Keyboard dead for a bunch of LG laptops

Keyboard dead for a bunch of LG laptops

[Joonas Saarinen]

Hello, people! There's still lingering an interesting regression where 
the internal keyboard of various older LG laptops is dead. It is tracked 
here:

https://bugzilla.kernel.org/show_bug.cgi?id=58991

Downstream:
https://bugzilla.redhat.com/show_bug.cgi?id=969550
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1243904
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1256776

Late last year, I personally did a git bisect on LG LW25 which ended up 
pointing at an atkbd patch:

be2d7e4233a4fe439125b825470020cef83407d5
Input: atkbd - fix multi-byte scancode handling on reconnect

Full description about that patch and some discussion:
https://groups.google.com/forum/#!topic/fa.linux.kernel/rUpZ8-HaQIM

Could it be that? The boot parameter i8042.dumbkbd=1 can be used to 
still make the keyboard functional. Of course that workaround leaves the 
keyboard a bit gimped (no lock LEDs, for example).

So, some questions:

1) What exactly happened? Is there something common with these LG 
machines' BIOS or keyboard controller firmware which causes the problem?

2) How could this happen? I thought the i8042 keyboard was one of the 
simplest devices in a PC. It's quite serious if it stops working like this.

3) Should something be done? Those are of course aging (Core 1 Duo era) 
machines, but I think Linux has always strived to work well with all 
sorts of hardware, old and new.


Joonas

Re: Keyboard dead for a bunch of LG laptops

[Shawn N]

Thanks for the bug report + bisect.

On Fri, Jan 24, 2014 at 8:21 AM, Joonas Saarinen <jza@saunalahti.fi> wrote:
> Hello, people! There's still lingering an interesting regression where the
> internal keyboard of various older LG laptops is dead. It is tracked here:
>
> https://bugzilla.kernel.org/show_bug.cgi?id=58991
>
> Downstream:
> https://bugzilla.redhat.com/show_bug.cgi?id=969550
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1243904
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1256776
>
> Late last year, I personally did a git bisect on LG LW25 which ended up
> pointing at an atkbd patch:
>
> be2d7e4233a4fe439125b825470020cef83407d5
> Input: atkbd - fix multi-byte scancode handling on reconnect
>
> Full description about that patch and some discussion:
> https://groups.google.com/forum/#!topic/fa.linux.kernel/rUpZ8-HaQIM
>
> Could it be that? The boot parameter i8042.dumbkbd=1 can be used to still
> make the keyboard functional. Of course that workaround leaves the keyboard
> a bit gimped (no lock LEDs, for example).
>
> So, some questions:
>
> 1) What exactly happened? Is there something common with these LG machines'
> BIOS or keyboard controller firmware which causes the problem?
>

It's hard to guess, it seems like a bug / quirk of the KB controller
implementation. It looks like ATKBD_CMD_ENABLE is not re-enabling the
keyboard after ATKBD_CMD_RESET_DIS is received. I will try to acquire
a LG LW25 laptop to debug.

> 2) How could this happen? I thought the i8042 keyboard was one of the
> simplest devices in a PC. It's quite serious if it stops working like this.
>

There are many independent KB controller implementations that use this
driver, each with different bugs / quirks. It's the danger of
modifying such a widely-used driver.

> 3) Should something be done? Those are of course aging (Core 1 Duo era)
> machines, but I think Linux has always strived to work well with all sorts
> of hardware, old and new.
>

I'm not sure what to do here, I will leave the decision to
"management" whether to revert. Meanwhile, I will acquire an LG laptop
that demonstrates this problem, and debug to get a better idea about
the failure case.

>
> Joonas

Shawn