Bug 208891

Summary: r300 does not work with 4GB on Intel x86_64
Product: [Fedora] Fedora Reporter: Adam Goode <adam>
Component: kernelAssignee: Dave Airlie <airlied>
Status: CLOSED RAWHIDE QA Contact:
Severity: high Docs Contact:
Priority: medium    
Version: rawhideKeywords: Reopened
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2008-02-09 09:19:11 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
Xorg log
none
xorg.conf none

Description Adam Goode 2006-10-02 15:00:23 UTC 
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.8.0.7) Gecko/20060913 Fedora/1.5.0.7-1.fc5 Firefox/1.5.0.7 pango-text

Description of problem:
When I start X on my Dell Precision 380 with PCIE FireGL V3100 (ATI Technologies Inc RV370 5B64 [FireGL V3100 (PCIE)] (rev 80)), X displays a black screen and uses 100% of CPU. X does not respond to kill -9.

I suspect DRI is at fault, but until #207614 is resolved, I can't test without DRI.

The only working solution is VESA driver.

Before I upgraded from FC5, I remember that a later xorg-x11-drv-ati release exhibited the same effects until I commented out 'Load "dri"'.

Version-Release number of selected component (if applicable):
xorg-x11-drv-ati-6.6.2-2.fc6

How reproducible:
Always


Steps to Reproduce:
1. Modify xorg.conf to have Driver "radeon"
2. Start X

Actual Results:
Black screen
X process takes 100% CPU, becomes unkillable

Expected Results:
X.

Additional info:
Comment 1 Adam Goode 2006-10-02 15:02:21 UTC 
Created attachment 137550 [details]
Xorg log
Comment 2 Adam Goode 2006-10-02 15:02:39 UTC 
Created attachment 137551 [details]
xorg.conf
Comment 3 Adam Goode 2007-05-04 00:36:15 UTC 
Tracked at freedesktop here:
https://bugs.freedesktop.org/show_bug.cgi?id=9074


Basically, Dave Airlie on IRC helped out a lot by poking through the code, then
he realized the existing code is doing completely the wrong thing wrt DMA and
cache coherency. It's triggered by swiotlb on Intel. I think the solution
involved "a lot of work" ?
Comment 4 Adam Goode 2007-06-06 02:21:39 UTC 
Fixed upstream in 2.6.22-rc4.
Comment 5 Adam Goode 2008-01-23 02:28:58 UTC 
This bug appears to have returned in rawhide.

Same symptoms, same solution, same "failed writeback test" when radeon.ko loads.

One notable difference, instead of stopping at a blank screen, it stops at a
screen filled with seemingly uninitialized garbage. The mouse pointer responds,
but nothing else.

kernel-2.6.24-0.157.rc8.fc9
xorg-x11-drv-ati-6.7.196-7.fc9
Comment 6 Adam Goode 2008-02-04 19:54:55 UTC 
Still broken in latest rawhide kernel-2.6.24-9.fc9.
Comment 7 Dave Airlie 2008-02-06 03:53:19 UTC 
any ideas when it started to break... sounds like we are going to have to bisect
the kernel to find this ... uggh....
Comment 8 Adam Goode 2008-02-06 04:00:04 UTC 
Not sure exactly, F8 was working, when I switched to rawhide it broke. So
somewhere between kernel-2.6.23.9-85.fc8 and kernel-2.6.24-0.157.rc8.fc9 ??


I'll give bisect a go, I guess.
Comment 9 Adam Goode 2008-02-06 04:36:16 UTC 
Ug, rawhide has gcc 4.3, which the kernel doesn't like...
Comment 10 Dave Airlie 2008-02-08 00:46:18 UTC 
I think we have a fix for this which I've just put in upstream, so hopefully the
next F8 stable and F9 kernels will solve this problem...
Comment 11 Adam Goode 2008-02-09 09:19:11 UTC 
Working great with kernel-2.6.24.1-26.fc9.x86_64 !