Re: [PATCH v2] iio: accel: bmc150: Check for a second ACPI device for BOSC0200

From: Andy Shevchenko
Date: Tue Jan 30 2018 - 14:05:42 EST


On Tue, Jan 30, 2018 at 8:34 PM, Steven Presser <steve@xxxxxxxxxxxxx> wrote:
> Andy,
>
> I apologize for the long response, but there's several issues to address
> here.

NP, it it a good explanation why. That's what commit message missed apparently.

> First, I believe the "bmc150" in the subject line is in some way a misnomer.
> You'd have to ask Jeremy for more details on what he intended it to refer
> to. However, I believe the device in question is actually the bma250[1],
> which does not have a magnetometer component. I'm unfortunately away from
> my notes, but I can check later if you need me to verify the exact chip.

Please do, I would really be on the safe side here.

> Second, we're seeing a difference between what's in the data sheet and
> what's exposed in the wild via ACPI. I own the laptop that started the
> process of building this patch and I did the original ACPI-tables
> investigation.
>
> The device in question (BOSC0200) appears in the Lenovo Yoga 11e (and
> possibly other laptops - this happens to be the one I own). These laptops
> have a 360-degree hinge between the screen and the keyboard, letting them
> convert into tablets, if the user desires. The 11e implements this
> mode-switching by placing an accelerometer in each of the screen and
> keyboard, then doing math with the resulting vectors to figure out the angle
> between the two.

This makes a lot of sense.

> For whatever reason, Lenovo chose to expose these two
> (physically separate) accelerometers via a single ACPI device which presents
> two i2c devices at sequential addresses.


> As part of my original investigation of the Yoga 11e, I wrote a
> proof-of-concept of pulling accelerometer data from the two devices exposed
> under the BOSC0200 ID and using that to calculate the position of the screen
> relative to the keyboard. So based on my empirical experience, I can tell
> you the BOSC0200 device ID can expose two accelerometers at sequential
> addresses in the wild.
>
> I don't understand why Lenovo has reused the BOSC0200 ACPI device ID for a
> device that is fundamentally different from the base device. The ID doesn't
> belong to them and we're (apparently) now stuck in this situation where this
> ACPI device ID could represent two different device layouts.

Bad, bad Lenovo. (DMI strings might help here)

> Finally - Andy, I apologize if I came across as challenging you in my
> initial mail. I was trying to strike a balance between brevity/respecting
> your time and asking a question. Evidently I struck the wrong balance and
> should have given you more background on why I was doubting what you saw.
> This is my fault and you have my sincerest apologies for any offense I have
> caused.

No need, the root cause is lack of description in the commit message.

Nevertheless, the approach chosen I don't like. It looks like an ugly hack.

What we can do here is:
- do not contaminate core part with I2C/SPI/etc
- do not create another driver via board_info, we already in *the same* driver,
so, the better approach here AFAICS is to add DMI quirk into i2c-core-acpi



> Steve
>
> [1]
> https://ae-bst.resource.bosch.com/media/_tech/media/datasheets/BST-BMA250E-DS004-06.pdf

--
With Best Regards,
Andy Shevchenko