Re: PROBLEM: atkbd: Incorrect (?) handling of BAT from laptop PS/2 keyboard

Hi Alexander,

On Sun, Jul 02, 2017 at 04:32:18AM +0300, Alexander Babayants wrote:

Hi,

HP Pavilion 15 keyboard controller sometimes (as far as I've found, at
least after keyboard (re-)connection and LED status change) emits 0xAA
scancode (BAT passed) and key-up events for LCtrl, LAlt, RCtrl, RAlt,
RShift. This, in turn, causes the keyboard to reconnect
(drivers/input/keyboard/atkbd.c, line 424) and to emit 0xAA scancode again,
which results in keyboard being stuck in endless loop of reconnecting.
While in this loop, keyboard is usable, but some keystrokes are being
occasionally dropped.
Is it necessary to handle BAT scancode in such particular way? Couldn't it
just be ignored?

We should not ignore BAT, as it normally indicates that the keyboard
lost its state and we need to re-initialize it (update LED state to be
in sync with the keyboard state, set up repeat rate, select proper set -
2 or 3, etc).

I would be interesting to see the command stream there. Can you try
booting with i8042.debug and reproducing the issue?

Thanks.

-- 
Dmitry