Re: [PATCH] platform/x86: Add new msi-ec driver

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

Hi,

Before diving into a full review of the driver I thought
it would be good to first catch up with the email thread sofar.

I agree with everything discussed so far. One clarifying remark
below:

On 2/15/23 22:27, Nikita Kravets wrote:

<snip>

>> This will start poking the embedded controller when the module is loaded, regardless of the platform. I am not sure that is desirable.
> 
> It only reads though, can it cause any harm?

I have seen cases where some weird (i2c) hw reacts to
reads as if they are writes.

But since this is using the standardized ACPI EC access
routines I believe that doing reads should generally
speaking be safe.

Also it seems that atm the module must always be loaded
manually ?

I don't see any MODULE_ALIAS or MODULE_DEVICE_TABLE entries
in the driver to make it auto-load.

I think this should get a dmi_system_id tables with known
MSI DMI_SYS_VENDOR() matches in there + a
MODULE_DEVICE_TABLE() pointing to the dmi_system_id table
to have the driver auto-load on MSI systems.

I think what we should do here is:

1. Get the module in the mainline kernel
2. Do a blogpost asking MSI laptop users to test
3. Assuming testing does not show any regressions / issues
   (it will likely show lots of unsupported fw versions)
   add the MODULE_DEVICE_TABLE().

Regards,

Hans