X (xorg) - Dropped keystrokes, random repeats, held metas

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Hello, all!

I've got a fairly new notebook (Acer Aspire 1522MLi, amd64 (but running Gentoo x86, *NOT* amd64)) that's been running Gentoo smoothly from the outset after some initial wrestling with ACPI.

In X, and *only* in X (running xorg 6.8.1), the keyboard is ... irritating

I actually also had trouble with the touchpad for awhile (synaptics driver). The touchpad threw random mouse movement events and button clicks (left, right, and middle), with TONS of "driver lost sync at byte 1..." messages in the kernel. I fixed that by adding "i8042.nomux" to the kernel boot parameters, so now the mouse is happy.

Now all that's left is this !@#$ing keyboard issue. Here are the symptoms (these all happen at random, whether the others are happening or not):

1) Meta-keys (shift, ctrl, alt, win, menu) stay "held down" even after being released. An example of this: since I use Enlightenment with virtual desktops, I use ALT-F1 through ALT-F4 frequently to switch between desktops. Even though I release ALT after the switch, it stays held down. Thus trying to type through to the application on that desktop ends up sending all those keys as if I was holding down ALT as I type them. This has caused Bad Things(tm) to happen

2) Keys repeat even though they're not being held down. Example: push [enter] to run a command at a bash prompt. You'll keep seeing new prompts every line (scrolling nonstop) as [enter] is held down (even though the key is pressed and released, not held).

[note, 1 and 2 seem to be symptoms of an underlying problem -- it looks like keypress release events aren't caught by X]

3) When holding down keys (the arrow keys are the worst), other keypresses are occasionally injected. Example: type out a nice long command line on the shell, or type something long-winded into a text input area in Firefox. Hold the left arrow key to go back to the beginning of the line (don't cheat and hit [HOME]

). There will now be, in random spots, the number "4" inserted. Now go back to the end by holding the right arrow key down. The number "6" gets put in. Same thing happens (with different numbers for each key) for up and down arrows.

4) Keystrokes are dropped. I'll re-type a part of item 3 here without correcting any keystroke problems. My typing doesn't suck, I promise:

hardcore · l33t Joined: 01 Nov 2003 Posts: 626 Location: MSU, MI

This may sound stupid, but has someone spilled something sugary on the keyboard? My parents had such a keyboard, where alt would stay pressed down unless you banged on the keyboard.
_________________
Nothing can stop me now, cuz I just don't care.

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Heh, I wish someone had -- then I'd know who to smack

It's a pretty new machine -- it's been out of its box less than a month. It is a notebook, but it is in good condition, no spills (I did "sniff" it to make sure someone didn't spill something on it when I wasn't looking, but I rarely leave it unattended and no one else uses it).

So, no spills as far as I know.

Watching output from xev, it seriously does just seem to miss "Key Released" events, and also appears to miss "Key Pushed" events too.

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Check that, I was mistaken -- in xev, when a key starts repeating by itself, both a keypress and keyrelease event are generated.

Just gets nastier and nastier

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Also, just a quick note, 2.6.10-gentoo-r2 makes no difference in this behavior (just finished upgrading to it).

hardcore · l33t Joined: 01 Nov 2003 Posts: 626 Location: MSU, MI

Try popping a knoppix cd in, and run it for a few days, see if you can replicate the problem. Also if you have access to another OS, say windows, see the problem repeats itself using that OS. If you have the same problem across multiple OS's and linux distro's, you have yourself a hardware problem, and you should contact your notebook manufacturer.

It also could be your kernel which you correctly stated. Try using an earlier 2.6.* kernel, or a 2.4 kernel and see if the problem persists.
_________________
Nothing can stop me now, cuz I just don't care.

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Nah, I mentioned above (in the first horribly long-winded post, sorry

) that in Windows (XP Pro SP2) on the same hardware, all is well.

Also, the Linux console is unaffected. If I don't start X, or if I switch out of X to the framebuffer console, all is well.

Thanks though -- I'll grab Knoppix and see if I can coax it into misbehaving.

rfdonnelly · Posted: Sat Jan 08, 2005 8:09 am Post subject:

In /var/log/messages do you see something similar to the following?:

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Yes, in droves. I've wondered about this but wasn't sure what to do about it (or whether it was related).

An example:

Jan 7 23:10:07 xerces atkbd.c: Unknown key released (translated set 2, code 0xe0 on isa0060/serio).
Jan 7 23:10:07 xerces atkbd.c: Use 'setkeycodes e060 <keycode>' to make it known.

rfdonnelly · Posted: Sat Jan 08, 2005 9:14 am Post subject:

Well I wasn't able to find a straight answer for why this happens but I got the message to stop and the weird keyboard behavior seems to have stopped.

Basically do what the log tells you: setkeycodes e060 <keycode> where keycode is any(?) number not being used already. Use getkeycodes to check keycodes in use. Looks like mostly 80-255 are used so you should be safe just using a smaller number: setkeycodes e060 55

willfe · n00b Joined: 08 Jan 2005 Posts: 16

When you mentioned the keycode error initially, I jumped the gun and started running setkeycodes against all the keys it complained about.

I've now set keycodes for each unknown event it spots, and the repeat/dropped keys issue persists

The irritating thing here is that it *still* throws the errors into /var/log/messages even though getkeycodes shows definitions for the keys I defined.

Bleh

Is it time for the mallet yet?

lothar · Posted: Sat Jan 08, 2005 6:29 pm Post subject:

I've got exactly the same problem!!! Its really annoying, and I get the same error in /var/log/messages...
Need a solution for this

willfe · n00b Joined: 08 Jan 2005 Posts: 16

So are there any other ideas for solving this?

lothar · Posted: Sun Jan 09, 2005 6:54 pm Post subject:

When the keys get messed up on my comp. I just press all the meta keys and hammer all I can on the keyboard and suddenly after 5min or something the keys are back to normal.. Cant loose the precious uptime

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Heh. The "bang on the keys" approach both feels good and makes meta keys stop sticking, but doesn't help with the repeating or dropped keystrokes

Jtb · Posted: Wed Jan 19, 2005 12:05 pm Post subject:

I have an Acer Aspire 1501LMi and my keys seems to stuck also sometimes..

But this is not also a Gentoo/Linux-Problem but also a Windows (XP) problem - sometimes while playing e.g. Q3 one of the cursorkeys get stuck and I strafe all the time to left/right/..
Very often CTRL or ALT gets Stuck - maybe I should call Acer if it's a known problem

_________________
Jens

.. God is real - unless declared integer!

willfe · n00b Joined: 08 Jan 2005 Posts: 16

See, lately I've thought it was hardware too, so to make sure I've been running the machine on Windows XP Pro for the last few days. I've played games, written journal entries, even coded (*gasp* -- ActiveState's ActivePython has a pretty damned nice development environment; I'm impressed), and haven't seen ONE dropped keystroke, repeat event, or other random nastiness exhibited in X.

As an update, I also ran for a couple days just on the framebuffer console (with X running in the background and without). The problem is definitely at its worst under X, however I *did* see very occasional keystroke drops even on the console. That and the "missed key release event" thing happens (as CTRL, SHIFT, ALT, etc., get held down) but not nearly as frequently. Whatever X is doing to poke this bug, it's sure good at it.

So to recap:

* In X, it's terrible. Lots of missed keystrokes, repeats, held-down meta keys, etc.
* On Linux console, it's definitely livable; one out of every five hundred strokes are missed, and over two days only one meta key got held down for me.
* On Windows, all is well. Not a single held key, dropped stroke, or repeat event.

If it's hardware, I'm stumped as to what it could really be.

stuart · Guru Joined: 27 Apr 2003 Posts: 552

i have same problem
if you have any solution please post here
_________________
Pochissime persone crescono. La maggior parte della gente invecchia.

willfe · n00b Joined: 08 Jan 2005 Posts: 16

Posting an update, since the notebook had a hardware problem.

The hard disk in the machine crashed. Acer shipped a replacement, I reloaded everything (Windoze first, *sigh*), then put a fresh shiny new Gentoo install on the second partition from scratch again.

Days of compiling, and ...

... the damned keys still stick, hold down, get dropped, etc. The problem persists.

Bleh. Any ideas, anyone?

lothar · Posted: Mon Jan 31, 2005 8:40 pm Post subject:

I've bough a new barebone computer.. The old computer is now just acting as server.. I've got all the same apps on the new computer, the only difference is that I'm not using ~x86 xorg, on the old computer I always had the "unstable" xorg..

Anyway, on the new computer I haven't got this problem at all. I'm even using my old home folder, shared via NFS from the server.. Weird..
Its either a hardware problem or a Xorg problem

rfdonnelly · Posted: Tue Feb 01, 2005 3:47 am Post subject:

rfdonnelly · Posted: Tue Feb 01, 2005 3:51 am Post subject:

willfe · n00b Joined: 08 Jan 2005 Posts: 16

I booted Knoppix 3.7 an hour ago and messed with it for awhile. Pretty spiffy actually, hadn't used it for a few years. ndiswrapper didn't work right though

However, the *keyboard* was flawless. X (XFree86, as opposed to xorg) started up fine (VESA-only; support doesn't yet exist for VIA Unichrome/S3 chipsets on notebook display panels), mouse and keyboard were both well behaved. No dropped keys, no repeats, no held-down metas, etc.

I did note in X's startup log that it was using the "legacy" driver to handle the keyboard. Does such exist on the x.org server?

lothar · Posted: Wed Feb 02, 2005 12:47 pm Post subject:

I want the best candy · n00b Joined: 03 Feb 2005 Posts: 1

Sorry for a long post. I feel really sorry especially considering that it won't give any definitive answers to the problem. I will do my best to make it short. I emphasize that this won't be helpful very much, as I know very little about it.

1. This seems to be a xfree86 problem. It happens with xfree86, and the xorg implementation in at least two different underlying OS. Here is the post about the incident in FreeBSD - http://lists.freebsd.org/pipermail/freebsd-questions/2004-September/059594.html
2. The condition that causes the problem can't be narrowed down at the moment, but it seems to have something to do with
hardware (motherboard, KVM, PS/2 extension cable),
compile option of X.
3. There are folks who somehow "solved" the problem. The most were able to recompile X in a way that doesn't bring up the issue by trying different compile options.
4. It also affects a CPU usage, and sometimes involves a mouse problem.
5. Alt and Ctrl seem to be the most involved keys, and that often make the problem appear to be something else. For example, with these keys stuck, your mouse operation may give a strange behavior, and a user might not suspect the keyboard problem. The other keys of course may confuse users, like keys repeating but no input field opened, but slowing down CPU, which may appear to be a problem not related to keyboards.
6. Those who escaped the problem did so by recompiling X, so your best bet at the moment as gentoo users is trying different sets of USE in make.conf that affect X and see any of those do the trick.
7. The bug seems reported to XFree86. See http://marc.theaimsgroup.com/?l=xfree86&m=104879558008189&w=2 Haven't read any response from XFree86.

more observations.
related to #1 above - I have probably tried more than a dozen of different kernel versions to no avail. I have read others who tried different glibc without success. Desktop environments don't seem to matter either.
related to #2 - I've seen one motherboard that has this problem with a very long PS/2 extension cable, and another that shows the symptom much, much less frequently without KVM. The former may still have the problem without an extension cable, but I couldn't see it in a short period of time that I used.

Here is the link to the gentoo forum discussion. It includes the links to more posts about the problem. https://forums.gentoo.org/viewtopic.php?t=156477