Bug 21244 - R600: ForceLowPowerMode causes display corruption
Summary: R600: ForceLowPowerMode causes display corruption
Status: RESOLVED INVALID
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/Radeon (show other bugs)
Version: 7.4 (2008.09)
Hardware: Other All
: medium normal
Assignee: xf86-video-ati maintainers
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
: 25059 (view as bug list)
Depends on:
Blocks:
 
Reported: 2009-04-17 05:22 UTC by Luca Tettamanti
Modified: 2011-04-11 12:37 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments
sample corruption (2.46 KB, image/png)
2009-04-17 05:23 UTC, Luca Tettamanti
no flags Details
sample corruption (4.60 KB, image/png)
2009-04-17 05:23 UTC, Luca Tettamanti
no flags Details
possible fix, only drop to x8 (741 bytes, patch)
2009-04-17 07:22 UTC, Alex Deucher
no flags Details | Splinter Review

Description Luca Tettamanti 2009-04-17 05:22:04 UTC
I'm using radeon driver from git (9dd33cc8), enabling ForceLowPowerMode causes some display corruption.
The corruption seems to be concentrated around icons (I'm attaching a couple of samples) and the mouse pointer (cannot capture on screenshot).
I'm using KDE 4.2, drm kernel module from git current, EXA acceleration is enabled.
Comment 1 Luca Tettamanti 2009-04-17 05:23:02 UTC
Created attachment 24890 [details]
sample corruption
Comment 2 Luca Tettamanti 2009-04-17 05:23:24 UTC
Created attachment 24891 [details]
sample corruption
Comment 3 Luca Tettamanti 2009-04-17 05:45:18 UTC
Forgot the mention: the chip is a M76 (Mobility HD2600).
More information: lspci reports this:
LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
I believe that the link width should be 4, no?
Comment 4 Michel Dänzer 2009-04-17 05:48:13 UTC
Does

    Option "EXANoDownloadFromScreen"

work around the problem?
Comment 5 Alex Deucher 2009-04-17 07:22:01 UTC
Created attachment 24893 [details] [review]
possible fix, only drop to x8

Reducing the number of PCIE lanes seems to negatively impact operations that use the gart (UTS/DFS) in some cases.  I'm not sure whether it makes more sense to not reduce the pcie lanes as much (8 or 16), or to disable UTS/DFS when the lanes are below a certain level.
Comment 6 Luca Tettamanti 2009-04-17 08:17:33 UTC
(In reply to comment #5)
> Created an attachment (id=24893) [details]
> possible fix, only drop to x8

I can't try it right now, but I believe that there's something wrong with the reprogramming of the lanes count; please note that lspci *always* report:

LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk+ DLActive- BWMgmt-
ABWMgm-

At the end of RADEONSetPCIELanes link_width_cntl is 0x000096 (LC_LINK_WIDTH_RD is 1) - regardless of the width requested (I was trying to force 16x).
Comment 7 Luca Tettamanti 2009-04-17 08:18:05 UTC
(In reply to comment #4)
> Does
> 
>     Option "EXANoDownloadFromScreen"
> 
> work around the problem?

Yes, it does.
Comment 8 Ronny 2009-11-13 13:05:12 UTC
*** Bug 25059 has been marked as a duplicate of this bug. ***
Comment 9 Luca Tettamanti 2011-04-11 12:11:16 UTC
The bug was related to reducing the active PCIe lanes regardless of what was present in Atom tables. I believe that this bug is fixed in UMS, and no more relevant for KMS (which leaves the PCIe alone as advertised in the BIOS).
Close?


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.