Bug 24455

Summary: [852GM] Intel drivers 2.8.0 and 2.9.0 have tiling corruption in xorg server 1.6.3
Product: xorg Reporter: Yury <yury.tarasievich>
Component: Driver/intelAssignee: Carl Worth <cworth>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium Keywords: regression
Version: unspecified   
Hardware: Other   
OS: All   
i915 platform: i915 features:
Description Flags
WindowMaker screen none

Description Yury 2009-10-11 01:20:45 UTC
Created attachment 30266 [details]
WindowMaker screen

The output is partially corrupted on xorg server 1.6.3 and Intel drivers 2.8.0 and newer, running on the Latitude 100L laptop with the Intel 852GM adapter. E.g. the xterm windows produce "burst and scattered" glyphs, and the glxgears windows produce 5 copies of cogwheels overlaid. Also, there's garbage after windows move/resize/etc.

Downgrading the Intel drivers to 2.7.0 and 2.6.0 (as 2.5.0 won't compile) resolves the issue.

I'm attaching the photo of the part of the WindowMaker screen with the following visible: 
- incorrect TTF rendering in the xterm window (notice the scattered fragments of the text near the left edge of the screen -- those would partially remain on-screen if xterm window's moved or closed)
- glxgears with the 5 copies of cogs
- correct TTF rendering in the window titles
- correct bitmap and TTF rendering in the WindowMaker Appearance application
Comment 1 Carl Worth 2009-10-26 11:28:01 UTC
Hi Yury,

Thanks for the bug report. You said:

> The output is partially corrupted on xorg server 1.6.3 and Intel drivers 2.8.0
> and newer, running on the Latitude 100L laptop with the Intel 852GM adapter.

I'm curious what "and newer" means? We've made some significant improvements
to support for 85x graphics adapters in the 2.9.0 release of the driver,
(2.8.0 was definitely very broken).

Could you please try version 2.9.1 of the driver and report back whether the
bug is fixed?



PS. Please clear the NEEDINFO keyword when you report back.
Comment 2 Yury 2009-10-26 12:27:20 UTC
To clarify: "and newer" meant 2.9.0 tarball. So, 2.8.0 and 2.9.0, in my installation (kernel 2.6.29, glibc 2.9, xorg server 1.6.3), are troubled with this.

Additionally, I noticed today that 2.7.0 (in fact, 2.6.3 and 2.7.0), in this same installation, is also troubled with total screen corruption, but of different kind. The pre-requisite for experiencing this is trying "clean" start of X after the reboot with 2.7.0 drivers. If some other version of intel drivers was installed and tried with X before that, then 2.7.0 may also be installed and normally used with X. I didn't determine which version of drivers "enables" correct work of 2.7.0 -- possibly it's 2.8.0 or 2.9.0.

The version 2.6.0 has bitblit issues when partially scrolling windows.

btw, setting option "Tiling" to "false" has no effect on any of these issues (2.6.0-2.9.0).


I had to switch to 2.6.0 and to endure the bitblit issues.

The info on 2.9.1 test will follow tomorrow.

More details on pre-2.9.1 versions issues will follow tormorrow, if needed.
Comment 3 Yury 2009-10-27 01:48:28 UTC
The 2.9.1 driver has this sort of corrected. Sort of, because the correct work REQUIRES having Option "Tiling" "false" in xorg.conf. With this option set to "true", the 2.9.1 driver exhibits old behaviour, shown on the original screenshot.

Thank you very much anyway, as I'm already able to work reasonably comfortably.
Comment 4 Chris Wilson 2010-07-04 07:59:10 UTC
In the versions of the driver tested, tiling was only used for the front buffer and for GL buffers. Now tiling is used by default. However, this has meant that we have caught quite a few tiling errors throughout the stack.

Yury, afaik we have fixed all the tiling bugs, in i915+ at least, and it would be useful if you could check the current tree on your hardware.
Comment 5 Chris Wilson 2010-07-17 07:26:46 UTC
I'm confident that tiling is working on gen2 hardware as well as it is on gen3...

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.