FreeBSD-amd64 STABLE/xorg7.4 -- System Freeze

[Robert Noland]

On Sat, 2009-04-04 at 16:18 +0300, Ilias Marinos wrote:
> * Robert Noland (rnoland at FreeBSD.org) wrote:
> > On Fri, 2009-04-03 at 21:38 +0300, Ilias Marinos wrote:
> > > * Robert Noland (rnoland at FreeBSD.org) wrote:
> > > > On Fri, 2009-04-03 at 19:52 +0300, Ilias Marinos wrote:
> > > > > Hello list,
> > > > > 
> > > > > My system freezes while trying to startx.Details:
> > > > > 
> > > > > FreeBSD kropotkin.liakushive.priv 7.2-PRERELEASE FreeBSD 7.2-PRERELEASE
> > > > > #0 r190662: Fri Apr  3 00:50:34 UTC 2009
> > > > > root at kropotkin.liakushive.priv:/usr/obj/usr/src/sys/GENERIC  amd64
> > > > > 
> > > > > -ATI HD 4670 GPU
> > > > > 
> > > > > -xorg.conf can be found here:
> > > > > https://marinos.ceid.upatras.gr/~admin/xorg.conf
> > > > > I've tried all the possible drivers : ati, radeon, radeonhd,vesa.Also I
> > > > > tried disabling modules like glx,dri etc.
> > > > 
> > > > Please set Option "AccelMethod" "EXA" and add Option "DRI" "On".
> > > > 
> > > > robert.
> > > >  
> > > > 
> > > > -- 
> > > > Robert Noland <rnoland at FreeBSD.org>
> > > > FreeBSD
> > > 
> > > Unfortunately , the problem remains.Some bright pixels appears at the
> > > center of the monitor and the system freezes(sometimes if you press any
> > > button it causes a hard reset).
> > 
> > So the only thing that looks interesting in the logs is that MTRR
> > failed.  What was the last working revision?
> > 
> > robert.
> > 
> > > 
> > -- 
> > Robert Noland <rnoland at FreeBSD.org>
> > FreeBSD
> 
> 
> Hello Robert,
> 
>   I am trying to make Xorg stop freezing the system trying various
> configurations and drivers but still no success.
> 
>   I am running startx in a ssh session, while the same time gathering logs
> (kernel messages and Xorg logs).The first startx runs ok without system
> freeze but the monitor (a Dell 2709W  connected with the GPU via DVI)
> gets in power save mode and suspends(like there is no signal) .From this
> state there is no way for the monitor to come back to previous operation
> as it doesn't gets signal even if I kill X(coming back to
> console).However, the system stays responsive and functional via remote
> usage(ssh).You can see the logs (all in one file) gathered here:
> 
> 	https://marinos.ceid.upatras.gr/~admin/log.0
> 
>   The only lines that look suspicious to me are:
> exaCopyDirty: Pending damage region empty!(/var/log/Xorg.0.log / last
> line)

This is reported to be harmless.

> failed to set mtrr: Invalid argument (startx output)
> 
>   After the first startx I kill X and run a second startx.Then a system
> freeze happens and no logs can be collected.

Unfortunately, restarting X is know to break things for at least some
cards/chips.  I'm planning to try to look into it, but it's a tricky
area, and afaik, it also often fails without drm so that opens up the
suspect code area a lot.

>   These logs were collected after the X kill:
> 
> 	https://marinos.ceid.upatras.gr/~admin/log.1
> 
> 
>   Is there any way to check what is causing mtrr failure?

If you are using drm, then drm handles setting mtrr.  If you aren't,
then I think this should be occurring in libpciaccess now.  Someone(tm),
needs to have a look and see what is going on there.  I'll try and get
to it, but I'm in ports mode right now, trying to get everything ready
before the freeze for 7.2.

robert.

> 
> Thanks for your time.
> Ilias 
>  
>   
> 
-- 
Robert Noland <rnoland at FreeBSD.org>
FreeBSD
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: This is a digitally signed message part
Url : http://lists.freebsd.org/pipermail/freebsd-x11/attachments/20090404/975a8a00/attachment.pgp