Bug 85016

Summary: xorg.conf.d/50-monitor.conf, xorg.conf, & EDID preferred are ignored
Product: xorg Reporter: Felix Miata <mrmazda>
Component: Driver/intelAssignee: Chris Wilson <chris>
Status: RESOLVED FIXED QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: major    
Priority: medium CC: ajax, samuel, sndirsch, tobias.johannes.klausmann
Version: unspecified   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
845G Xorg.0.log following no attempt to configure via /etc/X11/
none
xorg.conf that works none

Description Felix Miata 2014-10-14 18:56:24 UTC
Created attachment 107835 [details]
845G Xorg.0.log following no attempt to configure via /etc/X11/

openSUSE 13.2 (solution/workaround found on 865 host gx270):
3.0.0git-199.1 installed 13 July OK
2.99.916-139.1 (24 Sept)  broken
2.99.916-2.2  (current) broken
3.0.0git-216.1 OK
Mageia Cauldron (845 host t2240):
2.99.916-2 broken
Fedora 21 ( 945 host gx62b):
2.99.916-2 broken

Failures are on 845G, 865G, 915G, 945G and 4000 hosts using current and older 2.99.9xx versions. I discovered git works on 865G, then tried it on others. I'm unable to reproduce using Radeon rv200 or rv380.

I can make X start with desired resolution and DPI via xrandr commands in a startup script, but trying equivalent via xorg.conf.d/ or xorg.conf PreferredMode and DisplaySize are ignored. Omitting any configuration attempt produces 1024x768 instead of EDID preferred.

Possibly 'Option "AccelMethod" "UXA"' is also ignored. I can't tell because Fedora's X is freezing my 845 machines.
Comment 1 Chris Wilson 2014-10-14 20:03:01 UTC
Presumably the conf bug is fixed by

commit 1aefd16964c30264c05d136cef3e1e7de5eb9c53
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Sep 12 07:46:43 2014 +0100

    sna: Use default monitor options on the first output
    
    commit 6554cf0a69f04710b64f3488384fe5cb2748dc35
    Author: Chris Wilson <chris@chris-wilson.co.uk>
    Date:   Mon Aug 11 12:22:17 2014 +0100
    
        sna: Parse output options early during initialisation
    
    rearranged the monitor query to before the num_outputs increment. The
    result was that it choose the second output as the default and not the
    intended first.
    
    Bugzilla: https://bugs.gentoo.org/show_bug.cgi?id=522500
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

so I am confused as to what this report is for? What's happening on Fedora exactly?
Comment 2 Chris Wilson 2014-10-14 20:04:20 UTC
Also note that for Option "AccelMethod" "uxa" to work, the stanza must include Driver "intel".
Comment 3 Felix Miata 2014-10-14 20:58:41 UTC
Created attachment 107839 [details]
xorg.conf that works

15:10 -0400 After eliminating /etc/X11/xorg.conf.d/ on 845G, using this /etc/X11/xorg.conf makes all work as expected on Fedora 21, Rawhide, openSUSE 13.2, and Cauldron (no machine freeze; 120 DPI; 1600x1200; etc.). Commenting out 'Option "AccelMethod" "UXA"' restores 1024x768 (and in F21/22, freezing).

I repeated on F21/1.16.0 (last updated 25 Sept) on 945G: same results, except that X did not freeze. I yum distro-sync'd F21 16:00 -0400 and repeated with same result on 1.16.1.

Chris Wilson wrote:
> Also note that for Option "AccelMethod" "uxa" to work, the stanza must include
> Driver "intel".

I didn't try that initially, but did later after ajax mentioned it on test list.

> so I am confused as to what this report is for?

Only xrandr works to manually configure X unless UXA is enabled?

> What's happening on Fedora exactly?

Apparently same as Cauldron and Factory, commit hasn't gotten into any standard repos, and EDID and config files are mis- or not handled when UXA is not operative, I think. I'm frazzled by what seems to be a compound problem and too many interruptions to think straight or remember which machine is doing what. One would think a month old commit would have made it into standard repos of these devel distro versions. (Or, maybe that 12 Sept. commit isn't on point, but I sense not.)

F21 # rpm -q --changelog xorg-x11-drv-intel | head -nX
* Thu Sep 11 2014 Dave Airlie <airlied@redhat.com> 2.99.916-2
- backport some SNA and MST fixes.

* Wed Sep 10 2014 Dave Airlie <airlied@redhat.com> 2.99.916-1
- Rebase to 2.99.916

* Wed Sep 03 2014 Dave Airlie <airlied@redhat.com> 2.99.914-4
- Add UXA MST support as a fallback

This is the head of openSUSE BS xf86-video-intel git3.0.0 changelog:
* Wed Sep 24 08:00:00 2014 eich@suse.com
- U_sna-Validate-framebuffer-tiling-before-creation.patch
  Fix regression causing tiling settings not being validated
  which subsequently lead to failed surface allocations (bnc#898169).

* Tue Sep  9 08:00:00 2014 tobias.johannes.klausmann@mni.thm.de
Comment 4 Chris Wilson 2014-10-15 08:10:47 UTC
(In reply to Felix Miata from comment #3)
> Created attachment 107839 [details]
> xorg.conf that works
> 
> 15:10 -0400 After eliminating /etc/X11/xorg.conf.d/ on 845G, using this
> /etc/X11/xorg.conf makes all work as expected on Fedora 21, Rawhide,
> openSUSE 13.2, and Cauldron (no machine freeze; 120 DPI; 1600x1200; etc.).
> Commenting out 'Option "AccelMethod" "UXA"' restores 1024x768 (and in
> F21/22, freezing).
> 
> I repeated on F21/1.16.0 (last updated 25 Sept) on 945G: same results,
> except that X did not freeze. I yum distro-sync'd F21 16:00 -0400 and
> repeated with same result on 1.16.1.
> 
> > so I am confused as to what this report is for?
> 
> Only xrandr works to manually configure X unless UXA is enabled?

That's fixed already upstream, so nothing more to do there but make a release.

> > What's happening on Fedora exactly?
> 
> Apparently same as Cauldron and Factory, commit hasn't gotten into any
> standard repos, and EDID and config files are mis- or not handled when UXA
> is not operative, I think. I'm frazzled by what seems to be a compound
> problem and too many interruptions to think straight or remember which
> machine is doing what. One would think a month old commit would have made it
> into standard repos of these devel distro versions. (Or, maybe that 12 Sept.
> commit isn't on point, but I sense not.)

If you install xf86-video-intel.git on 845g with f21/f22 and see if you can extract any logfiles. Is the machine completely frozen or is it still network accessible?
Comment 5 Felix Miata 2014-10-17 03:05:15 UTC
(In reply to Chris Wilson from comment #4)
> If you install xf86-video-intel.git on 845g with f21/f22 and see if you can
> extract any logfiles. Is the machine completely frozen or is it still
> network accessible?

Instead, since I don't build, and couldn't find any such rpm for Fedora, on Rawhide on i845G I did:

# wget http://download.opensuse.org/repositories/home:/sumski:/test/openSUSE_Factory/i586/xf86-video-intel-3.0.0git-216.1.i586.rpm # dated 11 Oct on the mirror
# yum remove xorg-x11-drv-intel
# rpm -ivh --nodeps xf86-video-intel-3.0.0git-216.1.i586.rpm

That eliminated both machine lockup and ignoring valid xorg.conf, without specifying UXA, giving me 1600x1200 @ 120 DPI via PreferredMode and DisplaySize on a CRT with EDID preferred mode 1280x1024. It also gave me 1440x900 @ 108 DPI via PreferredMode and DisplaySize on an LCD advertising preferred mode 1280x1024 via EDID but preferred mode 1440x900 in its documentation.
Comment 6 Chris Wilson 2014-12-21 15:13:39 UTC
2.99.917 has been released with the fix.

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.