Re: [PATCH 1/2] platform/x86: silead_dmi: Add touchscreen info for PoV mobii wintab p800w

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 17-06-17 01:38, Darren Hart wrote:
> On Fri, Jun 16, 2017 at 03:22:45PM +0200, Hans de Goede wrote:
>> Hi,
>>
>> On 16-06-17 14:44, Andy Shevchenko wrote:
>>> On Thu, Jun 15, 2017 at 7:53 PM, Darren Hart <dvhart@infradead.org> wrote:
>>>> On Thu, Jun 15, 2017 at 08:48:31AM +0200, Hans de Goede wrote:
>>>
>>>>> +             /* Point of View mobii wintab p800w */
>>>>> +             .driver_data = (void *)&pov_mobii_wintab_p800w_data,
>>>>> +             .matches = {
>>>>> +                     DMI_MATCH(DMI_BOARD_VENDOR, "AMI Corporation"),
>>>>> +                     DMI_MATCH(DMI_BOARD_NAME, "Aptio CRB"),
>>>>> +                     DMI_MATCH(DMI_BIOS_VERSION, "3BAIR1013"),
>>>>> +                     /* Above matches are too generic, add bios-date match */
>>>>> +                     DMI_MATCH(DMI_BIOS_DATE, "08/22/2014"),
>>>>
>>>> This is the first time I've seen a BIOS date match used to determine hardware
>>>> features. DMI matching is a (necessary) hack to begin with (the vendors should
>>>> be providing this data via ACPI _DSD anyway) but a date match means we would
>>>> need a kernel patch every time one of these tablets gets a BIOS update...
>>>>
>>>> With words like "Aptio CRB" it's clear the vendor isn't doing their job and just
>>>> using unmodified reference code. The problem with this of course is that the
>>>> vendor is not providing a way to identify this hardware.
>>>>
>>>> Andy, I'd appreciate your thoughts on this... I'm leaning towards not accepting
>>>> bios date (or indeed, BIOS version) as a way to identify a platform.
>>>
>>> The question is what is the anticipated amount of affected devices
>>> with BIOS date included and otherwise?
>>
>> I expect there to be collisions (false positive matches) without the
>> BIOS_DATE check, a quick web-search finds other devices with a
>> 3BAIR1013 bios version. Those don't necessarily also use a Silead
>> touchscreen (which is needed for a collision to happen), but given
>> the popularity of Silead touchscreens on cheap devices a collision
>> is not unlikely.
>>
>> With the bios-date check added, I expect this match to be unique,
>> for it to not be unique we would need to be really unlucky.
>>
>>> If Hans believes that there will be no update for some devices,
>>
>> Yeah I'm pretty sure this specific device will not see any
>> BIOS updates ever.
>>
>>> while there are devices with the same DMI strings, but different date and
>>> _fixed_ issue, I think we have no other choice for now.
>>> Also can we use some other strings to distinguish group of devices
>>> which are affected?
>>
>> bios_date: 08/22/2014
>> bios_vendor: American Megatrends Inc.
>> bios_version: 3BAIR1013
>> board_asset_tag: To be filled by O.E.M.
>> board_name: Aptio CRB
>> board_serial: T80091A4C11B0848
>> board_vendor: AMI Corporation
>> board_version: To be filled by O.E.M.
>> chassis_asset_tag: To Be Filled By O.E.M.
>> chassis_serial: To Be Filled By O.E.M.
>> chassis_type: 3
>> chassis_vendor: To Be Filled By O.E.M.
>> chassis_version: To Be Filled By O.E.M.
>> product_name: To be filled by O.E.M.
>> product_serial: To be filled by O.E.M.
>> product_uuid: 03000200-0400-0500-0006-000700080009
>> product_version: To be filled by O.E.M.
>> sys_vendor: To be filled by O.E.M.
>>
>> The product-uuid is a known example uuid, so is
>> no good. The board_serial might be useful, but
>> only if it is unique for the model and not per
>> tablet. Unfortunately I only have 1 of these
>> tablets, so I cannot tell.
> 
> Do we have any indication that this BIOS Date isn't just the default value
> provided by AMI?

In my experience with devices with similar generic DMI strings,
the BIOS date is different for all of them, I believe the
tools use to build the BIOS set this to the actual build date.

> Does it offer any more information than the BIOS Version?

Yes it does a web search for "3BAIR1013" finds unrelated matches,
so that alone is not enough, where as the BIOS date tends to be
unique.

> I suppose we may be able to do some kind of a partial match on the Board Serial
> if even that is platform specific (I suspect it is with the T800 at the
> beginning.
> 
> The sloppy handling of this firmware really irks me. That's obviously not Hans'
> fault, so we'll take the patch.

Thank you.

Regards,

Hans