Bug 22092 - [G45] xorg crash when GEM is activated
Summary: [G45] xorg crash when GEM is activated
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: x86 (IA32) Linux (All)
: medium blocker
Assignee: Eric Anholt
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-06-04 12:24 UTC by David Amiel
Modified: 2009-10-18 00:22 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
dmesg log (16.00 KB, text/plain)
2009-06-17 12:28 UTC, David Amiel
no flags Details
Xorg log (15.92 KB, text/plain)
2009-06-17 12:28 UTC, David Amiel
no flags Details
Xorg log (1.66 KB, text/plain)
2009-08-06 13:39 UTC, David Amiel
no flags Details

Description David Amiel 2009-06-04 12:24:47 UTC
I have to disable GEM to be able to start xorg for a few months. At the moment I'm using an Intel DG45FC (G45) motherboard with no xorg.conf.

Software I'm using :
- libdrm : 2.4.11
- xorg : 1.6.1.901
- intel drivers : 2.7.1
- mesa : 7.5rc2
- kernel : 2.6.30-rc8

- error in dmesg with gem : http://pastebin.com/f3de5b70f
- xorg log with gem : http://pastebin.com/f6dd1e406

disabling GEM on this setup make everything working
Comment 1 Eric Anholt 2009-06-17 11:38:20 UTC
We need the full logs attached to the bug report, not excerpts in a pastebin that will expire.
Comment 2 David Amiel 2009-06-17 12:28:10 UTC
Created attachment 26896 [details]
dmesg log
Comment 3 David Amiel 2009-06-17 12:28:43 UTC
Created attachment 26897 [details]
Xorg log
Comment 4 Eric Anholt 2009-07-15 16:55:46 UTC
<3>[   16.287322] Xorg:3009 map pfn expected mapping type uncached-minus for a0000-b0000, got write-back
<3>[   16.287730] Xorg:3009 map pfn expected mapping type uncached-minus for a0000-b0000, got write-back

This looks like PAT screwing things up.  Does booting with nopat help?
Comment 5 David Amiel 2009-07-16 03:59:50 UTC
yes we tried this.
Booting with nopat option makes those error messages to disapear, but it doesn't help to start xorg
Comment 6 David Amiel 2009-07-20 14:12:20 UTC
I've done a test with kernel 2.6.30.2, mesa 7.5, libdrm  2.4.12 and intel 2.7.99.902 with the same errors
Comment 7 Gordon Jin 2009-07-22 01:08:55 UTC
clearing NEEDINFO
Comment 8 Eric Anholt 2009-08-05 15:16:55 UTC
I keep coming back to this bug, and I've got no idea what could be going wrong.  Do you know when the failure started occurring?  Could you bisect the kernel to find a guilty commit?

Does it occur with KMS?
Comment 9 David Amiel 2009-08-06 13:37:35 UTC
(In reply to comment #8)
> I keep coming back to this bug, and I've got no idea what could be going wrong.
>  Do you know when the failure started occurring?  Could you bisect the kernel
> to find a guilty commit?
> 
> Does it occur with KMS?
> 

The last working driver has been 2.5.1, with 2.6.27 (.9 I think) kernel, after that it has been impossible to have a working driver for a few months until a geexbox dev find the no-gem tips.

Actually 2.8.0 driver without GEM does not even work on my motherboard : I provide an extract of my xorg.log in attachment. Though we have successfull reports on older intel motherboards (still without GEM)

Xorg crash with KMS too, but the oops occurs in different modules, I can post the corresponding dmesg log if you want

The point is that we have to release a new version of our distribution (people ask for this for months now), and we can't do this with a broken intel driver.

Things are changing a lot (xorg, libdrm, mesa etc etc), and we could have miss something on our side which explains the crash, but actually intel driver is the only one which doesn't work

as there had tons of changes, bissecting seems impossible to do, but we have integrated intel tools, may I do something with them which can help you to identify the problem ?

Comment 10 David Amiel 2009-08-06 13:39:56 UTC
Created attachment 28409 [details]
Xorg log
Comment 11 David Amiel 2009-10-18 00:22:04 UTC
the problem has been solved by changing the kernel configuration, though we're unsure about the exact parameter(s) which make the all thing to work.

Configuring some drivers in modules instead of built-in has definitely helped ( http://hg.geexbox.org/geexbox/rev/76e1f4f53359 ), but hasn't been sufficient

Why a kernel configuration leads to an xorg dirty crash ? that is a big mystery ..

however intel drivers are working fine with GEM and KMS


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.