Bug 18024 - R300 crash (fix included)
Summary: R300 crash (fix included)
Status: RESOLVED NOTABUG
Alias: None
Product: DRI
Classification: Unclassified
Component: General (show other bugs)
Version: unspecified
Hardware: x86 (IA32) Linux (All)
: medium major
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords: regression
Depends on:
Blocks:
 
Reported: 2008-10-12 03:31 UTC by Kévin Ferrare
Modified: 2008-10-13 18:37 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments

Description Kévin Ferrare 2008-10-12 03:31:44 UTC
Hi,

Since the 6.7.197 version of the ATI drivers and independantly of the server / kernel version, I have been unsable to use my ATI card with X. Depending on the driver versions, it would either instantaneously freeze the system with screen corruption or take some seconds during which I can use the interface until it freezes.

My video card is detected as follow by X :
  Chipset: "ATI Radeon Mobility 9600/9700 (M10/M11) NP (AGP)" (ChipID = 0x4e50)

Using the git bisect tool, I have been able to find the commit which caused the regression :

09348a83d06ba9d3129499c4daedd44a68771530 is first bad commit
commit 09348a83d06ba9d3129499c4daedd44a68771530
Author: Alex Deucher <alex@botch2.(none)>
Date:   Tue Jan 29 12:09:24 2008 -0500

    Seems the default is more reliable... we can't win.

    Revert "radeon: Default to 1x again with non-v3 AGP cards."

    This reverts commit b653e5a628bfa4dfb168e96f93f41eb910f409fb.

:040000 040000 fc0f7cc2805d5fbd8999becb3fa50135d8863701 a339cd27696fbac9ccdb72e0d3d549a457f75afa M      src

( http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=09348a83d06ba9d3129499c4daedd44a68771530 )

I then downloaded the sources of the 6.9.0 version of the driver and reversed the commit (that part of radeon_dri.c has not changed since then so it was quite straightforward)

With this fix, I can now use the 6.9.0 version of the driver.
Sounds like an AGP mode detection problem with my chip ?

If you need further information, I can of course provide them.

Kévin
Comment 1 Alex Deucher 2008-10-13 09:03:15 UTC
You don't have to go through all that trouble.  That commit merely changes the default AGP mode.  Just adjust the AGPMode option in the device section of your xorg.conf:

Option "AGPMode" "X"

Replace X with 1 or 2 or 4

Basically AGP sucks and some chipset/GPU combos only work well in certain configurations.  The problem is no one default will work in all cases, so we try and pick a default that will work on the broadest range of hw possible.
Comment 2 Kévin Ferrare 2008-10-13 17:25:44 UTC
I will try to add that line to my xorg.conf when I will have access to my hardware.

If this is a known problem, may I suggest making the default settings less aggressive? (Supposing that AGP 1x would work better on a broader range of hardware)

Also I do not know whether this is relevant or not, but my card works flawlessly under windows in AGP 2x/4x mode ; that makes me thing that the bug I encounter may only be the tip of the iceberg.
Comment 3 Alex Deucher 2008-10-13 18:37:51 UTC
(In reply to comment #2)
> If this is a known problem, may I suggest making the default settings less
> aggressive? (Supposing that AGP 1x would work better on a broader range of
> hardware)
> 

We've tried various default settings and the current setting was what worked reliably for the most people.  We tried 1x, but that was worse than than the alternative, so we changed it back.  The commit in question does exactly that.

> Also I do not know whether this is relevant or not, but my card works
> flawlessly under windows in AGP 2x/4x mode ; that makes me thing that the bug I
> encounter may only be the tip of the iceberg.
> 

AGP is broken by design.  You card combo/chipset probably requires some special system specific workaround.


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.