Post new topic   Reply to topic
View previous topic Printable version Log in to check your private messages View next topic
Author Message
braveheartleoOffline
Post subject: Multiple display corruption with 2.6.38-x kernels on 865g  PostPosted: 15.04.2011, 04:10



Joined: 2011-04-08
Posts: 17

Status: Offline
Ever since the Linux kernel hit 2.6.38, I have been experiencing display corruptions, mostly manifesting on texts when scrolling (texts become blurry, some have multiple strike-throughs in them). This is on an Intel 82865G graphics controller. The corruption is not permanent, and can usually be cleared by scrolling the text again, or highlighting said text. So far, the corruption only happens on text and not on other graphic elements, and has not caused serious faults (desktop freeze, kernel panic).

My workaround has always been to keep using the 2.6.37-0.slh.18-aptosid Linux kernel that came with the aptosid 2011-01 iso release.

I was wondering if somebody could have a hint or two regarding this odd behavior. Using the latest 2.14 xorg-intel driver release didn't help.
      Code:
Host/Kernel/OS  "linux" running Linux 2.6.37-0.slh.18-aptosid-amd64 x86_64 [ aptosid 2011-01 Γῆρας - xfce - (201102051540) ]
CPU Info        2x Intel Pentium 4 2048 KB cache flags( sse3 ht nx lm ) clocked at [ 2400.000 MHz ]
Videocard       Intel 82865G Integrated Graphics Controller  X.Org 1.9.5  [ 1280x720@59.9hz ]
Network cards   Marvell 88E8001 Gigabit
Processes 100 | Uptime 5:59 | Memory 545.6/995.1MB | HDD ST3160815AS,ST3320620AS Size 480GB (58%used) | Client Shell | Infobash v3.35
 
 View user's profile Send private message  
Reply with quote Back to top
spockOffline
Post subject:   PostPosted: 15.04.2011, 17:18



Joined: 2010-09-11
Posts: 80
Location: Near Milton Keynes, UK
Status: Offline
I'm also seeing this behaviour - it looks like this: (RightClick and View Image to enlarge screenshot)

http://aptosid.com/index.php?name=PNphp ... =7314#7314

It doesn't bother me too much, and I find it kind of quaint that these things happen yet the system still somehow works normally. Confused Often just scrolling or refreshing or moving a window will remove the artefacts. It's rather like the equally inexplicable KATE backwards typing error (Google that), and I just accept it and hope that one day someone wiser than me can figure it out. Although there was one really epic crash last week where the graphic artefacts took over and screen redraws failed completely. Really spectactular and looked very odd but I've no idea what caused it, and I lost control of X and had to reboot so I can't properly report a bug. I'm keeping my eyes peeled though, now I've learned how to take screenshots. Any advice on how to diagnose X crashes after they occur is welcome. /var/log/Xorg.log.* give no clues to my untrained mind.
 
 View user's profile Send private message Visit poster's website  
Reply with quote Back to top
slhOffline
Post subject:   PostPosted: 15.04.2011, 17:36



Joined: 2010-08-25
Posts: 732

Status: Offline
Unfortunately I can't reproduce that on i830, i945GME or i945GZ, which makes debugging rather hard.
 
 View user's profile Send private message  
Reply with quote Back to top
braveheartleoOffline
Post subject:   PostPosted: 15.04.2011, 20:17



Joined: 2011-04-08
Posts: 17

Status: Offline
      spock wrote:
I'm also seeing this behaviour - it looks like this: (RightClick and View Image to enlarge screenshot)

That's exactly how the corruption manifests in my display.

Apparently using the 2.6.37 kernel doesn't manifest this problem. However, other problems cropped up behind the scenes like consistent "intel_drm: EIR got stuck" in the logs, and another quickly spamming "page flipped multiple times" messages that at one point grew my syslog to a whopping 530M (the messages are approximations).

Other than those error lines I haven't experienced any other graphic problems. Using 2.6.38 no (EE) message lines in Xorg.log that could indicate potential problems in the driver/hardware, nor related errors in syslog.

This has been observed in both 32- and 64-bit installs.

      slh wrote:
Unfortunately I can't reproduce that on i830, i945GME or i945GZ, which makes debugging rather hard.

The report from spock in the linked thread happens to be with an Intel 915G. Preceeding that is another from apogarte who happens to be using an Intel 852/855GM. While I on the other hand am using Intel 865G.

Could it be that there are specific Intel graphics hardware affected, and until one is using any of the above and experiencing the same issues, can we confirm that the issue is reproducible?
 
 View user's profile Send private message  
Reply with quote Back to top
slhOffline
Post subject:   PostPosted: 15.04.2011, 20:25



Joined: 2010-08-25
Posts: 732

Status: Offline
The hardware range driven by i945.ko/ xserver-xorg-video-intel ist by no means uniform and differs massively; 12 years of graphics card development.
 
 View user's profile Send private message  
Reply with quote Back to top
muchan
Post subject:   PostPosted: 15.04.2011, 22:45
Moderator


Joined: 2010-09-11
Posts: 468

I encounter this problem with:
Intel Mobile 945GME Express Integrated Graphics Controller

In my case, it occurs on Xfce4-terminal console, Monospace font 9,
anti-aliasing is on.
I din't notice it occurs on other software, (but once... I forgot where)
 
 View user's profile Send private message  
Reply with quote Back to top
braveheartleoOffline
Post subject:   PostPosted: 16.04.2011, 01:14



Joined: 2011-04-08
Posts: 17

Status: Offline
      muchan wrote:
I encounter this problem with:
Intel Mobile 945GME Express Integrated Graphics Controller

In my case, it occurs on Xfce4-terminal console, Monospace font 9,
anti-aliasing is on.
I din't notice it occurs on other software, (but once... I forgot where)

I also notice it on xfce4-terminal, but not inside a login shell (i.e. not inside X).

The odd thing about this display corruption is that it only affects text displays. Movies play fine without manifesting artifacts.
 
 View user's profile Send private message  
Reply with quote Back to top
braveheartleoOffline
Post subject:   PostPosted: 16.04.2011, 01:57



Joined: 2011-04-08
Posts: 17

Status: Offline
      slh wrote:
The hardware range driven by i945.ko/ xserver-xorg-video-intel ist by no means uniform and differs massively; 12 years of graphics card development.

Then it might be some regression for the problem to occur in .38 when it didn't occur in .37.
 
 View user's profile Send private message  
Reply with quote Back to top
spockOffline
19 Post subject: Computer freaking out as I write this  PostPosted: 18.04.2011, 12:41



Joined: 2010-09-11
Posts: 80
Location: Near Milton Keynes, UK
Status: Offline
OK, sorry for the epic length post but I'm not sure what is going on here, so any hints on who to direct this bug report to are welcome...

      spock wrote:
Although there was one really epic crash last week where the graphic artefacts took over and screen redraws failed completely. Really spectactular and looked very odd but I've no idea what caused it, and I lost control of X and had to reboot so I can't properly report a bug. I'm keeping my eyes peeled though, now I've learned how to take screenshots.


OK, my machine has gone weird again, and then regained enough sanity to post this. Text+icons disappearing, changing colour when you scroll, all in no consistent way. I can CTRL+ALT+F1 and the console is unaffected, and I can get back to X again, but the corruption continues. Iceweasel and KATE have largely recovered, but Conky, Fluxbox, ROX-Filer and XFCE-Terminal have invisble text (apart from my login prompt).

I'll try and add screenshots in the next post if this works, but this may be hard since the text has disappeared from my Fluxbox taskbar, although I can ALT+TAB switch normally.

      spock wrote:

Any advice on how to diagnose X crashes after^H^H^H^H^H while they occur is welcome.


/var/log/Xorg.0.log is being flooded with 100,000s of messages, growing at roughly 1MB per minute! Almost every line is:

      Code:
[  7458.230] (WW) intel(0): intel_uxa_prepare_access: bo map failed: input/output error


With a very occasional line like this every few thousand:

      Code:
[  7458.230] (WW) intel(0): failed to set cursor: input/output error


I've seen logs get filled up like this before, and once it filled / so I now have /var on its own 18GB partition (also for other reasons...).

I Googled the first error message above, and see threads from the past few months, some also mentioning the dreaded "Hangcheck timer elapsed... GPU hung" which I used to get on my older Intel 82845 machine (now mostly retired from X).

https://bugs.archlinux.org/task/22604

This bug from Jan 2011 is top Google hit for the first error message, but I don't have KDM installed, and this machine had been running happily for a few hours before flipping out, although it now operates sort-of normally except I can't see text (and some icons) in some places. Other weird things like Iceweasel mouse pointer not changing to a finger when you hover over a link. I can't tell what CPU usage is at since my Conky readouts are gone. Perhaps this is related to recent udev problems, but I understand very little... Confused
 
 View user's profile Send private message Visit poster's website  
Reply with quote Back to top
spockOffline
22 Post subject: Re: Computer freaking out as I write this  PostPosted: 18.04.2011, 14:03



Joined: 2010-09-11
Posts: 80
Location: Near Milton Keynes, UK
Status: Offline
[Apologies for any unintentional thread hijack if this is actually an issue with Fluxbox.]

      spock wrote:
Iceweasel and KATE have largely recovered, but Conky, Fluxbox, ROX-Filer and XFCE-Terminal have invisble text (apart from my login prompt).

I'll try and add screenshots in the next post if this works


My original screenshots that I took in the midst of the chaos aren't particularly helpful, since they went a bit wrong.

Now the machine has sort of recovered, but not totally. Imagine a Fluxbox taskbar with icons but no text whatsoever. The Fluxbox menu still works when you right-click the desktop to run programs, although some menu items are invisble until you select them (see screenshot below). Window titles are visible when not in focus, but disappear when you focus on that window. This is like the text on websites changing sometimes to the background colour, sometimes to other odd colours.

One instance of KATE that was running changed to a blank grey background, so I closed it and re-opened it fine. New ROX-Filer windows still have no text though unless items are selected (see screenshot below). Iceweasel is now only missing a few favicons, although at one point even the File Edit View History menu texts went invisible; they've now returned without restarting. Websites have returned to their usual colourschemes. Tooltips work fine in Iceweasel but hovering over the Iceweasel tab in the taskbar yields a blank tooltip.

Dragging windows around the desktop was always strange in Fluxbox (seeming drunk), but now the delay is vastly increased, so dragging a window with rapid movements of the mouse takes ages to catch up, and top reports Xorg using 50% CPU.

Perhaps what I'm seeing now is an issue with Fluxbox, although this is the post-crash scenario so it may just be a result of the earlier weirdness in X. I leave it to Mods to decide whether to break out into a separate thread.

As you can see from my desktop background in this screenshot, we have some purple-haired people looking into this matter and hopefully they'll issue a fix soon...

      spock wrote:

Any advice on how to diagnose X crashes after^H^H^H^H^H while they occur is welcome.


Is there anything else I can try while it's in this state?

top reports that the machine is only 30% busy (running Iceweasel and a few other things). I was able to run GIMP to resize this screenshot, without any problems. I'll leave the machine alone for a while before rebooting (which cured it before) in case anyone has any bright ideas. We have a few hours before the log file (currently at 320MB and rising) grows to fill up the remaining 14GB space... Meanwhile I'll take it as an excuse to go and play piano instead Very Happy



PurpleHairedPeople-screenshot_750x600.png
 Description:
Fluxbox desktop screenshot showing menu and ROX-Filer only displaying selected text
 Filesize:  493.6 KB
 Viewed:  6252 Time(s)

PurpleHairedPeople-screenshot_750x600.png


 
 View user's profile Send private message Visit poster's website  
Reply with quote Back to top
bjoOffline
Post subject: RE: Re: Computer freaking out as I write this  PostPosted: 19.04.2011, 03:20



Joined: 2010-09-15
Posts: 3

Status: Offline
I have also problems with the intel-driver and the .38 kernel (slh-image and also sid-image), sometimes the desktop freezes except the mousepointer (the "Hangcheck timer elapsed... GPU hung" issue) With .37 this problem never appears.


Host/Kernel/OS "ostrea" running Linux 2.6.37-2-686 i686 [ sidux 2009-04 Μόρος - xfce - (200912310312) ]
CPU Info Intel Pentium M clocked at [ 800.000 MHz ]
Videocard Intel Mobile 915GM/GMS/910GML Express Graphics Controller X.Org 1.10.0.901 [ 1024x768@60.0hz ]
Processes 148 | Uptime 11min | Memory 311.6/2017.9MB | HDD Size 40GB (58%used) | Client Shell | Infobash v3.35
 
 View user's profile Send private message  
Reply with quote Back to top
Display posts from previous:     
Jump to:  
All times are GMT - 12 Hours
Post new topic   Reply to topic
View previous topic Printable version Log in to check your private messages View next topic
Powered by Zafenio