FS#16587 - [KDE] Xorg crashes randomly on keyboard input

Attached to Project: Arch Linux
Opened by Mario K (empanyc) - Monday, 12 October 2009, 13:57 GMT
Last edited by Giovanni Scafora (giovanni) - Tuesday, 27 October 2009, 22:31 GMT
Task Type Bug Report
Category Packages: Extra
Status Closed
Assigned To Tobias Powalowski (tpowa)
Jan de Groot (JGC)
Pierre Schmitz (Pierre)
Thomas Bächler (brain0)
Andreas Radke (AndyRTR)
Architecture i686
Severity Medium
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 3
Private No

Details

Hi,
since last week i encounter a bug on two different PCs (both with ATI Radeon ~9600/9700).
If i try to login through SSH to some other servers, my Xorg crashes and i get the login
screen. The weird thing is, I cannot find a pattern from the affected servers and the
problem _disappears_ if I reinstall xf86-video-ati and restart Xorg manually. After this
everything works fine. Unfortunately i see no errors in the Xorg-Logfile. I'm not sure
if it's a coincidence, but it happens since i updated both PCs to KDE 4.3.2. One of the
two PCs has a fresh install of ArchLinux (3-4 days old).

As mentioned, i can't find any error message neither in Xorg.log nor in the kernel log,
respectively dmesg. Yesterday, I activated KMS, but the problem still remains.

Anyone has a hint?
This task depends upon

Closed by  Giovanni Scafora (giovanni)
Tuesday, 27 October 2009, 22:31 GMT
Reason for closing:  Works for me
Comment by Andreas Radke (AndyRTR) - Tuesday, 13 October 2009, 12:02 GMT
probably not a packaging bug... ask upstream.

new Xorg server will hit testing soon. You can try then if the problem is solved with it.
Comment by Mihael Pranjić (tux) - Tuesday, 13 October 2009, 13:43 GMT
Ok so I do not have ATI cards, but Intels...
I use x86_64 arch on 2 computers, where everything works fine
I also use i686 archlinux, on 2 EEEPCs, where nothing works :D

What happens:
I thought KDE was crashing after a random number of seconds, but it is not like that, when you type 9 it crashes immediately, when you hit enter it crashes sometimes

steps to reproduce:
use archlinux i686 testing
type "9"

a personal note: if this is a bug, its the best bug I have ever seen in my life.. it's pretty funny ^^
Comment by Mihael Pranjić (tux) - Tuesday, 13 October 2009, 13:46 GMT
one more thing: I downgraded back from testing, doesnt seem to work either
I also downgraded the kernel to a 2.6.30 kernel, and downgraded Xorg to an older version, doesnt work
stable version of xf86-video-intel does not work, and the AUR xf86-intel-newest doesnt work either

actually I dont know where this comes from
Comment by Mihael Pranjić (tux) - Tuesday, 13 October 2009, 18:38 GMT
EDIT: it crashes when you type "9" as often as you want and then hit enter.. so basically you can type as many nines as you want but it crashes when you hit enter
Comment by Musikolo (Musikolo) - Wednesday, 14 October 2009, 09:39 GMT Comment by Mario K (empanyc) - Wednesday, 14 October 2009, 16:27 GMT
andreas: how can i investigate whats happening during the reinstall of xf86-video-ati?
Maybe i can figure out why it's working after that.

Mihael: please stop hijacking my bug ;)
Comment by Musikolo (Musikolo) - Wednesday, 14 October 2009, 19:40 GMT
Hi mates,

I have looked a bit into it and have collected a few interesting files. Please, check them out and try to give me a hand.

Best regards! :-)
Comment by Musikolo (Musikolo) - Wednesday, 14 October 2009, 21:03 GMT
Hi again,

looking into the problem I have forced a crash by just typing the sequence '912 912'. Please, have a look to the attached file.

Best regards!
Comment by Musikolo (Musikolo) - Thursday, 15 October 2009, 05:32 GMT
Hi mates,

I have found out that for some unknown reason X server was starting in tty3 rather in tty7 as it should. All I have done is to move the 'kdm' daemon to the last place in my /etc/rc.conf file like this:

From this ---> DAEMONS=(syslog-ng @network netfs crond dbus hal kdm fam alsa ntpdate cups sensors athcool preload)
To this ---> DAEMONS=(syslog-ng @network netfs crond dbus hal fam alsa ntpdate cups sensors athcool preload kdm)

Whereas the first one uses tty3, the latter uses tty7 which does not crash at all. Don't know why this is happening as I have tried to move kdm daemon in a position in between preload and fam unsuccessfully.

Best regards! :-)
Comment by Andreas Radke (AndyRTR) - Thursday, 15 October 2009, 09:12 GMT
adding Tobias and Pierre beeing responsible for kernel and kde.
Comment by Mario K (empanyc) - Thursday, 15 October 2009, 11:15 GMT

The latest kernel update fixed my problem on one PC.
I will try the other this evening.

Btw, my part of the rc.conf:
DAEMONS=(syslog-ng network netfs crond ntpdate ntpd sshd dbus hal kdm cups)

Thanks!
Comment by Mihael Pranjić (tux) - Thursday, 15 October 2009, 11:25 GMT
Latest updates with testing enabled fix the problem I was talking about
Comment by Musikolo (Musikolo) - Thursday, 15 October 2009, 18:44 GMT
Hi again guys,

I have looked into the problem a bit more and next I'll show you my results. Starting from the only configuration I had found:
DAEMONS=(syslog-ng @network netfs crond dbus hal fam alsa ntpdate cups sensors athcool preload kdm)

I have tried to reach my wished daemon distribution:
DAEMONS=(syslog-ng @network netfs crond dbus hal kdm fam alsa ntpdate cups sensors athcool preload)

So, I have started moving 'preload' to the right as much as possible. Then, I have done the same thing with 'kdm' too. Next you have the further I have been able to reach that keeps everything working, i.e., using tty7 rather than tty3:
DAEMONS=(syslog-ng @network netfs crond dbus hal fam alsa ntpdate preload sensors cups athcool kdm)

If I move just single position to the left either 'preload' or 'kdm', tty3 is used and the keyboard issues come up again. I'm not sure whether or not this will shed any light on this issue, but it's as much as I can do.

I hope we find a solution this issue soon, because it's a pain in the neck.

Best regards! :-)
Comment by Kenni Lund (Kenni) - Thursday, 15 October 2009, 19:19 GMT
In case it isn't clear from previous comments, this issue doesn't seem to be related to ATI exclusively...I have the same issue on Intel graphics with the xf86-video-intel driver.

Therefore, the bug should be renamed making it easier for other people to spot it in the bugtracker - I skipped it to begin with, as it seemed to be ATI-related and I had Intel graphics.
Comment by Kenni Lund (Kenni) - Friday, 16 October 2009, 13:17 GMT
I can confirm the behaviour which Musikolo describes.

DAEMONS=(syslog-ng hal !network netfs crond wicd kdm bluetooth sshd)
-> X starts on tty3 and crashes when I open up Konsole and writes "192." a few times.

DAEMONS=(syslog-ng hal !network netfs crond wicd bluetooth sshd kdm)
-> X starts on tty7 and seems to be running perfectly stable.

No usable info from Xorg log or dmesg during crash :(

The only indicator that something went wrong, can be found in errors.log:
kdm[3438]: X server for display :0 terminated unexpectedly

xorg-server 1.6.3.901-1
Comment by Andreas Radke (AndyRTR) - Friday, 16 October 2009, 16:52 GMT
no idea what you all are talking about. if you start kdm in deamon mode and not from inittab it always should be the last one in the row. at least al modules have to be loaded long enough before to be initialized well.

is there still something to be fixed? what should the topic become?

btw: preload isn't supported for good reasons in Arch..
Comment by Kenni Lund (Kenni) - Friday, 16 October 2009, 18:11 GMT
> is there still something to be fixed?
Well, yes I suppose.

Placing KDM as the last item in DAEMONS works around the issue, but is it expected behaviour to get an unstable Xorg system when kdm is not the last in the DAEMONS-row, even though the only daemon following kdm is sshd (which should not interfere with X or keyboard input)? And why does it change to tty3 (which seems to be related to the system getting unstable)?

> is there still something to be fixed? what should the topic become?
It seems like KDE 4.3.2 was the trigger for the bug for most people, however, I don't know if it the cause of the bug or Xorg is.

Perhaps "[KDE] Xorg crashes randomly on keyboard input"?
Comment by Musikolo (Musikolo) - Saturday, 17 October 2009, 11:14 GMT
@AndyRTR
Upto KDE-4.3.1 I have had no problem at all using the next DAEMONS distribution:
DAEMONS=(syslog-ng @network netfs crond dbus hal kdm fam alsa ntpdate cups sensors athcool preload)

as you can see, kdm is loaded after the most important services are started up, but not the last one at all. For some reason, and this is the question of this thread, since KDE 4.3.2 the previous distribution does not work anymore. I have already posted the different behaviours corresponding to a varied series of tested configuration, but one common factor is that all dysfunctional settings use tty3 whereas all functional ones use tty7. This seems to be clearly bound to the place where the kdm daemon is set in the list and the only work-around is to start it up in the last position, like this:
DAEMONS=(syslog-ng @network netfs crond dbus hal fam alsa ntpdate cups sensors athcool preload kdm)

I haven't found any evidence that preload has anything to do nor a reason to stop using, as it is a Community package, presumably well supported. I haven't found any evidence that KDE itself is the culprit, it might be another package that was installed at the moment I upgraded to KDE 4.3.2, that taking into account I don't usually leave much time between updates; 1 each day or couple of days at the most.

Thus, everyone should agree in one thing: Houston, we have problem! What the reasons are, it's something one knows for the moment being, but as said, something must be done.

Thanks a million for your time and effort.
Regards.
Comment by Mario K (empanyc) - Saturday, 17 October 2009, 13:15 GMT

Personally, i think it's not a good approach to mix different bugs in a single bug report and neither changing the title afterwards.

Well, for me, the original bug reporter, this bug is _fixed_. With the latest kernel update, both PCs work well again, thanks to everyone involved!

Loading...