[X1950] Vertical bands of interference around HW cursor

[myxiplx]

After changing monitors, I've removed the proprietary ATI drivers and switched to the open source ones to allow me to run the monitors at different resolutions. Since doing this I'm noticing bands of graphical interference around the mouse cursor.

It looks to me like the interference bands are 20 pixels wide, and repeated every 250 pixels across the screen, on both screens. Locations I've measured (very approximately) are:
1770 - 1989
2025 - 2045
2290 - 2299
2540 - 2559

With the mouse cursor in any vertical position on one of these bands, a bunch of small flickering horizontal lines appear within the band, running from the right hand edge of the mouse cursor, to the right side of the band.

The lines don't move around, with the mouse cursor in any one spot you get 2-3 lines that just constantly flicker between black and white.

This is on a ATI 1950 (PCI-e), with the first screen running at 1280x1024, and the second 1680x1050. I'm running Ubuntu 8.10, and apart from the Intel graphics driver update released today, it's fully patched and up to date.

[AlexC (alex-j-crowe)]

Created an attachment (id=21347)
Pic of corruption, edited to illistrate the area affected

[AlexC (alex-j-crowe)]

Created an attachment (id=21348)
Pic of corruption, second type, single pixel

[Ancoron Luciferis (ancoron-luciferis)]

I have exactly the same issue here.

To me it appears that this issue only appears on the second display and it mostly takes some time until this issue appears. At least for me from the beginning all things are fine.

I sometimes use KDE4 or XFCE using Ubuntu 8.10 and when using XFCE it appears within minutes whereas in KDE it take several hours of work until the corruption appears.

All methods to get that corrupted mouse pointer captured failed and in framebuffer mode it doesn't seem to appear, so it is likely to be a problem in the driver.

In addition to the corruptions that Alex mentioned I have an additional corruption with the mouse pointer itself. When I see that corruption the cursor can sometimes be corrected by moving the mouse into the first display and back again.

I'm using a RV515 (ATI Radeon X1300) although the driver complains in the Xorg.0.log that DRI may not work correctly for R600 cards. (???)

[Ancoron Luciferis (ancoron-luciferis)]

Created an attachment (id=22182)
behavior of "type 1" corruption

This is an exact reproduction of the "type 1" mouse pointer corruption as I can see it.

[Ancoron Luciferis (ancoron-luciferis)]

Created an attachment (id=22183)
"type 1" and "type 2" areas on my dual-head

I managed to exectly find out at which positions the corruption areas are on my system. I created that image (2560x1024) because there may be some mathematical logic in there.

[Ancoron Luciferis (ancoron-luciferis)]

Created an attachment (id=22184)
"type 3" corruption

I am experiencing this corruption in addition to the other two ones which always come together. This one seem to appear a bit later and it is not restricted to areas.

Sometimes it is only visible on the second display but after some time it also gets to the first display.

[myxiplx (myxiplx)]

After changing monitors, I've removed the proprietary ATI drivers and switched to the open source ones to allow me to run the monitors at different resolutions. Since doing this I'm noticing bands of graphical interference around the mouse cursor.

It looks to me like the interference bands are 20 pixels wide, and repeated every 250 pixels across the screen, on both screens. Locations I've measured (very approximately) are:
1770 - 1989
2025 - 2045
2290 - 2299
2540 - 2559

With the mouse cursor in any vertical position on one of these bands, a bunch of small flickering horizontal lines appear within the band, running from the right hand edge of the mouse cursor, to the right side of the band.

The lines don't move around, with the mouse cursor in any one spot you get 2-3 lines that just constantly flicker between black and white.

This is on a ATI 1950 (PCI-e), with the first screen running at 1280x1024, and the second 1680x1050. I'm running Ubuntu 8.10, and apart from the Intel graphics driver update released today, it's fully patched and up to date.

[Tormod Volden (tormodvolden)]

Thank you for taking the time to report this bug and helping to make Ubuntu better. Please attach your X server configuration file (/etc/X11/xorg.conf) and X server log file (/var/log/Xorg.0.log) to the bug report as individual uncompressed file attachments using the "Attachment:" box below. Thanks in advance.

[Tormod Volden (tormodvolden)]

It would be nice if you could take a picture and attach it.

Does it help to use:
 Option "SWcursor" "true"
in the Device section of your /etc/X11/xorg.conf?

[myxiplx (myxiplx)]

Tried to take a screenshot, but it didn't show up. Yes, setting the swcursor option does seem to fix it. Does that make it a graphics driver bug?

[Tormod Volden (tormodvolden)]

Yes, that's bug in the driver. Does it happen when you use only one screen?

[myxiplx (myxiplx)]

Hey there, sorry, took a while to get around to testing this. No, it's fine if I reboot with just one monitor connected, but comes back immediately when I connect both and reboot again.

To give a little more detail in case it helps, the monitor connected to the first output of the graphics card uses a DVI connection, the second monitor is a VGA one using a DVI-VGA adapter, so you have a mix of analog and digital being used here.

[Tormod Volden (tormodvolden)]

I think I have found this reported upstream at http://bugs.freedesktop.org/show_bug.cgi?id=19215 - can you please check if your issue is the same?

[myxiplx (myxiplx)]

I've been hit by this too, but had reported it on the Ubuntu site:
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/320771

With help from there we've found that:
- This only happens with the free driver (ATI's driver is fine)
- It only happens with two monitors connected (one monitor is fine)
- It only happens with the hardware cursor SWcursor=true removes the problem.

It looks to me like the interference bands are 20 pixels wide, and repeated every 250 pixels across the screen, on both screens. Locations I've measured (very approximately) are:
1770 - 1989
2025 - 2045
2290 - 2299
2540 - 2559

It may also be worth knowing that this issue only occurs on my home computer, not at work, with the main differences between them being:

Home: ATI 1950 Pro, 1x DVI monitor @1280x1024, 1x VGA monitor @1680x1050
Work: ATI 2600XT, 2x DVI monitors @1920x1200

[myxiplx (myxiplx)]

Yes, that's it. Absolutely identical symptoms - cursor corruption in vertical bands on dual screens.

[Michael Sparmann (theseven)]

Same issue here, I get flickering black and white lines when the cursor is in those approx. 20 pixel wide corruption bands on both screens, with the following setup:
ATI Radeon Mobile X1400, Internal LCD @1680x1050, VGA monitor @1920x1200

[Mhopf-suse (mhopf-suse)]

Bug 13405 is a duplicate, only this time it's on the radeonhd driver.

[Bryce Harrington (bryce)]

Thanks.

[Stephen-boddy (stephen-boddy)]

Something that might be relevant: The corruption appears to be visible when the cursor spans a strip that is the last 10 or so pixels of the stride value given here:
http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/tree/src/radeon_cursor.c#n204
i.e. pixels where x=245 to 255. This could be expressed as every (256*n-10) to (256*n) where n>=1. If you use a jumbo sized pointer you can see that the strips of cursor on either side of the range given are OK.

There is an additional corruption which is seen as the cursor vertically bouncing. This occurs at 204+(256*n) where n>=0.

I see both symptoms on both screens with an X1400 Lenovo Z61m using LVDS (1680x1050) and CRT (1280x1024). According to xdpyinfo these are treated as screen #0 @ 2960x1050 pixels.

This was with 6.11.0.99+git20090305.5dc4b69f-0ubuntu0tormod~intrepid package.

[Ancoron Luciferis (ancoron-luciferis)]

I'm currently also experiencing corruptions type 2 and 3 in an irregular basis across two screens using the proprietary fglrx driver (8.54.3) with a Radeon HD2400 PRO PCI-E. It is not really a problem because it appears and then it takes a few minutes and the corruption is gone completely.

This is rather unusual and I didn't had such issues previously using Ubuntu 8.04 but now I did a fresh install of Kubuntu 8.10. In addition it indicates that this may not be a bug in the xrandr implementation as stated in some other bug as the fglrx 8.54.3 doesn't use xrandr.

Machine data:
- AMD Athlon 64 X2 4400+ Dual-Core @ 2.3 GHz
- ATI Radeon HD 2400 PRO (RV610, PCI-E)
- 2x 19" CRT @ 1280x1024@85Hz each

Software stack:
- Kubuntu Intrepid Ibex (8.10)
- Kernel 2.6.27.11-generic x64_64
- X.org 7.4.2.0
- fglrx 8.54.3.2 (BigDesktop mode)
- KDE 4.1.3

All drivers/software from the repos as I couldn't get the newer fglrx drivers to work with my two monitors. Xrandr implementations doesn't seem to be stable yet. In addition xrandr implementation in the drivers often use modes that are not appropriate for CRT's.

Might this be a result of shared code between the proprietary driver and the free ones? I know that AMD already pushed a lot of code out to the world for free use and this code might be buggy. As these corruptions can be seen regardless of the Hardware actually used (or at least in a range of chips R5xx-R6xx) and regardless of the driver (+version) used.

Can we also be sure that this involves only 64bit versions of the driver(s) used?

[agd5f (agd5f)]

I'm pretty sure the cursor issues are related to the watermark setup for the displays. Unfortunately, I haven't yet gotten any info on how to properly program them, but I got some new leads today.

The open drivers do not share any code with the proprietary drivers.

[Ancoron Luciferis (ancoron-luciferis)]

OK then, if driver code is completely separately developed but the same corruptions can be seen this bug is only partly related to the drivers then (as far as I suspect as I don't know the meaning of "watermark setup" exactly, as I'm just a lazy web developer).

:-)

[Kevin Puetz (puetzk)]

Ancoron: I am seeing this on my Macbook 1,1 (Core Duo), so it's not specific to 64-bit.

I'm running Ubuntu 9.04 beta (radeon 6.12.1, x.org, 1.6.0, kernal 2.6.28) with a radeon M56P (mobility X1600).

I'm also seeing the corruption on both screens, apparently unlike some others. If I disable one screen, or set them to clone rather than left-of/right-of, it clears up, but when it's happening both outputs are affected.

[Ancoron Luciferis (ancoron-luciferis)]

Two weeks ago I upgraded to KDE 4.2.0 from the Kubuntu testing repos plus a newer kernel and I've yet to see corruptions type 3. Also I have no input freezes anymore.

                   Kernel: 2.6.27-14-generic
                    X.org: 1.5.2
xserver-xorg-video-radeon: 6.9.0+git20081003.f9826a56-0ubuntu2.1

It is the same radeon driver version that ships with the stable Kubuntu repos and only the kernel and DM version changed. And some issues seem to be fixed for me.

@Kevin: What type of corruptions are you experiencing and when?

[Ancoron Luciferis (ancoron-luciferis)]

Well I must withdraw my previous statement. Today I got that type 3 corruption again after a week of a running system.

[Mandor-ngs (mandor-ngs)]

System:

RV635
Core2 Quad 6600
Debian testing.
xserver-xorg-video-radeon 6.12.2
xserver-xorg 7.4.
xrandr 1.3
kernel 2.6.29 with radeon drm module from git.

Display 20" LCD (1680x1050)

Problem:
Change orientation via "xrandr -o left".
All is ok until I press mouse button - cursor corrupt. It restored when I move mouse to some place, where cursor must change (to link in web-browser for example).
When I enter some text I see artefacts on previos place of blinking cursor.
If I enable SWcursor and enter command "xrandr -o left" - xserver fail and themself restarted.

xserver-xorg-video-radeonhd 1.2.5 fail on command "xrandr -o left" - my computer is locked - keyboard doesn't response, but mouse work.

This is my first bug report. Please ask me, if need more information.

Thanks.

[Bugzi09-fdo-tormod (bugzi09-fdo-tormod)]

Andrew, you will have to open a new bug report here: http://bugs.freedesktop.org/enter_bug.cgi

[agd5f (agd5f)]

I've fixed the "single pixel" corruption in e3c564114eea4d032f920fcebd4040f1934c21b0

The other one is watermark related and I should have a fix for that soon.

[agd5f (agd5f)]

*** Bug 16865 has been marked as a duplicate of this bug. ***

[myxiplx (myxiplx)]

Well, this is back with a vengeance in 9.04. I'd switched back to proprietary drivers in 8.10 since this was so annoying, but in 9.04 those no longer work with my machine:
https://answers.launchpad.net/ubuntu/+question/68783

In fact, 9.04 is a bit of a nightmare all around when it comes to graphics:
- Unusable system if I try proprietary drivers
- Can't login by default if I use the open source ones (I have to disable compiz effects first)
- Once I finally get the open source drivers going I have cursor corruption all over the screen, and can't run my monitors at the proper resolutions.

Not impressed.

[agd5f (agd5f)]

Should be fixed now in git.

[AlexC (alex-j-crowe)]

Great Alex!

I will attempt to build from Git and confirm the fix on my system. Any instructions for compiling from Git from a standard Ubuntu 9.04 driver?

[agd5f (agd5f)]

basic instructions here:
http://www.botchco.com/agd5f/?page_id=2

[Bpaterni (bpaterni)]

When transitioning from screen to screen I'm experiencing left over cursor images on the screen I transitioned from. My cursor is still functional on the active screen, however, and the image goes away once I return to the other screen.

The behavior was first introduced in commit: http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=4e948e2aaa59267a9aa797396add2c6b7b4c1e7c

[agd5f (agd5f)]

(In reply to comment #24)
> When transitioning from screen to screen I'm experiencing left over cursor
> images on the screen I transitioned from. My cursor is still functional on the
> active screen, however, and the image goes away once I return to the other
> screen.
>
> The behavior was first introduced in commit:
> http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=4e948e2aaa59267a9aa797396add2c6b7b4c1e7c
>

I think this should be fixed in 35183ef528411eb0122cf48550d93f921d291d12

[AlexC (alex-j-crowe)]

Looks good to me Alex, I have been using my machine all this evening and the artifacts are not appearing.

I saw the behavior Brian described briefly but I believe the reason was I hadn't done a full restart after installing, only an X restart

Great job!

[Tormod Volden (tormodvolden)]

This should now be fixed upstream, can you please try the -ati/-radeon packages from my PPA https://launchpad.net/~tormodvolden/+archive/ppa ?

[ar-jar (arto-jarvinen-letter)]

Seems to work fine now on my laptop running Jaunty and a Radeon Mobility X1600 in a dual screen "extended desktop" configuration. Visual appearance matter a lot (especially when advocating open source for your friends) so I'm glad you prioritized this. Thank you, and thank you "upstream" for the good work! -Arto

[Arto-jarvinen (arto-jarvinen)]

I installed Tormod Volden's -ati/-radeon packages for Ubuntu that I believe should have these fixes. When I restart the computer after stand-by the cursor flickering comes back. The flickering disappears (for good in that session it seems) when I click once on the second screen in my laptop setup. I'm running ubuntu 9.04 on a laptop with a Radeon Mobility X1600 gfx board. Thanks! -Arto

[ar-jar (arto-jarvinen-letter)]

I'm sorry but I have to modify the above a bit. When I start the computer from stand-by, the flickering in certain "bands" of the display area returns. It disappears again when I click once on screen 2 of my laptop setup. A clue perhaps? I'll make a comment of this to Bugzilla – Bug 19215 too. -A

[Stephen-boddy (stephen-boddy)]

Just to add to the chorus of approval, I'm also on the Tormod Volden packages for Ubuntu, and it is totally fixed for me too. Many thanks (I know this was minor, but it really used to irritate me!)

[ComBatXL (florian-eberle)]

i compiled 35183ef528411eb0122cf48550d93f921d291d12 a week ago and it works fine till now. thanks alex!

---
ubuntu 9.04
radeon
VGA compatible controller: ATI Technologies Inc RV516 [Radeon X1300/X1550 Series]

[Ronald van Engelen (ronalde)]

We are experiencing horizontal glitches showing very briefly (multi color single pixel rows while moving cursor) using 1:6.12.1-0ubuntu2 on our ltsp-clients (HP t5735).

[ar-jar (arto-jarvinen-letter)]

There is still a little flickering left in the 6.12.2.99+git20090508.76fdae90-0ubuntu0tormod_i386 versions. Not sure if those are the latest and the greatest. I'm running an ATI Radeon Mobility X1600 on a Compac nc8430 laptop with a fully updated Jaunty (except for the drivers that are from Tormod's PPA).

As I mentioned above, the flickering comes when returning from suspension and it disappears when I click once on the secondary screen. I think the last fact should be a powerful clue to finding the last little piece of this bug. I don't think this ticket thus should have the status fix released.

Thanks! -A

[Ancoron Luciferis (ancoron-luciferis)]

Yepp, tested on a HD2400Pro and an X1950XTX and finally the cursor got fixed.

Thanks Alex!

[Bryce Harrington (bryce)]

Looks like the upstream fix has propagated into the version we now have in Karmic.

If you can still reproduce the problem on Karmic, either file a new bug or reopen this one.

[Ancoron Luciferis (ancoron-luciferis)]

Sorry for the reopen but today the problem reappeared (similar to type 2 corruption, but not thinner at the top and not wider at bottom).

This appeared on a system with an X1950XTX using GIT master up to commit 2b9a5be3bb22fad1d52ec9eea7733c5e956250b7.

The system was running for about a whole day (no system standby or suspend in between, just monitor blank/off via DPMS).

Regards,

Ancoron

[Ancoron Luciferis (ancoron-luciferis)]

Well, I have to add something here.

Shortly after the corruption appeared (around an hour or so) the cursor changed into a square as I moved the cursor from one screen to the other. Additionally the cursor moved with a delay and not smooth at all. After that the keyboard was inaccessible and logging in via SSH remotely (which took quite some time) I saw that Xorg was using almost all CPU (dual-core AMD X2) and what I was curious about was that even 'top' was using very much CPU (~20-60%).

The system didn't come down correctly after issuing a 'shutdown now' so I had to hard-reset the system.

Although the mouse pointer was movable (hardware cursor) the system didn't respond to anything.

My current setup:

                  xorg: 7.4~5ubuntu18
          xserver-xorg: 7.4~5ubuntu18
     xserver-xorg-core: 1.6.3+git20090801+server-1.6-branch.f274e595-0ubuntu0sarvatt~jaunty
xserver-xorg-video-ati: 6.12.99+git20090729.2b9a5be3-0ubuntu0sarvatt~jaunty

Regards,

Ancoron

[agd5f (agd5f)]

(In reply to comment #32)
> Well, I have to add something here.
>
> Shortly after the corruption appeared (around an hour or so) the cursor changed
> into a square as I moved the cursor from one screen to the other. Additionally
> the cursor moved with a delay and not smooth at all. After that the keyboard
> was inaccessible and logging in via SSH remotely (which took quite some time) I
> saw that Xorg was using almost all CPU (dual-core AMD X2) and what I was
> curious about was that even 'top' was using very much CPU (~20-60%).
>
> The system didn't come down correctly after issuing a 'shutdown now' so I had
> to hard-reset the system.
>
> Although the mouse pointer was movable (hardware cursor) the system didn't
> respond to anything.

That's a GPU lock-up.

[Ancoron Luciferis (ancoron-luciferis)]

(In reply to comment #33)
>
> That's a GPU lock-up.
>

OK, that's why I asked about the ability to monitor temperatures a while back on the mailing list.

The system runs fine now till then, although I experienced the mouse cursor corruption from time to time but it disappeared after some minutes when I move the mouse from one monitor to the other.

Is it possible that the corruption itself is caused by high temperatures of the GPU? Seems unlikely to me but I'm not a GPU expert.

[G-trevor (g-trevor)]

This has reappeared for me. I had not seen it for a long time, but after recent upgrade to Fedora 14 it has reappeared. It is does not always seem to be present, and may be related to suspend/resume. That is I think it only happens sometimes after the system resumes from a suspended state.

[G-trevor (g-trevor)]

Created attachment 41465
Xorg.0.log when corruption present

[Roman Šmakal (schmakerisko)]

I have type 3 and Single Pixel (vertical and vertical + horizontal .. ill say 128x128 px area) type corruptions when playing Heroes of Newerth. There is no way to reproduce it easily, but its quite often issue. It happens when you scroll the map on edge of the screen (for some reason i think it happens only at corners).

GPU is AMD Sumo on 3870K APU. It happens for quite longer time. Im using Linux 3.11 with DPM enabled (but that does not matter as it happened on older kernels as well) and git mesa from ubuntu PPA.

Feel free to ask for more informations.

[penalvch (penalvch)]

Fixed as per https://bugs.freedesktop.org/show_bug.cgi?id=19215#c26 .