2.6.28-rc4-mmotm1110 shift-lock oddness

2.6.28-rc4-mmotm1110 shift-lock oddness

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 1310 bytes --]

Twice today, I've managed to get my laptop into an odd state where the keyboard
got quite confused:

1) For the characters [a-z] the shift bit was *inverted* - if the shift-lock
LED was off, you'd get UPPER-case, and if it was on, you'd get lower-case.

2) To add to the fun, the 0-9 keys were stuck shifted - I'd get the special
characters no matter what setting the shift or shift-lock key had.  I suspect
that the special-char keys like [{ and ;: were also stuck in shifted state,
but can't totally confirm it.

The previous two issues combined meant that I wasn't able to su or login
as root to do much of anything about it (guess who has special chars in their
root password? ;)

3) ctl-alt-Fn to switch virtual consoles was spotty/erratic - one time
I was unable to use it at all, another time I was able to use c-a-F1 to get
to a green-screen console, but then F2 to get another one or F7 to get
back to the X server screen wouldn't work.

4) Fortunately for my sanity, alt-sysrq was still working as advertised,
so I was able to reboot and thus clear the problem.

I was thinking this might be a simple "missed a key press/released" event
for a shift key, except that doesn't explain the numerics getting stuck shifted.

Any suggestions/hints for instrumentation/etc, in case I can replicate it
again?

[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: 2.6.28-rc4-mmotm1110 shift-lock oddness

[Andrew Morton]

On Mon, 17 Nov 2008 16:56:55 -0500 Valdis.Kletnieks@vt.edu wrote:

> Twice today, I've managed to get my laptop into an odd state where the keyboard
> got quite confused:
> 
> 1) For the characters [a-z] the shift bit was *inverted* - if the shift-lock
> LED was off, you'd get UPPER-case, and if it was on, you'd get lower-case.
> 
> 2) To add to the fun, the 0-9 keys were stuck shifted - I'd get the special
> characters no matter what setting the shift or shift-lock key had.  I suspect
> that the special-char keys like [{ and ;: were also stuck in shifted state,
> but can't totally confirm it.
> 
> The previous two issues combined meant that I wasn't able to su or login
> as root to do much of anything about it (guess who has special chars in their
> root password? ;)
> 
> 3) ctl-alt-Fn to switch virtual consoles was spotty/erratic - one time
> I was unable to use it at all, another time I was able to use c-a-F1 to get
> to a green-screen console, but then F2 to get another one or F7 to get
> back to the X server screen wouldn't work.
> 
> 4) Fortunately for my sanity, alt-sysrq was still working as advertised,
> so I was able to reboot and thus clear the problem.
> 
> I was thinking this might be a simple "missed a key press/released" event
> for a shift key, except that doesn't explain the numerics getting stuck shifted.
> 
> Any suggestions/hints for instrumentation/etc, in case I can replicate it
> again?

That's the first such report I've seen.  Let's add some cc's.

Re: 2.6.28-rc4-mmotm1110 shift-lock oddness

[Jiri Kosina]

On Mon, 17 Nov 2008, Andrew Morton wrote:

> > Twice today, I've managed to get my laptop into an odd state where the keyboard
> > got quite confused:
> > 
> > 1) For the characters [a-z] the shift bit was *inverted* - if the shift-lock
> > LED was off, you'd get UPPER-case, and if it was on, you'd get lower-case.
> > 
> > 2) To add to the fun, the 0-9 keys were stuck shifted - I'd get the special
> > characters no matter what setting the shift or shift-lock key had.  I suspect
> > that the special-char keys like [{ and ;: were also stuck in shifted state,
> > but can't totally confirm it.
> > 
> > The previous two issues combined meant that I wasn't able to su or login
> > as root to do much of anything about it (guess who has special chars in their
> > root password? ;)
> > 
> > 3) ctl-alt-Fn to switch virtual consoles was spotty/erratic - one time
> > I was unable to use it at all, another time I was able to use c-a-F1 to get
> > to a green-screen console, but then F2 to get another one or F7 to get
> > back to the X server screen wouldn't work.
> > 
> > 4) Fortunately for my sanity, alt-sysrq was still working as advertised,
> > so I was able to reboot and thus clear the problem.
> > 
> > I was thinking this might be a simple "missed a key press/released" event
> > for a shift key, except that doesn't explain the numerics getting stuck shifted.
> > 
> > Any suggestions/hints for instrumentation/etc, in case I can replicate it
> > again?
> That's the first such report I've seen.  Let's add some cc's.

First, we'd need a little bit more information about the hardware in 
question, I guess.

What laptop is that, please? Is the keyboard PS/2-connected or 
USB-connected? Could you obtain evtest output from the buggy cases please?

-- 
Jiri Kosina

Re: 2.6.28-rc4-mmotm1110 shift-lock oddness

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 975 bytes --]

On Tue, 18 Nov 2008 01:26:51 +0100, Jiri Kosina said:

> First, we'd need a little bit more information about the hardware in 
> question, I guess.

Dell Latitude D820 laptop, with a USB-connected Microsoft Natural keyboard. I
don't think it's the USB subsystem going *completely* bonkers, as the
USB-connected trackball mouse functions just fine when this happens...

> What laptop is that, please? Is the keyboard PS/2-connected or 
> USB-connected? Could you obtain evtest output from the buggy cases please?

Hmm.. Got a pointer to a usable copy? The best I could find googling for
'linux evtest' was:

https://svn.openwrt.org/openwrt/packages/utils/event_test/src/evtest.c

which wants a /dev/inputN - all I have is /dev/input/ and entries under that
for the laptop's onboard mousepad and the trackball mouse, no keyboard entry.

Hmm part two:

% zgrep EVDEV /proc/config.gz 
# CONFIG_INPUT_EVDEV is not set

Should I be looking at rebuilding the kernel with this =m?



[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: 2.6.28-rc4-mmotm1110 shift-lock oddness

[Jiri Kosina]

On Mon, 17 Nov 2008, Valdis.Kletnieks@vt.edu wrote:

> > First, we'd need a little bit more information about the hardware in 
> > question, I guess.
> Dell Latitude D820 laptop, with a USB-connected Microsoft Natural 
> keyboard. I don't think it's the USB subsystem going *completely* 
> bonkers, as the USB-connected trackball mouse functions just fine when 
> this happens...

Does this happen only when Microsoft natural keyboard is connected, or 
does it happen also when you use the built-in keyboard?

What vendor/product ID does the USB keyboard have, please?

> > What laptop is that, please? Is the keyboard PS/2-connected or 
> > USB-connected? Could you obtain evtest output from the buggy cases please?
> Hmm.. Got a pointer to a usable copy? 

All decent distros ship this, usually as part of input-utils package.

> The best I could find googling for
> 'linux evtest' was:


> https://svn.openwrt.org/openwrt/packages/utils/event_test/src/evtest.c
> which wants a /dev/inputN - all I have is /dev/input/ and entries under that
> for the laptop's onboard mousepad and the trackball mouse, no keyboard entry.

Yes, /dev/input/eventN is used these days (when CONFIG_INPUT_EVDEV is 
turned on).

-- 
Jiri Kosina
SUSE Labs