. First of all, some 64-bit ARM systems do not have any
>> system RAM below 4 GB, so there is not way they can implement the
>> 32-bit entry point.
>
> I didn't know that, thanks for the notice. No big deal anyway, these
> systems did not support SMBIOS before version 3.0 so there cannot be
> any regression on these systems.
>
>> Also, the 64-bit entry point does not limit the
>> structure size or the entire table to 64 KB like the 32-bit one does,
>> so it may be necessary to create a whole separate table with a subset
>> of the contents of the real table to stay within limits for the 32-bit
>> entry point.
>
> I doubt this is an issue in practice. I have been around for quite some
> time now and the largest table I've ever seen was 9043 byte long, which
> is nowhere close to the limit.
>
>> And the 32-bit entry point could well be 3.0 anyway, if
>> it uses any of the new enum values for the data items that were
>> undefined before 3.0.
>
> This is true but irrelevant to the discussion.
>

To clarify, the SMBIOS 3.0 spec explicitly allows the 32-bit entry
point to either point to the same table as the 64-bit entry point, or
point to a separate table, in which case the contents of the latter
should be a subset of the contents of the former. It doesn't specify
anything about the version number to be used in the 32-bit entry point
in case they point to separate tables. This means the presence of the
32-bit entry point does not guarantee that the table contents are
compatible with the pre-3.0 tools. So perhaps it would make sense to
export the 32-bit entry point separately only if it points to a
different table, and has a different version number?

-- 
Ard.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/