Bug 58166 - RV670 AGP regression between drm-fixes and drm-next - no display.
Summary: RV670 AGP regression between drm-fixes and drm-next - no display.
Status: RESOLVED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Radeon (show other bugs)
Version: XOrg git
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-12-12 00:03 UTC by Andy Furniss
Modified: 2012-12-22 20:59 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:

Attachments
working dmesg drm-fixes (27.09 KB, text/plain)
2012-12-12 00:03 UTC, Andy Furniss 		no flags Details
failing kernel log drm-next (42.01 KB, text/plain)
2012-12-12 00:16 UTC, Andy Furniss 		no flags Details
Fix dma fence driver (1.91 KB, text/plain)
2012-12-12 21:45 UTC, Jerome Glisse 		no flags Details
Force wb bo into vram (3.20 KB, patch)
2012-12-12 21:45 UTC, Jerome Glisse 		no flags Details | Splinter Review
View All

Description Andy Furniss 2012-12-12 00:03:59 UTC 
Created attachment 71363 [details]
working dmesg drm-fixes

Just put my RV670 AGP back in an old box that's been running a r300 for the past 6 months.

Booting with drm-fixes kernel works, drm-next fails - I loose display as soon as I modprobe radeon, sysrq works. Exact line used =

modprobe radeon modeset=1 agpmode=8 no_wb=1

agpmode=-1 doesn't help.

dmesg from working and log from fail attached.
Comment 1 Andy Furniss 2012-12-12 00:16:57 UTC 
Created attachment 71366 [details]
failing kernel log drm-next
Comment 2 Alex Deucher 2012-12-12 04:03:51 UTC 
Does it work if you disable AGP? Boot with radeon.agpmode=-1 on the kernel command line in grub.
Comment 3 Alex Deucher 2012-12-12 04:05:01 UTC 
Sorry, missed the line about agpmode in comment 1.
Comment 4 Michel Dänzer 2012-12-12 09:26:20 UTC 
Can you bisect?
Comment 5 Andy Furniss 2012-12-12 12:34:59 UTC 
(In reply to comment #4)
> Can you bisect?

commit 4d75658bffea78f0c6f82fd46df1ec983ccacdf0
Author: Alex Deucher <alexander.deucher@amd.com>
Date:   Thu Sep 27 15:08:35 2012 -0400

    drm/radeon/kms: Add initial support for async DMA on r6xx/r7xx
    
    Uses the new multi-ring infrastucture.  6xx/7xx has a single
    async DMA ring.
Comment 6 Jerome Glisse 2012-12-12 21:45:07 UTC 
Created attachment 71412 [details]
Fix dma fence driver

This patch should fix it for you. If it doesn't please also apply on top of it the patch that i will attach next.
Comment 7 Jerome Glisse 2012-12-12 21:45:51 UTC 
Created attachment 71413 [details] [review]
Force wb bo into vram

If previous patch didn't fix the issue on its own please also apply this patch on top of it
Comment 8 Andy Furniss 2012-12-13 00:22:42 UTC 
(In reply to comment #6)
> Created attachment 71412 [details]
> Fix dma fence driver
> 
> This patch should fix it for you. If it doesn't please also apply on top of
> it the patch that i will attach next.

This fixes it, thanks.
Comment 9 Florian Mickler 2012-12-22 09:22:12 UTC 
A patch referencing this bug report has been merged in Linux v3.8-rc1:

commit 86a1881d08f65a42c17071a59c0088dbe2870246
Author: Jerome Glisse <jglisse@redhat.com>
Date:   Wed Dec 12 16:43:15 2012 -0500

    drm/radeon: fix fence driver for dma ring when wb is disabled

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.