Bug 14331 - Radeon XPRESS 200M: System hang with radeon DRI and Fedora 10 userspace unless DRI=off
Summary: Radeon XPRESS 200M: System hang with radeon DRI and Fedora 10 userspace unles...
Status: CLOSED UNREPRODUCIBLE
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks: 14230
  Show dependency tree
 
Reported: 2009-10-06 00:29 UTC by Alex Villacis Lasso
Modified: 2009-12-29 16:20 UTC (History)
4 users (show)

See Also:
Kernel Version: 2.6.32-rc3
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments
Config file for kernel build (93.16 KB, text/plain)
2009-10-06 00:29 UTC, Alex Villacis Lasso
Details
Xorg log up to the point where kernel crashes (15.36 KB, text/plain)
2009-10-06 14:54 UTC, Alex Villacis Lasso
Details
dmesg output after boot with graphical login and DRI=off (35.94 KB, text/plain)
2009-10-06 14:55 UTC, Alex Villacis Lasso
Details

Description Alex Villacis Lasso 2009-10-06 00:29:44 UTC
Created attachment 23278 [details]
Config file for kernel build

System is Fedora 10 with the following graphics userspace:
xorg-x11-server-common-1.5.3-17.fc10.i386
xorg-x11-server-Xorg-1.5.3-17.fc10.i386
xorg-x11-drv-ati-6.10.0-2.fc10.i386
mesa-dri-drivers-7.2-0.15.fc10.i386
mesa-libOSMesa-7.2-0.15.fc10.i386
mesa-libGL-7.2-0.15.fc10.i386
libdrm-2.4.0-0.21.fc10.i386

xorg.conf is configured for driver "radeon" with:
Section "Device"
	Identifier  "Videocard0"
	Driver      "radeon"
	Option "AccelMethod"  "EXA"
	Option "AGPMode" "4"

	Option "AGPFastWrite" "on"
	Option "AccelDFS" "on"
EndSection

Virtual    1440 1080

lspci of video chipset is:
01:05.0 VGA compatible controller: ATI Technologies Inc RC410 [Radeon Xpress 200] (prog-if 00 [VGA controller])
	Subsystem: Intel Corporation Device d600
	Flags: bus master, 66MHz, medium devsel, latency 64, IRQ 17
	Memory at d8000000 (32-bit, prefetchable) [size=128M]
	I/O ports at ee00 [size=256]
	Memory at fdef0000 (32-bit, non-prefetchable) [size=64K]
	[virtual] Expansion ROM at fde00000 [disabled] [size=128K]
	Capabilities: <access denied>
	Kernel modules: radeon

BIOS configured for 64 MB VRAM.

With distro kernels kernel-2.6.27.30-170.2.82.fc10.i686 and kernel-2.6.27.35-170.2.94.fc10.i686, and also with vanilla kernels 2.6.31 and 2.6.31.1, the graphics system works normally with DRI fully enabled.

With recently-released 2.6.32-rc3, the graphical login results in a black screen and no keyboard response (but keyboard leds do not blink). I have not yet managed to capture a dmesg output or a xorg log. I will try to produce those later.


Setting
Option "DRI" "off"

on xorg.conf makes the graphical login show up, but with no compiz or any other DRM support.

To reproduce:
Install Fedora 10 with latest updates
Compile and install 2.6.32-rc3
Attempt to use graphical login, or otherwise start X with DRI on.

Actual results:
System hang with black screen and unresponsive keyboard.

Expected:
Either 1) DRI works normally as it did on earlier versions, or 2) DRI refuses to work with too-old userspace but allows X to start without DRI and without additional configuration.

Build 2009-10-05 on Fedora 10 i386
Comment 1 Alex Villacis Lasso 2009-10-06 14:54:29 UTC
Created attachment 23284 [details]
Xorg log up to the point where kernel crashes
Comment 2 Alex Villacis Lasso 2009-10-06 14:55:10 UTC
Created attachment 23285 [details]
dmesg output after boot with graphical login and DRI=off
Comment 3 Alex Villacis Lasso 2009-10-06 14:56:26 UTC
A ssh session that was made into the test computer went unresponsive after triggering the crash. This, in my opinion, is confirmation that the problem is kernel related, and not xorg-related.
Comment 4 Alex Villacis Lasso 2009-10-07 14:49:24 UTC
I have checked 2.6.32-rc2 and it also hangs the machine when trying to use DRI.
Comment 5 Andrew Morton 2009-10-14 02:46:42 UTC
What's the latest version of the kernel which did not have this bug?

Thanks.
Comment 6 Alex Villacis Lasso 2009-10-14 15:40:58 UTC
2.6.31, 2.6.31.1, and 2.6.31.3 do not have the bug. 2.6.31.3 is what I am running right now. 2.6.32.rc1/2 do have the bug. I am aware that rc1 and rc2 are actually the same tag.

I posted the same question at xorg-driver-ati@lists.x.org and got the following response:
----------------------------
>>> > >> Adding Option "DRI" "off" to xorg.conf allows me to start up X, but I
>>> > >> lose DRI and compiz.
>>> > >>
>>> > >> I would like your opinion about the following issues:
>>> > >> * Should I expect DRI to continue working for radeon in a setup with a
>>> > >> newer kernel and old userspace?
>>> > >> * Have any of you experienced anything similar with 2.6.32-rc2
>>> onwards?
>>> > >>     
>> > >
>> > > F10 and F11 radeon kernel and userspace are not compatible with
>> > > upstream.  Both need to be upgraded to use the latest bits.
>> > >
>> > > Alex
>> > >
>> > >   
> > But, is it acceptable to make the system hang up on this 
> > incompatibility? I think a better response would be for the -rc3 kernel 
> > to refuse the queries on DRI that come from incompatible userspace, and 
> > then DRI would be disabled, but X would still start up. Or even that 
> > xserver refuses to start at all and drop to the command line, but still 
> > have control of the machine. Either of these would be better than a 
> > system lockup that forces a hard reset. Is either of these even possible 
> > if only the kernel is updated?
> > 
> > 

No possible without introducing crap in the kernel to check userspace
process which will never be accepted upstream. You should rather try
Fedora12 livecd or update to rawhide we made several improvement that
we are adding to fedora12.
----------------------------

But I still think that the kernel should behave in a way that will at least give me back control of the machine without a hard reset.
Comment 7 Jérôme Glisse 2009-10-14 22:03:17 UTC
So you are not enabling KMS in 2.6.32 right ? Did 2.6.31-rc4 had the bug ? I can't remember any change to old radeon drm which might play a role in what you are seeing.
Comment 8 Alex Villacis Lasso 2009-10-14 22:34:49 UTC
I disabled KMS long ago because my monitor needs a gamma setting of 2.08 and with the distro kernel KMS results in non-working gamma support (screen is too dark, xgamma does not work). In addition, KMS from around vanilla 2.6.30 onwards no longer works with my distro-supplied video drivers, but I did not mind it very much because I worked without KMS anyway.

Are you sure that you mean 2.6.31-rc4 and not 2.6.32-rc4 ? 2.6.31-rc4 is older than 2.6.31, which I know for a fact that works. The change that broke my setup was introduced somewhere between 2.6.31 and 2.6.32-rc1. But I can try 2.6.31-rc4 anyway if you think it will demonstrate something.
Comment 9 Alex Villacis Lasso 2009-10-28 00:49:41 UTC
Still present in 2.6.32-rc5.
Comment 10 Rafael J. Wysocki 2009-10-28 18:34:13 UTC
On Wednesday 28 October 2009, Alex Villací­s Lasso wrote:
> Rafael J. Wysocki escribió:
> > This message has been generated automatically as a part of a report
> > of recent regressions.
> >
> > The following bug entry is on the current list of known regressions
> > from 2.6.31.  Please verify if it still should be listed and let me know
> > (either way).
> >
> >
> > Bug-Entry   : http://bugzilla.kernel.org/show_bug.cgi?id=14331
> > Subject             : Radeon XPRESS 200M: System hang with radeon DRI and
> Fedora 10 userspace unless DRI=off
> > Submitter   : Alex Villacis Lasso <avillaci@ceibo.fiec.espol.edu.ec>
> > Date                : 2009-10-06 00:29 (21 days old)
> >
> >   
> Sorry for answering late. This bug is still present in 2.6.32-rc5 - I 
> have just checked.
Comment 11 Evgeniy 2009-12-13 19:28:27 UTC
I too have rc-7 and last 32 stable version of the kernel rendering a system freeze just before the gdm login screen shows up.
Moreover, KMS seems to be extremely dangerous for our card right now (X200M rc410). The earlier versions of the kernel 31-14 and up are all giving system hangs sooner or later with the radeon.modeset=1 option enabled. https://bugs.launchpad.net/ubuntu/+source/linux/+bug/489447
It might be a different issue and I will report this here as a separate bug. There is no way I can provide the logs: they are either missing or garbled by the crashes.
Comment 12 Evgeniy 2009-12-14 00:29:15 UTC
On ubuntu karmic 9.10 the last kernel image called drm-next has also this bug.
Booting Linux version 2.6.32-996-generic (root@zinc) (gcc version 4.2.3 (Ubuntu 4.2.3-2ubuntu7)) #200912071003 SMP Mon Dec 7 10:56:26 UTC 2009 with the radeon.modeset=1 does the same thing.
Comment 13 Rafael J. Wysocki 2009-12-29 16:19:52 UTC
On Tuesday 29 December 2009, Alex Villací­s Lasso wrote:
> El 29/12/09 10:28, Rafael J. Wysocki escribió:
> > This message has been generated automatically as a part of a report
> > of regressions introduced between 2.6.31 and 2.6.32.
> >
> > The following bug entry is on the current list of known regressions
> > introduced between 2.6.31 and 2.6.32.  Please verify if it still should
> > be listed and let me know (either way).
> >
> >
> > Bug-Entry   : http://bugzilla.kernel.org/show_bug.cgi?id=14331
> > Subject             : Radeon XPRESS 200M: System hang with radeon DRI and
> Fedora 10 userspace unless DRI=off
> > Submitter   : Alex Villacis Lasso<avillaci@ceibo.fiec.espol.edu.ec>
> > Date                : 2009-10-06 00:29 (85 days old)
> >
> >
> >
> >    
> I will not be able to test this regression anymore. I recently switched 
> to Fedora 12 on the target machine, and of course the userspace works 
> with the distro kernel.

Note You need to log in before you can comment on or make changes to this bug.