RE: [PATCH 2/2] acpi,backlight: MSI S270 laptop support - driver

[Brown, Len]

>If you look closely you'll see that the patch already adds a new
>entry to MAINTAINERS.

Excellent.

>I put the the driver in drivers/acpi/ because it relies heavily on the
>ACPI EC stuff. But OK, what is a better place for the driver? 

It is possible that parts of it should live in different places,
but I agree that it is also desireable to keep the compnents of a
platform
specific driver together in as few files as possible.

>drivers/char/ 
>  It doesn't even register any character device.

I believe that is okay, and that creating a /dev file isn't a
requirement.

>drivers/video/backlight/
>  It doesn't just do backlight control.

Perhaps the backlight part should live there.

>drivers/misc/
>  Seems to be the last resort for everything that doesn't fit it
>  otherwise.
>
>Unless anyone has a better idea I will move it to drivers/misc/, then.

Yeah, there is probably a better place than misc.

The only thing I can say for sure is that it implements
platform-dependent
features rather than standard features in the ACPI spec,
and thus it must live outside of drivers/acpi/.

>> lcd brightness platform support should talk to the existing
>> backlight stuff under sysfs.
>
>It already does exactly that, you can find the backlight class driver
>in /sys/class/backlight/s270/.

Excellent.

>I cannot map the "automatic brightness control" feature to the
>backlight class driver, that's why I duplicated the brightness stuff
>in /proc/acpi/s270/.

Better to extend the backlight code so that it can handle the
special features of this platform than to duplicate brightness
control in multiple places.

>> wlan and bluetooth indicators/controls need to appear under
>> generic places under sysfs -- not under platform specific
>> files under /proc/acpi.
>
>What are those "generic" places? I cannot think of any besides a
>"platform" device.

They should appear as properties under their associated
devices in /sys/devices tree rather than invening a new place.

>Any ideas on that ec_transaction() patch I sent earlier?

I agree 100% that ec.c needs to be whacked.  Unfortunately
there seem to be 3 people doing it at the same time,
so we'll have to sort that out.

thanks,
-Len

ps. Lennart, please understand that I didn't intend to be gruff.
I figured that by the quality of your work -- which is better than most
--
that I'd go for an immediate reply, even though I was still holding a
sharp pointed stick at the end of a 6-hour bug scrub marathon.
-
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/