FC3 upgrade - X is hosed :( (LONG)

Jim Cornette fc-cornette at insight.rr.com
Sun Nov 14 04:42:40 UTC 2004


Mike Klinke wrote:
> On Tuesday 09 November 2004 10:45, sz beam wrote:
>  
> 
>>The laptop is a Sony Vaio with the i815 chipset. The Xorg log is
>>below, I can't see anything amiss in it but most of it makes no
>>sense to me. Apologies to those of you on dialup for this but I
>>don't know what else to say, this is a pretty stock upgrade path
>>and the i815 driver is loaded per normal as far as I can tell.
>>
> 
> 
> Most likely, anything that uses the i810 driver is going to be 
> problematic.  Try some of the hints here:
> 
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=132267
> 
> I use a Sony Vaio too and have had good results using 16bit, 
> (thousands of colors) with FC3T3, but this seems to vary with the 
> version of graphics controller you're using.
> 
> Regards, Mike Klinke
> 


Just to add some experience and some history regarding the breakage of 
the i810 driver. There were some changes made in xorg-x11 that related 
to the 830 cards and dual head. This caused the server to not even come 
up on the first CVS version resulting in the current release of the 
server. Later, there was a patch applied resulting from Mike's bug 
report. Code was changed to get the 810/815 video cards to work again.
When I started the server with the patch which brought up the server 
again for the 815 cards, I experienced the refresh/lockup and hovering 
mouse over portion of screen would refresh partially that part of the 
screen.
Resolutions that worked for me were to replace the
/usr/X11R6/lib/modules/drivers/i810_drv.o
file with a driver extracted from the pre-cvs release or to change my 
xorg.conf file to run at a higher resolution 1280x1024 at 24 depth.
The later was eventually broken by various kernel versions and I just 
gave in and added Option "NoAccel" to the xorg.conf file and things ran 
acceptably.
As Mike points out, if you run at a resolution that DRI is active, the 
refresh problem is not present. There was a bug where changing to a 
terminal via ctl-alt-Fn would kill the server and it would need to be 
restarted. Therefore, it seems that DRI has a problem when the screen 
transitions from graphics to text, it just bails out.
Now, I like higher resolutions and deeper color depth. This runs on the 
815 card with no DRI.

However, I also have experience on a Dell computer with an Intel 865G 
video card. This card runs with DRI at higher resolutions and deeper 
color with no problem. The changes in xorg-x11 actually resolved a 
problem with unpredictable crashes of the server which left a residual 
image of whatever was on the screen before the the X server dies out. 
The residual leftover on the screen problem disappeared and the server 
has not crashed with overnight tested with all of the screensavers cycling.

Lets just say that DRM and AGP might be faulty at the kernel level. It 
might also be that the driver might need to be split up to prevent 
breakage with the newer cards that also use this i810 driver having 
changes, which break or introduce limiting options, which need to be 
applied to the xorg.conf to hide this horrible refresh problem.

Basically, the 810/815 are very effected by same driver and changes to 
improve 830 and above video cards.
865G cards work a lot better and have outlived the xscreensaver daemon 
during my tests with this driver.

We need this fixed and the problem is showing up with other releases 
also. Input regarding all of the different video cards needs to be found 
out and which cards act which way. 810/815/830/845/865/9xx and others. 
Also information regarding how certain kernels effect this problem or if 
the kernel has no effect on th problem.

Sorry Mike for replying to your post and referencing your prior 
findings. Thanks for pointing out the 24 depth in my xorg.conf file 
during the testing phase. Also, does changing to a terminal still cause 
your server to crash when using DRI.

Jim




More information about the fedora-list mailing list