Bug 230123

Created attachment 110610 [details]
hwinfo of the graphic card

Created attachment 110611 [details]
hwinfo of the monitor done at runlevel 3

Created attachment 110613 [details]
xorg.conf

Created attachment 110614 [details]
Xorg.0.log

Created attachment 110615 [details]
Version of the x-server, x-video seams not to be installed

Hmm ...

Ok. This issue needs to get fixed upstream. I suggest to check the latest
driver from time to time by updating xorg-x11-server/xorg-x11-driver-video
packages from the xorg73 project.

  http://software.opensuse.org/download/xorg73/openSUSE_10.2/

Closing as LATER.

I don't expect this still to be fixed in "i810" driver, so better switch to new "intel" driver in xorg.conf for testing.

Sorry for asking, but which other driver would be the most appropriate?

"intel".

Thanks for the information - unfortunately this doesn't work. When I put "intel" in the xorg.conf the X-server doesn't start telling me that the module wasn't found. modprobe intel gives the feedback that there is no module named "intel" and sax2 doesn't offer the possibility to select another driver (as in older versions). What can/must I do in order to get this driver?

Obviously you didn't update the X.Org RPMs mentioned in comment #7. "intel" is a X.Org module, not a kernel module.

Sorry, you're right - forgot it. Well, after the update I first tried the driver 'intel' but the screen began to flicker. Than I configured the x-server once more with sax2. Now suspend to ram works! Thanks! 

reopen to close as fixed

fixed with new X.Org, i.e with openSUSE 10.3.

Sorry, but the it seems as if the bug is back: Yesterday when I updated several packages unfortunately I also did an update of xorg from http://software.opensuse.org/download/xorg73/openSUSE_10.2/
(xorg-server 7.2-250.4, xorg-x11-driver-video 7.2-243.1)
After resume from s2ram the windows of the opened programs are transparent as before (with xorg 7.2). Is there a repository of xorg 7.3 where older version are available, especially the one which was the last one at the beginning of april?  

Unfortunately we don´t have older repos available.

JFYI.

This is very sad...
Would it be worthwhile to try to compile the xorg-server and the xorg-x11-video-driver myself in order to find out if a self-compiled server works better?
In case it would pay, where should I try get the sources?

I don't think this would help, since we don't know which sources worked for you. :-(

I think the version must be 7.2-152.2 (The first update of xorg I did on april 13. At this day I also updated my desktop Pentium 4 and on this machine I still did not update to the newest xorg-version, but unfortunately I left the yast-update-mechanism on default, which means, that all rpm-packages of each update are deleted after installing. Could it be a possibility to copy file by file of the to packages xorg-server and xorg-video or won't they work with the rest of the xorg-system?

Not sure if I understand you correctly. I thought the RPM has been deleted and is no longer installed? So how do you want to copy it file by file?

Stefan, I think Michael suggests to copy the files over from another machine which still has the version that worked for him.
Michael, it should be possible as long as the ABI hasn't changed. You could try and copy just the driver that you are using. Best you do this in runlvl 3. Be sure to make a backup copy of the old driver.
To test you can just do type 'X' as root which will bring up the bare server without any clients. You can kill it with ctrl-alt-backspace if it comes up.
Otherwise it will be more difficult as you will also have to copy the Xorg binary (/usr/bin/) and several modules (/usr/lib/xorg/modules) - or better the entire
directory.
Stefan, what have been the changes since early April?
It looks like this is a problem with an uninitialized 2D engine. 
Michael: To verify you can set the Option 'noaccel' in /etc/X11/xorg.conf.

Sorry, but what is the ABI? (with google I only found this: The Linux ABI is a patch to the Linux kernel that allows a linux system to run foreign binaries.)
I suppose, I should set the option 'noaccel' after having copied the driver or should I try this with the driver of the last update still installed?

> Stefan, what have been the changes since early April?
A lot. See "rpm --chnangelog -q xorg-x11-server", "rpm --changelog -q xorg-x11-driver-video". Among the changes are

- update to xserver 1.3.0.0 release
- update to Mesa 6.5.3
- update to intel driver release 2.0.0

ABI -- > http://en.wikipedia.org/wiki/Application_binary_interface

Please test the "noaccel" option with the driver of the last update still installed.

Still with the newest driver I tried the noaccel option in the screen section and with this s2ram works correctly.
Then I copied -as you suggested in runlevel 3 - the i810-driver from my desktop-machine (usr/lib/libI810XvMC.la, usr/lib/libI810XvMC.so.1.0.0, usr/lib/xorg/modules/driver/i810_drv.la and ../i810_drv.so) and restored the original xorg.conf (without the noaccel option) but this doesn't work. The x-server starts but even after restarting the machine the desktop appears with transparent windows. So it seems that there are major changes in the xorg-system. Don't know if it pays to try to copy the whole bunch of files of the server and the video-driver or is there a chance to get this fixed in one of the next updates? 
Is there something else what I could try or test?
BTW: Thanks for the hint about ABI.

You need to copy the intel_drv* and not the i810_drv* files since you're currently using the intel and not the i810 driver.

No, I don't think so. Please see comment #13 - after I tried 'intel' without luck I used sax2 to configure and this restored the i810-driver and it worked...(until the last update of xorg).

i810 driver didn't change since a long time. So it must be something in the Xserver in general. So what's the latest changelog of the xorg-x11-server package on the machine, on which you think you still have installed the working Xserver?

rpm --changelog -q xorg-x11-server

Created attachment 140166 [details]
Changelog of xorg-x11-server from april 2007

Changes since this time:

-------------------------------------------------------------------
Wed May  2 22:39:31 CEST 2007 - sndirsch@suse.de

- commit-c09e68c:
  * Paper over a crash at exit during GLX teardown

-------------------------------------------------------------------
Mon Apr 30 22:21:38 CEST 2007 - sndirsch@suse.de

- updated to Mesa 6.5.3 sources
- obsoletes the following patches:
  * bug-211314_mesa-destroy_buffers.diff
  * bug-211314_mesa-framebuffer-counting.diff
  * bug-211314-patch-1.diff
  * bug-211314-patch-2.diff
  * bug-211314-patch-3.diff
  * bug-211314-patch-4.diff
  * bug-211314-patch-5.diff
  * bug-211314-patch-6.diff
  * bug-211314-patch-7.diff
  * bug-211314-patch-8.diff
  * bug-211314-patch-9.diff
  * bug-211314-patch-10.diff
  * bug-211314-patch-11.diff
  * bug-211314_mesa-refcount-memleak-fixes.diff
  * Mesa-6.5.2-fix_radeon_cliprect.diff
- remove__GLinterface.patch/
  support_mesa6.5.3.patch
  * required Xserver changes for Mesa 6.5.3

-------------------------------------------------------------------
Sat Apr 28 09:54:28 CEST 2007 - sndirsch@suse.de

- xorg-x11-server-1.2.99-unbreak-domain.patch:
  * This patch fixes some multi-domain systems such as Pegasos with
    xorg-server 1.3. Since pci-rework should get merged soon and
    this patch is a bit of a hack, it never got pushed upstream. 
    (X.Org Bug #7248)

-------------------------------------------------------------------
Fri Apr 20 07:11:16 CEST 2007 - sndirsch@suse.de

- xserver 1.3.0.0 release
  * Syncmaster 226 monitor needs 60Hz refresh (#10545).
  * In AIGLX EnterVT processing, invoke driver EnterVT before 
    resuming glx.
  * Disable CRTC when SetSingleMode has no matching mode. Update 
    RandR as well.
  * Rotate screen size as needed from RandR 1.1 change requests.
  * Add quirk for Acer AL1706 monitor to force 60hz refresh.
  * RandR 1.2 spec says CRTC info contains screen-relative geometry
  * typo in built-in module log message
  * Use default screen monitor for one of the outputs.
  * Allow outputs to be explicitly enabled in config, overriding
    detect.
  * Was accidentally disabling rotation updates in mode set.
  * Disable SourceValidate in rotation to capture cursor.

Once more after mid-air-crash:
Please find attached the changelog - and I yes, I'm pretty sure, that this is
the version which worked fine on the laptop. After I succeeded with the laptop
I integrated the update-url for xorg7.3 in my desktop-machine and updated this
one. As you see from comment #13 I was lucky on friday 13th, and rpm -qi
xorg-x11-server on my desktop says Sat Apr 14 00:03:50 2007 as installation
date...

1) You could copy 

/usr/lib/xorg/modules/extensions/libGLcore.so
/usr/lib/xorg/modules/extensions/libglx.so

from the old version to see if it's related to the Mesa udpate. It is not
if using the old files doesn't help. 

2) I can build you a version which only includes the update to xserver 1.3.0.0 release.

I suggest to begin with option 1. :-)

Following your advice I just tried option 1:
On the desktop I found libGLcore.so* only in /usr/lib and but none in /usr/lib/xorg/modules/extensions, but on the laptop I found it in /usr/lib/xorg/modules/extensions.
So I did the following:
From the desktop I copied libglx.so.1.0.9755 to the same directory on the laptop and then changed the link libglx.so to this file.
From the desktop I copied libGLcore.so.1.0.9755 to /usr/lib and I changed the symbolic link libGLcore.so.1 to this file. Next I renamed (on the laptop) /usr/lib/xorg/modules/extensions/libGLcore.so and the I did ln -s /usr/lib/libGLcore.so.1.0.9755 /usr/lib/xorg/modules/extensions/libGLcore.so
(all in runlevel 3)
Then I tried this. Freshly rebooted the x-server works correctly but after resume from s2ram the windows changed to transparent. It seems that I have to keep with the the noaccel-option... :-(
So if it could be the solution it would be nice if you could build the version you mentioned

Well. Now you begin to completely mess up your system with files from a NVIDIA driver installation, trying to use libraries as Xserver modules, etc. I'm not sure if I can fix again the system with building a new package. :-(

I suppose you are talking about the two *.9755 files. I didn't know that these were from the nvidia installer. Nevertheless after my try I set back the laptop in the original state by deleting the two copied files, renaming the original files and setting back the links as before. 

Ok. I'll give you a package for testing soon.

--> ftp://ftp.suse.com/pub/people/sndirsch/RPMS/bug230123

(in some hours, need time for syncing)

Thanks! I suppose it's xorg-x11-server7.2-84.i586.rpm what must install. I'm not sure if I can give a feedback till tomorrow as I have to prepare luggage. I have to travel to Venezuela - unfortunately not for vacation but for work.I will be back on 31/05.

Well, I tried the xorg-x11-7.2-84.i586.rpm package, but without luck. So I left the laptop with the updated xorg-packages (7.2-250.2). In order to use s2ram during my stay in Venezuela I worked with the noaccel-option in xorg.conf.
Is there a chance that with OSS 10.3 s2ram will work? As I also had problems with s2disk the guys who manage bug # 230668 ask me to try the s2disk with the new kernel. At the same time I could try the s2ram, too.

Ok. So the update to Xserver 1.3.0.0 (massive changes) broke s2ram for you.

You know http://en.opensuse.org/S2ram ?

See my previous comment #44. Did you try all s2ram options described on this website?

Yes, I did  them all with different results - the only common one was that there was no correct resume: with some combinations I still could use the keyboard, with  others the machine froze completely, only -s -f worked. 

> only -s -f worked.
But this is now no longer working, correct? 

In combination with the NoAccel-Option in xorg.conf it still works. But if you think it's worthwhile to try all the other combinations once more I'll do it.

Ok. noaccel is no solution. Testing all other combinations would take you some hours I'm afraid.

Since you've updated to a newer Xserver and Xorg driver version and we'll use "intel" driver as default anyway for openSUSE 10.3, could you verify if s2ram works with the "intel" driver? Replace "i810" with "intel" in your xorg.conf to switch to the new driver.

About the "intel" driver see my comment # 13 (13.04.07): My laptop didn't work with this driver. As I wrote, after that unsuccessful trial I configured xorg once more with sax2 and after this set the configuration back to i810.
Nevertheless, if there have been changes with the "intel"-driver since my last trial I could try it once more.

Sorry, forget about this. But many issues have been fixed in intel driver since 13.04.07. So I think it would be worth a try.

Well, I tried the following:
1. I changed >Driver "i810"< to >Driver "intel"< in xorg.conf, but there is no change on my machine: it doesn't work with this driver. As before the screen remains dark - slightly darker the right half of the screen, and on the left side one can see it flicker.
2. I tested once more the different options with s2ram (as indicated at http://en.opensuse.org/S2ram). The behaviour of the notebook hasn't changed since the last try:
At init=/bin/bash the only option which works is s2ram -f -s. I tried all listed on the site. There is one thing I couldn't try: In the text you find an indication to use the -v option, but the installed version of s2ram doesn't know this option inspite of having installed the newest version (pm-utils-0.20.0.20061114, suspend-0.50-6).
I tried all combinations: With some the keyboard still worked, with others not even that. As I said with s2ram -f -s the notebook comes back and it is possible to work but there are some strange shadows on the screen. (When I do a s2ram from kde (with NoAccel) and after resume change to terminal (Ctrl-ALT-F1) the screen doesn't have this shadows which appear after a resume from init=/bin/bash. 
(Collateral damage: During this various trials something went wrong with the root-account and/or the identification of the laptop: I can open Yast in kde but when I try to open the software-installation part, there is no reaction. In a terminal YaST seems to work - I can get to the software-installation but in the upper right corner where should be YaST @ mymachine I find YaST @ noname - and I don't know how to fix this...)

Ok. Thanks for giving it a try.

As I had to try s2disk with the 10.3 Alpha version (see bug 230668) I tried s2ram too. It's all the same as with 10.2: The intel-driver doesn't work. The automatic configuration wrote 'intel' in the xorg.conf. With this the x-server doesn't work,
so I had to put manually i810 in it - fortunately this still works. Should I open a new bug about that? I still couldn't try all the posible combinations for the s2ram-options but perhaps within the next days I'll find the time...

>I still couldn't try all the posible combinations for the s2ram-options but >perhaps within the next days I'll find the time...
Any news on this?

Sorry for the delay. Now I tried this too. The result is very similar to that on OSS 10.2. The difference is that the OSS-10.3-version of s2ram knows the option -v. Now there are two combinations which work:
s2ram -f -s
s2ram -f -s -v
but this doesn't make any difference when I try to suspend from init 5.
(BTW: I couldn't find the config-file in /etc/pm, can you tell, where I can find it)

Ok. It seems s2ram works for you now with the intel driver. So what are the remaining issues?

I don't know what happens to /etc/pm.

pm-utils configuration changed slightly. See http://en.opensuse.org/Pm-utils#Configuration

> Ok. It seems s2ram works for you now with the intel driver. 
Sorry, but this is a misunderstanding: NOT on the OSS 10.2 NOR on 10.3 installation the intel driver works. On both installation I had to set the xorg.conf manually to i810 driver in order to the x-server working. Please see my comment # 55. What I described with comment # 57 was testing s2ram at init=/bin/bash on OSS 10.3 Alpha 5.

> So what are the remaining issues?
- The intel driver doesn't work (Saturday I posted a bug on the xorg buglist).
- After resume from s2ram the x-server only works correctly with i810 when the option NoAccel is set in xorg.conf.

> Remaining issues:
> The intel driver doesn't work (Saturday I posted a bug on the xorg buglist).
Thanks. I remember. Duplicate of Bug #289088

> After resume from s2ram the x-server only works correctly with i810 when
> the option NoAccel is set in xorg.conf.
Ok.

LATER again.

openSUSE 11.0 Beta1 is planned for this week. In case you're considering to test again with this Beta release, whether this issue still occurs, please give me feedback about your results. Otherwise please let me know, so I will close it as LATER again and reopen it once openSUSE 11.0 final is available for testing.

Ok, I'll give it a try report the results here.

Thanks. Please try with intel driver. There has been no changes in i810 driver since openSUSE 10.3.

Looks like testing the Beta release is no option here.

Oops.

First attempt: With an iso-image for network-installation. I was unable to install it because the configuration for the network-interfaces fails: The wlan didn't get the connection and when I tried to setup the lan (with static IP-Address) the setup-programm tells me (incorrectly) that the IP-Address is in use by another machine.
Second attempt: With the kde-live-cd. The system starts and X is configured automatically with the i810-driver. Therefore I changed the xorg.conf driver to 'intel' and restarted the x-server. The x-server works with this configuration. Then I tried suspend2ram.  The  machine resumes afterwards but with a black screen. 
Now I am downloading the dvd-image and tomorrow will give it a try, too.

What does comment #66 mean? Is it useful that I keep trying?

Please forget about comment #66. My fault. Sure it's useful that you keep trying. And please try intel also with 'Option "noaccel"'.

Created attachment 210978 [details]
automatically created at installation

Created attachment 210979 [details]
xorg.conf used for the different trials

Created attachment 210980 [details]
difference between live-cd xorg.conf and xorg.conf created by installation routine

1. Attempt: I downloaded the dvd-image and installed OSS 11 (very fast!) The x-server was set up automatically (see xorg.conf_i810) with i810-driver. So I changed the driver in xorg.conf to 'intel' and tried to restart the x-server but without success, the x-server crashed and the screen remained black. So started once more with the live-cd in order to look for the difference. I compared them with:
diff xorg.conf xorg.conf_live
You'll find the result in the attached xorgdiff.txt
2. Attempt: I changed the xorg.conf setting the default depth to 16 and cut of the lines between
Option       "NoDDC"
and
Option       "SaXDualOrientation" "LeftOf"
After this on booting the machine gets up a working the x-server. Without any change to /etc/pm I then tried s2ram and it worked once.The laptop went asleep and after pressing a key it came back with a working x-server. The only strange observation is that on selecting the s2ram-option from kpowersafe without having connected nothing to usb/firewire I got a message telling me that external media couldn't be unmounted because DCOP Interface couldn't be called.(The original message is in German but I hope you know which message is meant.) At a second try the x-server didn't come back, the screen remained black. I rebooted and tried it once again but this time after resume the x-server came back with a white screen without mouse and a black frame of the kwrite-window which I left opened before s2ram.
Then i copied the config file from OSS 10.2 /etc/pm to  OSS 11 /etc/pm/config.d - contains the option S2RAM_OPTS="-f -s" but it had the same result, the first s2ram works but at the second s2ram the screen remains black.
As a last step I removed the config file from /etc/pm/config.d and put the "NoAccel"-Option in the screen-section of xorg.conf (still the one with the intel-driver) but with this configuration not even the first try with s2ram works. On resume at the left side of the screen you can see running vertically something like a compressed image of the screen. After Ctrl - Alt - Backspace the x-server comes back correctly but I have to login once more... 
Is there something more I can try?

Ok. This sounds, that the experiences with openSUSE 11.0 are even worse than with 10.3. An easier way to use the intel driver for configuration would have been to run 'sax2 -r -m 0=intel'. But I don't expect the result would have been much different to your manually edited xorg.conf.

So all in all with openSUSE 11.0 you still need the i810 driver with 
'Option "noaccel" ' set. :-(

openSUSE 11.0 has been released today. Please give it a try and let me know, if the issue is fixed now. Thanks.

needinfo

About a week later. In case you give openSUSE 11.0 a try and the
issue still exists, please provide feedback and reopen. Thanks.

Sorry for the delay - too much work, I had no time left. I installed OSS 11.0 and first tried suspend out of the box. The automatic configuration after installation set the graphic driver to i810 Result: The suspend works correctly but the resume ends up at a black screen. Then I changed the driver to 'intel' and tried this. Result: The laptop suspends but the resume doesn't work. It seems that the resume-routine tries starting the x-server again and again without success. This means that after the beginning of the resume there is a black screen, then for some seconds apears the mouse pointer on the black background, then the screen gets black and empty once more, then the pointer appears again... and so on.
Last try: in the config-file at /etc/pm/config.d I added the line
S2RAM_OPTS="-f -s" (still with the intel driver) but with this even the suspend doesn't work. The laptop goes to sleep like an anxious child: with the lights turned on, e.g. the backlight of the screen keeps switched on. No resume posible (no reaction at any key stroke). Do you think, it would be worth to try some other combinations?