[RV250] Regression: Screen corruptions on X + massive RAM- & Swap-usage

I'm using Ubuntu 11.04 – since I upgraded to Natty Narwhal I see heavy graphics corruptions in Gnome 2.32.1, the Gnome panel is especially affected. I'm sure I already saw those corruptions in Ubuntu 10.10 (but I had the backports activated…). I'm sorry but I can't narrow this further down…

In addition I'm noticing increasing usage of my swap-partition. I have 4GB swap (don't ask me why) – within a few hours the graphics corruptions and the swap-usage increases until the whole 4GB are in use and the system freezes. I'm quite sure those symptoms are related…

I attached two screenshots showing different corruptions on Launchpad:

https://launchpadlibrarian.net/70632490/screenshot_stefan-nagy_01.png shows a permanent symptom only affecting context menus of the Gnome panel - in my case the gnome clock applet, the indicator applet and the applet used to switch between open windows using buttons (in german called 'Fensterwähler'). I *don't* see those corruptions when I open the context menus of the menu bar and the netspeed applet.

https://launchpadlibrarian.net/70632558/screenshot_stefan-nagy_02.png shows a symptom which appears after some time (an hour or more) and tends to get worse i.e. one Gnome panel part after another is affected.

I'd really appreciate any help with this, please tell me if I can provide additional information.

Bugreport on Launchpad: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/771370

Linux rosa 2.6.38-8-generic #42-Ubuntu SMP Mon Apr 11 03:31:50 UTC 2011 i686 i686 i386 GNU/Linux

01:00.0 VGA compatible controller [0300]: ATI Technologies Inc Radeon RV250 [Mobility FireGL 9000] [1002:4c66] (rev 01)

Please attach the full Xorg.0.log file.

Does

       Option "RenderAccel" "off"

work around any of these problems?

Created attachment 46260
Xorg.0.log

Thanks – option "RenderAccel" "off" seems to solve the problem, I tested this for some hours now and didn't see any corruptions, swap-partition usage looks normal… I'll let you know if I come across any corruptions again.

Regarding the problem shown in my first screenshot it seems like it has nothing to do with my graphics card – see https://bugs.launchpad.net/ubuntu/+source/light-themes/+bug/751226.

After testing this for 10 days now I have to say that I've seen some minor corruptions (can't say more about that at this point). But that's not the main problem… ram- and swap-usage aren't normal at all.

Normal ram-usage is now at 800 MB and this doesn't change so much over time – but at some point the swap-usage increases without any obvious reason. When it reaches 60% I usually restart gdm because the performance sucks… Once or twice in the last 10 days the x-server crashed – but I didn't observe the ram- and swap-usage… :-/

All in all it's possible to work with option RenderAccel turned off, but the performance is quite bad at the beginning and gets even worse over time… so this is no long-term solution.

After several weeks of testing this I have to correct myself. I don't really see any corruptions anymore & the ram & swap-usage seems to be quite normal. After all I'm working every day for several hours with this computer & never had any problems (since my last posting). Sure, the performance is much worse with option RenderAccel deactivated, but apart from that everything works fine.

Please tell me if I can help in any way to get this fixed. I'd love to use my old notebook for at least another decade & the increase in performance with RenderAccel turned on is just amazing… ;)

Created attachment 50096
screenshot showing corrupted graphics (option "RenderAccel" is "off")

Sorry, I have to correct myself again (I hope this is the last time). After writing the last comment I kept an eye on both symptoms (massive swap-usage, graphic corruptions) and now I can say that both symptoms are still present with option "RenderAccel" "off" – up to a moderate extent…

· swap-usage
When RAM-usage reaches around 40%, swap-usage begins to increase continuous. For example when I open a window (nautilus, gedit… whatever) swap-usage increases, closing windows/programs doesn't have the effect of decreasing swap-usage. Even when I change workspaces swap-usage increases. After some hours of work I'll log out and in again to empty swap & get the normal performance of my system again.

· graphic corruptions
I see some minor corruptions which are quite stable. For example the graphic you see in the middle of the screen when you change workspaces in Gnome looks crappy quite often and when it started once it will stay like this (I attach screenshot 1 showing this corruption). I'll also attach another example, I had to close the program to get rid of the corruption (see screenshot 2).

In Summary: option "RenderAccel" "off" doesn't solve any of the problems I'm having with this graphics card but it extenuates the symptoms to an extend so I can work with the computer.

Created attachment 50097
2nd screenshot showing corrupted graphics (option "RenderAccel" is "off")

After I tried out several radeon options I found a better solution than deactivating Render Acceleration: forcing the card into PCI mode (by setting AGP mode to -1). Performance is much better, I can see no corrupted graphics at all & I have no problems with massive RAM & swap usage.

The performance is inferior to AGP mode 4x, 2x and 1x, but apart from that it works great.

Could someone please take a look at the screenshots of the original reporter of Bug 771370 in Launchpad & mine – I suppose although we have the same graphics chip (with PCI-ID 1002:4c66) we are affected by different bugs; as far as I can tell the corruptions look quite different…

Besides I have a question: given the above mentioned workaround (forcing AGP to PCI mode while render acceleration stays activated) does that mean that VRAM to AGP transports could be the problem?

AGP is always the problem. There are too many wonky AGP chipsets out there.

Thanks for your reply! It seems that in the meantime I've found an even better way to avoid these symptoms: Option "EXAPixmaps" "off" works great for me (in AGP Mode 4x) – performance is much better now.

I asked the original reporter of launchpad bug #771370 if he could test the different radeon options suggested here & it turned out he doesn't see any corrupted graphics anymore – so it seems he was affected by another bug.

FYI: My original report on launchpad is bug #766693
https://bugs.launchpad.net/xserver-xorg-driver-ati/+bug/766693

Regarding my experiences using Option "EXAPixmaps" "off": It still works great, I just saw a minor screen corruption once; ram & swap usage look quite normal – but I have to test this over a longer period and report back later.

I tested Debian Stable (Wheezy) for some days & the graphics performance was just great (without any radeon-options) – the only thing I had to do was to install the firmware-blob (included in firmware-linux-nonfree).

I didn't see any screen corruptions in Squeeze and the performance was really incredible – so maybe this is a regression?

I tested this on Ubuntu 10.10 (maverick) for some time now and couldn't reproduce this bug anymore, so I closed my original report on Launchpad and opened a new one (see https://launchpad.net/bugs/860509). The only bug I'm affected by in Debian 6.0 & Ubuntu 10.10 (Xorg 7.5) with my graphics hardware is this one: https://launchpad.net/bugs/539163 (same symptoms as the original reporter).

The bug described here is new in Ubuntu 11.04 (Xorg 7.6). Since I see screen artifacts not only when I use Compiz but also with Metacity I'm going to change the title of this report. I decided to keep Ubuntu 11.04 installed for now so I can test this – I'd be glad to help if someone needs more information on this.

I'm testing this now with xorg packages from the ubuntu xorg-edgers PPA ("Xorg packages fresh from git") & the problem is still present.

BTW: Seems like the corruptions accumulate much faster when I have my external monitor connected (VGA-0).

I can't reproduce this bug on Debian Testing (Wheezy) – which really confuses me…
-> Xorg 1:7.6+9, xserver-xorg-video-radeon 1:6.14.2-2.

Maybe this information could help to narrow this bug down.

(please note that I *didn't* test this on wheezy before; in comment #12 I just mixed up the codenames "Squeeze" and "Wheezy" – sorry for the confusion)

I tested this on Ubuntu 12.04 now and since I can't reproduce the bug anymore I'm closing this report. Thanks.