Bug 1176

Summary: keyboard stops responding after 1 .. 2 minutes
Product: xorg Reporter: danny backx <danny.backx>
Component: Input/KeyboardAssignee: Xorg Project Team <xorg-team>
Status: RESOLVED WORKSFORME QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium CC: cs, nigel, orion
Version: 7.0.0Keywords: NEEDINFO
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Bug Depends on:    
Bug Blocks: 890    
Attachments:
Description Flags
files as described in the message
none
new logfile
none
logfile with old driver
none
xorgs log none

Description danny backx 2004-08-24 21:00:32 UTC
When using the X.org X server, the keyboard of my PC stops working after a 
short while. 
Applications still run, the mouse pointer follows and still operates the 
applications. 
Apparent "keyboard" functions such as toggling NumLock doesn't make the NumLock 
LED change state; same for Caps Lock. Also key combinations such as Ctrl-Alt-Fx 
to switch to another virtual console don't work any more. 
 
This happens both with the line Driver "kbd" and Driver "keyboard" in 
XF86Config. 
 
I'm running on Mandrake 10.0 with all patches until a couple of days ago 
installed, X from X.org CVS as of a couple of hours ago. 
Platform is a regular PC (built it myself), Pentium-4, 768MB RAM. 
Keyboard is PS/2 style, mouse is USB. Login is via KDM. 
 
I'm hoping I can attach XFConfig and Xorg.0.log to this report, if I don't find 
how to do that, find them at 
http://home.scarlet.be/~md032472/download/xorg-logs.tar.gz . 
 
In that file the Xorg.0.log.old is from a session with the problem. 
The Xorg.0.log is from a session in which I was fast enough to type the 
commands to switch back to my old configuration. So if there are messages 
relating to the problem, they may be in one file and not the other.
Comment 1 danny backx 2004-08-24 21:02:44 UTC
Created attachment 722 [details]
files as described in the message

Obviously the type of this attachment is a gzipped tar, containing 3 files.
Comment 2 Kristian Høgsberg 2004-08-25 06:06:05 UTC
In recent CVS "kbd" and "keyboard" both map to the new driver. Could you try to add

  #define UseDeprecatedKeyboardDriver YES

to config/cf/host.def and recompile (you can just do "make Makefiles all" in
programs/Xserver) and then set the keyboard driver to "keyboard" in your config
file.  This should set you up with the old driver, so we can see if it's an
issue with the new driver.
Comment 3 danny backx 2004-08-26 09:49:43 UTC
Created attachment 737 [details]
new logfile
Comment 4 danny backx 2004-08-26 09:52:31 UTC
Used your #define.
Did "make world" (after some build error in lib/Xdmcp).
Then the build went fine.

The result was not better than before.

How can I see whether there's a different driver in use ? Diffing the
Xorg.0.log files doesn't appear to show that.
Comment 5 Kristian Høgsberg 2004-08-26 13:10:24 UTC
(In reply to comment #4)
> Used your #define.
> Did "make world" (after some build error in lib/Xdmcp).
> Then the build went fine.
> 
> The result was not better than before.

Are you still using the "kbd" driver?  That will always load the new driver. 
Make sure your keyboard InputDevice section is like this

Section "InputDevice"
    Identifier "Keyboard1"
    Driver "keyboard"
    Option "XkbModel" "pc105"
    Option "XkbLayout" "en_US"
    Option "XkbOptions" ""
EndSection

I.e. make sure it has the 'Driver "keyboard"' line.

> How can I see whether there's a different driver in use ? Diffing the
> Xorg.0.log files doesn't appear to show that.

There should be a warning in the log about using the old driver.
Comment 6 danny backx 2004-08-27 10:53:21 UTC
Created attachment 746 [details]
logfile with old driver

Changed the config file to use the old driver; this time the warning about
using that is in the log file.
Behaviour is unchanged : keyboard is dead either immediately (did I write that
already? A KDE session takes a while to start up so the 'live period' of the
keyboard may have passed already; I don't know.) or after a while.
Comment 7 Kevin E. Martin 2004-09-01 10:27:17 UTC
On today's release wranglers call, we decided to set a deadline for patches of
today at 6pm EDT.  If patches are available by then, are small and
self-contained, they can still be integreated into the release.  Otherwise, we
will move the remaining open bugs to the release notes bug 999 for documentation.
Comment 8 Kevin E. Martin 2004-09-01 18:27:45 UTC
Moving to release notes bug 999 as noted earlier.
Comment 9 danny backx 2004-09-13 11:20:56 UTC
I may have found the real reason now; curing this is not something I can do,
unfortunately - leaving that to more knowledgeable folks.

The reason this is happening to me appears to be related to linux's boot.
I am guessing that Xorg's keyboard dies when linux's /bin/login or so
initialises the keyboard. (XFree86 never did this.)

This originally happened to me, I think, because the linux RC scripts
launched X before all other services were finished. One of the services
starting later was - in my initial setup - sendmail with an invalid
configuration file which caused it to pause for two minutes. After that,
the end of the boot would kill my keyboard.

This is now gone, and I've seen Xorg without ever responding to keyboard
events. (Which is worse than initially.)

I've now delayed X startup (/etc/inittab only brings me to level 3).
Then I log in to the console and say "telinit 5" to start X.

In this configuration, everything works.

Modifying inittab again brings me back to the bad state.

So I'm now thinking that Xorg and the bootup interfere with eachother,
leaving Xorg in a bad state no matter in which order they pass by
eachother.

Does this help anyone ?

Otherwise, I'm pleased to say that X11r6.8 appears to work nicely
on my Mandrake 10.0 .
Comment 10 Eric Shattow 2004-09-17 00:26:04 UTC
Confirmed on athlon-xp Gentoo Linux with both via and nforce2 chipsets.
Switching to "evdev" protocol appears to be helping.  still no idea what is
causing this or why it is happening.  interesting note?   logging out via use of
just the mouse, gdm loads up and the keyboard works again.   repeatable, sort
of.  WTF?
Comment 11 Greg Ryman 2004-09-17 00:32:28 UTC
Confirmed on a AMD Athlon 64. Running Gentoo 64 with 512 mb's ram. Xorg 6.8.0
with KDE 3.3.0 and Kernel 2.6.9-rc2. Keyboard stops responding, all applications
work as if there were no problem, however the keyboard acts as if it has been
unplugged. Video drivers are Nvidia, motherboard chipset is Via on a Abit KV8.
Comment 12 Eric Shattow 2004-09-27 13:52:22 UTC
switched to evdev protocol for input devices (2 mouse devices, 2 keyboards). 
have not experienced problem since. 
Comment 13 Nicolas Dubuit 2004-10-15 07:02:19 UTC
(In reply to comment #9)

(Mandrake 9.2, P4-2.4, 1GB RAM, xorg 6.8.1)

Same pb for me; the bug appears when starting X at boot (with kdm), and
disappears if I log in on console and then startx.
Comment 14 Russel Roeber 2004-10-19 06:54:29 UTC
(Mandrake 10.1, P4-3.0, 1GB RAM, xorg 6.8.1)
I get the same reaction with 1 exception, my keyboard never works.
I tried everything listed above with no luck.
Comment 15 Joost Wüstefeld 2004-10-20 16:35:57 UTC
I also see this, and like Russel can't do anything to get the keyboard 
working. I'm on a centrino notebook. The only way I can type anything is via 
an attached USB Keyboard, which before my update from 6.7.0 to 6.8.0/1 just 
worked in parallel with the internal keyboard. I just hope somebody has got a 
hint how to cure this?  
Comment 16 Joost Wüstefeld 2004-10-21 01:04:27 UTC
Today I found that things go back to normal for me if I switch from my freshly 
installed 2.6.9 kernel back to 2.6.8.1. Hope that helps... 
Comment 17 Russel Roeber 2004-10-21 05:51:55 UTC
Update.
If boot to level 3, wait 2 mins and then start X with "telinit 5" then my 
keyboard works.

Just like danny backx in comment #9.
Comment 18 Stiffux 2004-12-28 02:27:08 UTC
Hello All,

Just another comment to confirmed this bug.
My motherboard is an Asus A7N266-E with nforce2 integrated.

The server X (6.7.0-3mdk: Mandrake10.1) freezes just after login. The keyboard
is dead whereas the mouse still active but nothing can be done.

There is no special error message within logs !!!

I hope that someone will resolve this bug.
Comment 19 Steffen Schwientek 2006-01-15 08:02:04 UTC
I hope someone is reading this old bugreport, because this describes exactly 
the behaviour of my keyboard with the modular Xorg.  
 
I'll attach xorgs.log, if this happens again. 
 
 
 
 
Comment 20 Steffen Schwientek 2006-01-15 08:40:43 UTC
Created attachment 4350 [details]
xorgs log

I think the problem only occurs, if there is some mouse and keyboard
action...Never experience this bug during hacking text.
Comment 21 Andreas. S. 2006-05-30 11:57:55 UTC
I'm also having this problem on x.org 6.9/slackware current. I boot into console and start X 
with startx.
It sometimes takes hours and sometimes only half an hout till my keyboard stops 
responding. The Num key doesn't toggle the light. 
Everything is back to normal as soon as I exit X (keyboard works on console).
Comment 22 Daniel Stone 2007-02-27 01:23:55 UTC
Sorry about the phenomenal bug spam, guys.  Adding xorg-team@ to the QA contact so bugs don't get lost in future.
Comment 23 Nigel Cunningham 2007-09-02 19:37:34 UTC
+1. I think bug 7270 is a duplicate of this.

I'm running current xorg git, seeing the issue on an AMD64 (Ubuntu Gutsy pre-release based) laptop. I'm happy to do some work on diagnosing this if noone else has the time. It's annoying enough to make me want to!

Not sure why this was put in the driver/via category. Changing to input/keyboard if bugzilla will let me.
Comment 24 Orion Poplawski 2007-10-08 12:56:03 UTC
(In reply to comment #23)
> +1. I think bug 7270 is a duplicate of this.

Person in 7270 seems to have basic functionality, just not VT switching or keyboard maps.  What one of my F7 users is experiencing, like this report, is complete loss of keyboard responsiveness.  It's not hardware as /dev/input/input1 does show keyboard activity.
Comment 25 Nigel Cunningham 2007-10-11 20:42:31 UTC
Ok. I'm still seeing output from xev, but some higher level in the feeding chain seems to be throwing away some keystrokes. In my case, when it happens, modifiers stop working, and so do some other keys (Home, End etc), but a-z and backspace keep working. That's not what you're seing?
Comment 26 michael bane 2008-01-02 11:06:50 UTC
I also see this - a seemingly dead keyboard (num lock doesn't toggle light) but yet the /dev/input is seeing the input.

I also see problems with the mouse too (hardly to repro).

Both are USB and I'm running Debian etch on AMD64 with
ii  xorg                                       7.1.0-19                                   X.Org X Window System

Thanks, Michael 
Comment 27 Peter Hutterer 2008-04-30 23:56:55 UTC
(In reply to comment #25)
> Ok. I'm still seeing output from xev, but some higher level in the feeding
> chain seems to be throwing away some keystrokes. In my case, when it happens,
> modifiers stop working, and so do some other keys (Home, End etc), but a-z and
> backspace keep working. That's not what you're seing?

If you have a second machine, you could figure out where the stuff disappears.
Run X from gdb and put a breakpoint on xf86PostKeyboardEvent. If this breakpoint is hit (but nothing happens), the server loses the event. If not, then the driver does.


Do you have a reproducable testcase?

Comment 28 Nigel Cunningham 2008-07-16 16:02:52 UTC
I'm now convinced it's triggered by VMware; I don't see the issue anymore unless I use VMware and when I do see the issue it's fixed by going into system->preferences->keyboard (Gnome) and selecting a different "Keyboard model" in the Layouts tab.

HTH
Comment 29 Peter Hutterer 2008-07-16 17:03:44 UTC
> --- Comment #28 from Nigel Cunningham <nigel@nigel.suspend2.net>  2008-07-16 16:02:52 PST ---
> I'm now convinced it's triggered by VMware; I don't see the issue anymore
> unless I use VMware and when I do see the issue it's fixed by going into
> system->preferences->keyboard (Gnome) and selecting a different "Keyboard
> model" in the Layouts tab.

does this mean that it happens when you run _inside_ vmware or when you run
vmware as a client?
Comment 30 Nigel Cunningham 2008-07-16 17:06:31 UTC
When I run vmware as a client and switch focus from VMware to another app - that's when I have a chance of seeing it. I think it may be related to whether keys are being pressed in the VMware guest as I move the mouse out of the guest window, but am not completely sure of that. (I really should give it a test, but never seem to find the time).
Comment 31 Jeremy Huddleston Sequoia 2011-09-25 16:43:44 UTC
No response in a few years and the previous assumption that this is a VMWare 
bug.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.