Bug 5201

Something sounds very broken for this to happen.

Have you got a logfile ?

Created attachment 3952 [details]
failure log

This is a log of the (noisy) failure of the driver. Only on about half the runs
do I actually get the "QWord boundary" error message, but the failure to draw
things on the screen is consistent.

Created attachment 3953 [details]
my xorg.conf

You'll notice I've tried a few things to try to pacify the driver. I've tried
with glx and dri either explicitly loaded or commented out, and I tried
knocking down the resolution to something friendly. None of that accomplished
anything.

If you look furthur at the top of the log you'll see INVALID MEM ALLOCATION.

I suspect this is the problem.

Who makes this machine and can you see if there is a later BIOS for your hardware ?

(In reply to comment #4)
> If you look furthur at the top of the log you'll see INVALID MEM ALLOCATION.
> 
> I suspect this is the problem.
> 
> Who makes this machine and can you see if there is a later BIOS for your
hardware ?

The machine is a custom notebook from unitedmicro.com -- it's all Intel
components, stuck together in an unbranded case. (It's the N259EI on their site,
if you care to take a look.) Intel's site says to see the manufacturer for BIOS
upgrades, and unitedmicro.com has nothing worthwhile to say at all, so I'll try
to take it up with their tech support people in the morning. I'll keep you posted.

The alternative fix is to fix the Xserver source code.

Find the file xf86Bus.c and comment out the function call to ValidatePci().

Rebuild the Xorg server and try again.

Created attachment 4201 [details]
martin's xorg.0.log log

i have exactly the same problem with the intel 855GM graphics chip in my
notebook. The screen is corrupted, the mouse cursor isn't displayed and my
keyboard doesn't work. If i kill the X server over an ssh session the screen
holds it contens and doesn't switch back to the virtual console.

I attached the Xorg.0.log. Notice the reapting error messages at the end, like:

"ADVANCE_LP_RING: outring (0x1c37) isn't on a QWord boundary(WW) I810(0):
Fixing display offsets."

I think the BIOS of my notebook doesn't offer the correct native video mode
(1280x800), because with X11R6.8 (which worked great with i810) i had to use
the 855resolution tool to reprogram the vbios so that it supports that mode. A
newer BIOS isn't available. My Notebook is a Cebop MEX 100 and it's identical
in construction with the Uniwill 223II0
(http://www.uniwill.com/products/mobility/223ii0/223ii0.php?HL=3&M=3). I dead
it runs with the latest uniwill bios, because cebop doesn't provide bios
updates.

If you don't load agpgart and intel-agp - does the problem still happen ??

For me - yes.

Can you upload a log without them loaded ?

Martin - you need to rebuild X with the ValidatePci() command commented out as I
said in comment #6.

*** Bug 5443 has been marked as a duplicate of this bug. ***

Martin - are you running 2.6.15 kernel ??

Did you run an older kernel previously ?

*** This bug has been marked as a duplicate of 5443 ***

Alan, i'm now using Kernel 2.6.15.1 and commented out the ValidatePci() function
call. Everything works smooth now (incl. dri/glx with i915) and there are no
more error messages in the logfile except: 

(WW) I810(0): Bad V_BIOS checksum
(WW) I810(0): Extended BIOS function 0x5f05 failed.

I'm running the 1280x800 resolution without the 855resolution tool.

*** Bug 5685 has been marked as a duplicate of this bug. ***