[Arrandale] gnome-display-properties unable to correctly enable monitors connected to VGA

Attaching xrandr --verbose output when both LVDS and external monitor are enabled.

Thank you for your bug report. The issue is an upstream one and it would be nice if somebody having it could send the bug the to the people writting the software (https://wiki.ubuntu.com/Bugs/Upstream/GNOME)

Same problem here, same hardware.

This is a regression from Maverick.

The GNOME versions didn't change in natty since we stay on GNOME 2.32 and both users having the issue have it on the same hardware seems to me it's likely an issue on the xorg or driver side rather than in the GNOME library

Sebastien,

We tried to figure this out with Bryce, and was determined that this is not related to the driver side as this bug is not seen when disabling LVDS with xrandr, but rather when disabling LVDS with gnome-display-properties.

well, it can also means that the xrandr command and the libgnome-desktop api using x calls exerce different codepath on the xorg side

just to clarify the previous comment, the fact that the xrandr command doesn't have an issue doesn't mean that there is not a bug somewhere else on the xorg side (the apis gnome-desktop are using are different from what the xrand utility is doing)

@seb128, it appears to me that in fact they *do* use the same API into X. Please provide additional references as to the difference in api usage.

gnome-desktop's gnome-rr.c file is a wrapper around libXrandr:

#ifdef HAVE_RANDR
#include <X11/extensions/Xrandr.h>
#endif

The xrandr tool also uses libXrandr.h. Both codebases implement similar logic with analogous but not equivalent code, but underneath they're both utilizing libXrandr. So, where a problem exists when using one tool but not the other it does indeed point to something wrong with the utility and not in the underlying X code.

it would need debugging from someone having the issue then, let's not argue but GNOME didn't change and it's hardware specific it's likely the driver side which is the issue

Tell you what... you think it's X, I think it's gnome, sounds like we're not in agreement yet, so for now let's leave it open against *both*.

Chatting with kirkland and RoAkSoAx, they believe they have the same issue, and are also on arrandale chips.

Nothing arrandale-specific has been added to the xrandr utility recently (in fact the only changes to xrandr since maverick are man page fixes, autoconf tweaks, and one minor fix that only comes into play when using Xinerama... nothing at all relevant to this bug.)

On the X side, pretty much with -intel now all the XRANDR commands are pass-thru to the kernel. There is some logic in X to handle things but it mostly is hardware-independent; the hardware specific modesetting logic is down in linux's drm module.

Andres, can you confirm that this happens when booted into Classic Desktop (no effects)? It appears from your bug report that compiz (or kwin compositing) is turned off, but I want to be doubly sure this is true.

The reason is that upstream says that this bug is because there is a 4096x4096 maximum texture hardware limitation for compositing with this Arrandale chip. So you either need to disable compositing, or arrange your screens to not exceed this limit.

From your log files it appears your laptop is 1280x800 and you have a 3840x1080 monitor, which totals to 5120, and since that's larger than 4096 the graphics card can't handle it.

(Is this correct, that you have a 3840x1080 monitor?)

A way to work around this limitation would be to have your laptop monitor positioned above or below your external monitor.

Bryce,

Yes I confirm that this happens when using Classic Desktop (No effects).

Now, in terms of the monitors, the logs are when plugged to the DualHead2Go, displaying in the two monitors at 3840x1080, 1920x1080 each.

However, the *same* happens when plugged it to a *single* monitor at 1920x1080.

I'll test the position of the laptop monitor bellow the external monitors connected via the dualhead2go by the end of the day.

I believe Dustin can also confirm this issue.

Hope this helps.

Cheers,

Ok, well the case of compiz + >4096 width is essentially a dupe of bug 555641 (which is a wishlist X bug - basically X can't do anything about this until Xrandr 1.4, and so config tools will have to check for the invalid configuration).

However, that issue doesn't occur with compiz turned off.

Andres, try connecting a single monitor without the DualHead2Go unit at all in the mix and see if you can reproduce the problem.

Bryce,

Yes I can confirm that this happens when connecting the laptop directly to the external monitor (not using the dualhead2go).

Now, what I did is (all from Classic Desktop no effects - and from gnome-display-settings:

1. Connect external monitor to VGA
2. Enable mirror screens: WORKED
3. Disable mirror screens and Enable external monitor at 1920x1080: WORKED
4. Disable LVDS and leave VGA as sole monitor: FAILED (You can see what happens in the video: [1])

Now, if I do step 4 with xrandr (disable LVDS with xrandr), it works perfectly.

Furthermore, if I save the configuration (LVDS disabled, VGA1 Enabled) and reboot the computer, then LVDS is off by default and VGA1 should be displaying the image, however, the same occurs as with video [1].

Finally, if I do the steps above but place LVDS *under* the external monitor (using gnome-display-settings), the same as video [1] occurs.

Hope this helps. Please let me know if you need any further info.

[1]: http://me.roaksoax.com/dualmonitor.3gp

Hi Andres,

I can't tell from the logs if it is a dupe, but I posted a fix to a vaguely similar issue (bug #747205) to this PPA: https://launchpad.net/~bryce/+archive/cherry

I've a hunch that fix might help with this one too, would you mind installing that package and testing? Test it with the setup you describe in comment #16 first; it sounds like there may be a separate issue when total horiz res is >4096 so test that type of config only once the single-external-monitor case works.

Hi Bryce,

Will sure do. Though right now I'm out of town, and in Texas for the Texas Linux Fest. I'll test this Monday first thing in the morning.

Cheers,

Bryce,

No luck here :-( Still very, very broken in my home setup.

A few pieces of information, though...
 1) GDM seems to work correctly. ie, if I reboot my system with my external 24" Samsung 1920x1080 monitor hooked up, GDM comes up in mirror mode, full resolution and grandeur. When I log in, though, Classic or Unity, I get nothing on either monitor. Well, I get the flashing that Andres showed in his video.
 2) I gave a talk today at Texas Linux Fest, and brought along a 10.04 LiveUSB. However, when I tested Natty against the monitors the conference provided, everything worked just fine. So perhaps the problem is monitor specific? My monitor is a relatively common Samsung SyncMaster 2333 24", that has worked perfectly from Hardy -> Maverick. It's only Natty that's batty.

Dustin

@Andres, how has your testing gone so far?

@Dustin, the issue seems to have some relationship to the resolution(s) selected, but I don't *think* it is strongly monitor-specific. It might be interesting if you tested against a couple other monitors to see if there is a pattern. But I think the bug has stalled out upstream because as originally reported it sounded to them like a texture buffer size issue (>4096). But it sounds like this is a more fundamental error.

Dustin, it might be helpful to have a new bug report, filed with a very minimal setup (no more than one monitor attached, using the fewest tools to reproduce, etc.) We could shoot that upstream and hopefully get a different set of eyes on it.

Bryce!

Sorry, I guess I really never posted the results (Thought I did) hence, the testing was unsuccessful.

Would you like another bug report against xserver-xorg-video-intel only connected to one monitor?

Hi Andres, yes a new bug report for that configuration would be helpful. Reference this bug report so we can associate it with this one.

Andres, also would you be able to test booting Maverick via a livecd with this monitor configuration, using gnome-display-properties to configure things, and see if you get the same bug there? If it works properly on maverick, that would tend to confirm seb's hypothesis that it's not a gnome regression. If it does happen there, then it would make me the more convinced that it's something wrong in gnome-display-properties.

Hi Bryce,

I booted from a Maverick livecd and the bug does *not* happen. Everything works as expected, so yeah I guess that we can confirm that this is not a gnome regression. I'll file the new bug now.

Bryce,

I have file bug #761236. To file the bug report I did the following:

 1. Connected to my x201 through remote desktop.
 2. Launched gnome-display-properties and enabled output through VGA1: WORKING
 3. Disabled LVDS: ERROR showed as previously discussed (flickering)
 4. While having the external Monitor flickering and LVDS off (and connect through remote Desktop) i filed the bug with:

 - ubuntu-bug xserver-xorg-video-intel

I have the same issue as Dustin. x201 on Ultrabase docking station, connected to a Samsung 225BW monitor.

If I connect the laptop to my smallish 17" 1280x1024 Sony, I manage to hook it up in dual display mode. But if I try to disable the laptop screen it fails with a black screen /cursor. So I can't really use it in single-screen mode on the docking station.

If I connect the laptop to my widescreen Samsung 22", it directly goes to the parasited flashing screen effect.

So it looks like there are two separate bugs, one of them being specific to large screens / Samsungs.

This is also very similar to what Anmar described at bug 732050.

I managed to get it kinda working. The flashing issue with the Samsung seemed to be in a stale monitors.xml file. Posting for reference:

Boot laptop separate, then remove ~/.config/monitors.xml and /etc/gnome-settings-daemon/xrandr/monitors.xml

Shutdown. Place laptop on Ultrabase. Boot, system comes up in cloned screens mode.

gnome-display-properties, disable cloned mode, set monitor screen resolution, apply. We are now in dual head mode, with the unity launcher on the laptop screen. The only way to get it on the main screen seems to be to disable laptop screen.

disable laptop screen, apply. The confirmation countdown apparently runs somewhere else and will revert your change in 30 seconds... unless you press apply again and confirm it.

I still have to experiment on how I can make that setup stick... or work with the laptop lid closed on the Ultrabase.

Okay, I'm *finally* back home and was able to test this...

I can confirm part of what Thierry suggested. I removed ~/.config/monitors.xml and that got rid of the Samsung flashing monitor on reboot.

It comes up in cloned monitor mode, dropping down to the lowest common resolution, which is 1280x800. I am now working on my 24" monitor at 1280x800, which is better than the unusable flashing screen.

Okay, so now I try to change the resolution of my Samsung monitor to its max. I disabled the laptop screen entirely, and then I set the external Samsung to 1920x1080, and boom, flashing green/black screen all over again :-(

Give it 30 seconds, and it restores.

Next, I tried enabling both screens, with my laptop at 1280x800 and my external monitor at 1920x1080. Surprisingly, this worked just fine. Unfortunately, this isn't my preferred working setup :-( I prefer to put my laptop in my docking station, lid closed, and go straight from a single screen with 1280x800 resolution, to a single external monitor with 1920x1080.

Note that all of this works fine in 10.10. And also note that all of my testing is in the Gnome Classic Desktop.

Okay, one more interesting tidbit...

I am actually unable to run *any* resolution on my external monitor, if I disable my laptop LCD in gnome's monitor configuration utility. Seems that something is very broken there...

I installed Natty Beta2 this weekend on my HP laptop. I run it with an Acer 22" external monitor. I prefer the big screen (to the left) as primary and the laptop screen (to the right) for email and IM.

I only had to edit the ~/.config/monitors.xml to set the VGA screen to primary to get the setup I needed. After some updates this weekend, i can no longer get the external monitor to work with Unity. On bootup I get both screens with the login screen on the laptop. If I choose Unity, the external monitor freaks out like this...

http://www.youtube.com/watch?v=pK5AfUzg6UA

Mean while the laptop screen has a freely moving mouse, but not much else. If I go into Ubuntu classic mode, I get Gnome 2 and the resolutions work just fine. When I try Unity again it freaks out.

I've tried deleting the ~/.config and restarting. That seemed to have helped......until I restarted. The odd thing is that somehow my "left is primary" was retained even without the monitor.xml. When the view DOES come up it looks like this...

http://www.flickr.com/photos/tubatodd/5657189985/

> I am actually unable to run *any* resolution on my external monitor, if I disable my laptop LCD in gnome's monitor configuration utility.

This is what I am seeing as well, in bug 767319(which is marked as duplicate, though slightly different than this bug's description).

https://bugs.launchpad.net/bugs/767319

This bug is critical for anyone docking a laptop or solely using an external monitor.

All,

Note that I've filed bug [1] with updated information. On such bug, Bryce pointed out that a newer kernel should contain patches fixing this issue. After installing the newer kernel in the Ubuntu's daily builds the issue was fixed.

Unfortunately, I've been unable to bisect the patches to determine which one is fixing this issue. IS someone has a bit a free time and can help with this, it would be awesome. See more information in [1].

Thank you!

[1]: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/761236

Right, and 761236 is looking like a dupe of a common Arrandale bug, #745112. That in turn *might* be a sub-case of bug #761065 which affects late-model cards and for which a kernel fix is being verified.

So, as to the disposition of this bug report: We've factored out the single-monitor issue Andres had, so what remains is the multi-monitor case, where total X width is >4096. As mentioned above in comment #15, the X portion of that issue is essentially just bug #555641 and not something we need to track here. The bug watch on the upstream FDO bug appears to be focusing on that aspect of this bug.

After factoring out those two X bugs, the one remaining bit is that gnome-display-properties is failing to check for screen width >4096 when compiz is running. Basically, gnome-desktop must include a check that target resolution <= max texture width if compiz is enabled.

Similar problem with lenovoT410. when I wake up or reboot my laptop the X session doesn't start completely.

Not much to add besides "me too", but I can specifically confirm that using xrandr(1) does NOT trigger the problem, whereas using the GNOME Monitors applet does trigger the problem. Also, the GNOME Monitors applet produces some very bizarre geometry at times (half-height LVDS screen, typically), for which I will probably open another bug soon.

Sorry, forgot to add: this is in 2D "classic" mode, NOT with compiz enabled, so I'm not hitting the >4096px problem. Nor do my two screens together add up to >4096px anyway, so this is not merely the Arrandale hardware limit that I'm hitting.

I guess I am affected by this bug too. I have a Dell D820 with Intel Mobile GM965/GL960 Integrated Graphics Controller. I use a docking station to which 2 identical monitors HP LA1951g are hooked, one DVI, one VGA. The monitors "native resolution" is 1280 x 1024 at 60 Hz.

With Maverick I had to create a xorg.conf file to get both to work at native resolution in "extended desktop" setup (so virtual screen of 2560 x 1024) - but at least it worked fine (LVDS was shut off). The fact that compiz was off didn't bother me at all.

With Natty ... oh boy ! Not a chance (all effect were off). Natty seems to ignore xorg.conf file completely. Moreover, it's not even offering/detecting the native 1280 x 1024 resolution on the VGA display (?!?!?) any more. The max resolution available on it with Natty is 1024 x 768 ...

I am also hit by bug #745112.

The Intel chipset video support is so badly broken on Natty that I had to go back to Maverick.

Same annoying problem for me on HP ProBook 4520s with IntelHD graphics card.

This bug is still affecting me in Oneiric. Adding an oneiric task.

Boy, I sure would like to see this solved for Oneiric...

Actually, rather than tasking oneiric on this bug report (which is a bit stale at this point) I would recommend opening a new bug report, because so much has changed in arrandale support in oneiric. For instance, mesa now supports 8k textures on arrandale, so the issue I alluded to still remaining should now no longer be a problem. There could still be external monitor configuration troubles (they seem a dime a dozen), but it would be easiest to handle them as separate bugs.

I take it all back. I now have an arrandale laptop on hand for testing and am able to reproduce the bug exactly as originally reported. I'll do some diagnostics and report back.

While in broken state:

bryce@clanfield:~$ xrandr --auto
xrandr: cannot find crtc for output LVDS1
bryce@clanfield:~$ xrandr --verbose
Screen 0: minimum 320 x 200, current 1920 x 1080, maximum 8192 x 8192
LVDS1 connected (normal left inverted right x axis y axis)
 Identifier: 0x41
 Timestamp: 15385367
 Subpixel: horizontal rgb
 Clones:
 CRTCs: 0
 Transform: 1.000000 0.000000 0.000000
             0.000000 1.000000 0.000000
             0.000000 0.000000 1.000000
            filter:
 EDID:
  00ffffffffffff0006af2c1100000000
  01120103901d10780af9d59555549327
  21505400000001010101010101010101
  010101010101121b565a500019303020
  360025a41000001a0c12565a50001930
  3020360025a41000001a000000fe0058
  58333147804231333358573100000000
  00000000000000000001010a20200076
 BACKLIGHT: 0 (0x00000000) range: (0,15)
 Backlight: 0 (0x00000000) range: (0,15)
 scaling mode: Full aspect
  supported: None Full Center Full aspect
  1366x768 (0x44) 69.3MHz +HSync -VSync +preferred
        h: width 1366 start 1414 end 1446 total 1456 skew 0 clock 47.6KHz
        v: height 768 start 771 end 777 total 793 clock 60.0Hz
  1366x768 (0x45) 46.2MHz +HSync -VSync
        h: width 1366 start 1414 end 1446 total 1456 skew 0 clock 31.7KHz
        v: height 768 start 771 end 777 total 793 clock 40.0Hz
  1360x768 (0x46) 84.8MHz -HSync +VSync
        h: width 1360 start 1432 end 1568 total 1776 skew 0 clock 47.7KHz
        v: height 768 start 771 end 781 total 798 clock 59.8Hz
  1360x768 (0x47) 72.0MHz +HSync -VSync
        h: width 1360 start 1408 end 1440 total 1520 skew 0 clock 47.4KHz
        v: height 768 start 771 end 781 total 790 clock 60.0Hz
  1024x768 (0x48) 65.0MHz -HSync -VSync
        h: width 1024 start 1048 end 1184 total 1344 skew 0 clock 48.4KHz
        v: height 768 start 771 end 777 total 806 clock 60.0Hz
  800x600 (0x49) 40.0MHz +HSync +VSync
        h: width 800 start 840 end 968 total 1056 skew 0 clock 37.9KHz
        v: height 600 start 601 end 605 total 628 clock 60.3Hz
  800x600 (0x4a) 36.0MHz +HSync +VSync
        h: width 800 start 824 end 896 total 1024 skew 0 clock 35.2KHz
        v: height 600 start 601 end 603 total 625 clock 56.2Hz
  640x480 (0x4b) 25.2MHz -HSync -VSync
        h: width 640 start 656 end 752 total 800 skew 0 clock 31.5KHz
        v: height 480 start 490 end 492 total 525 clock 59.9Hz
VGA1 connected 1920x1080+0+0 (0x4c) normal (normal left inverted right x axis y axis) 477mm x 268mm
 Identifier: 0x42
 Timestamp: 15385367
 Subpixel: unknown
 Gamma: 1.0:1.0:1.0
 Brightness: 1.0
 Clones:
 CRTC: 0
 CRTCs: 1 0
 Transform: 1.000000 0.000000 0.000000
             0.000000 1.000000 0.000000
             0.000000 0.000000 1.000000
            filter:
 EDID:
  00ffffffffffff0010ac51a053393631
  0b1401030e301b78eeee95a3544c9926
  0f5054a54b00714f8180d1c001010101
  010101010101023a801871382d40582c
  4500dd0c1100001e000000ff004b3732
  3650303339313639530a000000fc0044
  454c4c205358...

Working state

bryce@clanfield:~$ xrandr --verbose
Screen 0: minimum 320 x 200, current 1920 x 1848, maximum 8192 x 8192
LVDS1 connected 1366x768+231+1080 (0x44) normal (normal left inverted right x axis y axis) 293mm x 164mm
 Identifier: 0x41
 Timestamp: 15385367
 Subpixel: horizontal rgb
 Gamma: 1.0:1.0:1.0
 Brightness: 1.0
 Clones:
 CRTC: 0
 CRTCs: 0
 Transform: 1.000000 0.000000 0.000000
             0.000000 1.000000 0.000000
             0.000000 0.000000 1.000000
            filter:
 EDID:
  00ffffffffffff0006af2c1100000000
  01120103901d10780af9d59555549327
  21505400000001010101010101010101
  010101010101121b565a500019303020
  360025a41000001a0c12565a50001930
  3020360025a41000001a000000fe0058
  58333147804231333358573100000000
  00000000000000000001010a20200076
 BACKLIGHT: 0 (0x00000000) range: (0,15)
 Backlight: 0 (0x00000000) range: (0,15)
 scaling mode: Full aspect
  supported: None Full Center Full aspect
  1366x768 (0x44) 69.3MHz +HSync -VSync *current +preferred
        h: width 1366 start 1414 end 1446 total 1456 skew 0 clock 47.6KHz
        v: height 768 start 771 end 777 total 793 clock 60.0Hz
  1366x768 (0x45) 46.2MHz +HSync -VSync
        h: width 1366 start 1414 end 1446 total 1456 skew 0 clock 31.7KHz
        v: height 768 start 771 end 777 total 793 clock 40.0Hz
  1360x768 (0x46) 84.8MHz -HSync +VSync
        h: width 1360 start 1432 end 1568 total 1776 skew 0 clock 47.7KHz
        v: height 768 start 771 end 781 total 798 clock 59.8Hz
  1360x768 (0x47) 72.0MHz +HSync -VSync
        h: width 1360 start 1408 end 1440 total 1520 skew 0 clock 47.4KHz
        v: height 768 start 771 end 781 total 790 clock 60.0Hz
  1024x768 (0x48) 65.0MHz -HSync -VSync
        h: width 1024 start 1048 end 1184 total 1344 skew 0 clock 48.4KHz
        v: height 768 start 771 end 777 total 806 clock 60.0Hz
  800x600 (0x49) 40.0MHz +HSync +VSync
        h: width 800 start 840 end 968 total 1056 skew 0 clock 37.9KHz
        v: height 600 start 601 end 605 total 628 clock 60.3Hz
  800x600 (0x4a) 36.0MHz +HSync +VSync
        h: width 800 start 824 end 896 total 1024 skew 0 clock 35.2KHz
        v: height 600 start 601 end 603 total 625 clock 56.2Hz
  640x480 (0x4b) 25.2MHz -HSync -VSync
        h: width 640 start 656 end 752 total 800 skew 0 clock 31.5KHz
        v: height 480 start 490 end 492 total 525 clock 59.9Hz
VGA1 connected 1920x1080+0+0 (0x4c) normal (normal left inverted right x axis y axis) 477mm x 268mm
 Identifier: 0x42
 Timestamp: 15385367
 Subpixel: unknown
 Gamma: 1.0:1.0:1.0
 Brightness: 1.0
 Clones:
 CRTC: 1
 CRTCs: 1 0
 Transform: 1.000000 0.000000 0.000000
             0.000000 1.000000 0.000000
             0.000000 0.000000 1.000000
            filter:
 EDID:
  00ffffffffffff0010ac51a053393631
  0b1401030e301b78eeee95a3544c9926
  0f5054a54b00714f8180d1c001010101
  010101010101023a801871382d40582c
  4500dd0c1100001e000000ff004b3732
  3650303339313639530...

After using xrandr to disable lvds1:

Screen 0: minimum 320 x 200, current 1920 x 1080, maximum 8192 x 8192
LVDS1 connected (normal left inverted right x axis y axis)
 Identifier: 0x41
 Timestamp: 32659
 Subpixel: horizontal rgb
 Clones:
 CRTCs: 1
 Transform: 1.000000 0.000000 0.000000
             0.000000 1.000000 0.000000
             0.000000 0.000000 1.000000
            filter:
 EDID:
  00ffffffffffff0006af2c1100000000
  01120103901d10780af9d59555549327
  21505400000001010101010101010101
  010101010101121b565a500019303020
  360025a41000001a0c12565a50001930
  3020360025a41000001a000000fe0058
  58333147804231333358573100000000
  00000000000000000001010a20200076
 BACKLIGHT: 0 (0x00000000) range: (0,15)
 Backlight: 0 (0x00000000) range: (0,15)
 scaling mode: Full aspect
  supported: None Full Center Full aspect
  1366x768 (0x44) 69.3MHz +HSync -VSync +preferred
        h: width 1366 start 1414 end 1446 total 1456 skew 0 clock 47.6KHz
        v: height 768 start 771 end 777 total 793 clock 60.0Hz
  1366x768 (0x45) 46.2MHz +HSync -VSync
        h: width 1366 start 1414 end 1446 total 1456 skew 0 clock 31.7KHz
        v: height 768 start 771 end 777 total 793 clock 40.0Hz
  1360x768 (0x46) 84.8MHz -HSync +VSync
        h: width 1360 start 1432 end 1568 total 1776 skew 0 clock 47.7KHz
        v: height 768 start 771 end 781 total 798 clock 59.8Hz
  1360x768 (0x47) 72.0MHz +HSync -VSync
        h: width 1360 start 1408 end 1440 total 1520 skew 0 clock 47.4KHz
        v: height 768 start 771 end 781 total 790 clock 60.0Hz
  1024x768 (0x48) 65.0MHz -HSync -VSync
        h: width 1024 start 1048 end 1184 total 1344 skew 0 clock 48.4KHz
        v: height 768 start 771 end 777 total 806 clock 60.0Hz
  800x600 (0x49) 40.0MHz +HSync +VSync
        h: width 800 start 840 end 968 total 1056 skew 0 clock 37.9KHz
        v: height 600 start 601 end 605 total 628 clock 60.3Hz
  800x600 (0x4a) 36.0MHz +HSync +VSync
        h: width 800 start 824 end 896 total 1024 skew 0 clock 35.2KHz
        v: height 600 start 601 end 603 total 625 clock 56.2Hz
  640x480 (0x4b) 25.2MHz -HSync -VSync
        h: width 640 start 656 end 752 total 800 skew 0 clock 31.5KHz
        v: height 480 start 490 end 492 total 525 clock 59.9Hz
VGA1 connected 1920x1080+0+0 (0x4c) normal (normal left inverted right x axis y axis) 477mm x 268mm
 Identifier: 0x42
 Timestamp: 32659
 Subpixel: unknown
 Gamma: 1.0:1.0:1.0
 Brightness: 1.0
 Clones:
 CRTC: 0
 CRTCs: 0 1
 Transform: 1.000000 0.000000 0.000000
             0.000000 1.000000 0.000000
             0.000000 0.000000 1.000000
            filter:
 EDID:
  00ffffffffffff0010ac51a053393631
  0b1401030e301b78eeee95a3544c9926
  0f5054a54b00714f8180d1c001010101
  010101010101023a801871382d40582c
  4500dd0c1100001e000000ff004b3732
  3650303339313639530a000000fc0044
  454c4c20535832323130540a000000fd
  00384c1e5311000a202020202020008b
  1920x1080 (0x4c) 148.5MHz +HSync +VSync *cur...

This prints out whenever I do a screen configuration change:

[15844.649388] dell_wmi: Received unknown WMI event (0x11)

Steps to reproduce for me are slightly different from OP:

1. Connect external monitor to VGA output.
2. Launch gnome-display-properties and make sure LVDS1 is on, VGA1 is off.
3. Turn LVDS1 off, VGA1 on
4. LVDS1 turns off, and VGA1 shows linear corruption pattern (different than the corruption in original post)

If I e.g. enable both monitors and then disable LVDS1, it works properly.

Still can't reproduce using xrandr. This works fine:

bryce@clanfield:~$ xrandr --output VGA1 --off
bryce@clanfield:~$ xrandr --output VGA1 --auto --output LVDS1 --off
bryce@clanfield:~$ xrandr --output VGA1 --off --output LVDS1 --auto
bryce@clanfield:~$ xrandr --output VGA1 --auto --output LVDS1 --off

[ 417.493074] [drm:i915_get_vblank_timestamp], crtc 0 is disabled
[ 417.504230] [drm:i915_driver_irq_handler], pipe A underrun
[ 417.509705] [drm:i915_get_vblank_timestamp], crtc 0 is disabled
[ 417.520990] [drm:i915_driver_irq_handler], pipe A underrun

[ 1301.290842] [drm:i915_driver_irq_handler], pipe A underrun
[ 1301.307600] [drm:i915_driver_irq_handler], pipe A underrun
[ 1301.309934] [drm:drm_mode_getconnector], [CONNECTOR:5:?]
[ 1301.309944] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:5:LVDS-1]
[ 1301.309967] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:5:LVDS-1] probed modes :
[ 1301.309972] [drm:drm_mode_debug_printmodeline], Modeline 18:"1366x768" 60 69300 1366 1414 1446 1456 768 771 777 793 0x48 0x9
[ 1301.309981] [drm:drm_mode_debug_printmodeline], Modeline 19:"1366x768" 40 46200 1366 1414 1446 1456 768 771 777 793 0x40 0x9
[ 1301.309991] [drm:drm_mode_getconnector], [CONNECTOR:5:?]
[ 1301.310261] [drm:drm_mode_getconnector], [CONNECTOR:12:?]
[ 1301.310267] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:12:VGA-1]
[ 1301.324067] [drm:intel_crt_detect], CRT detected via hotplug
[ 1301.324322] [drm:i915_driver_irq_handler], pipe A underrun
[ 1301.341081] [drm:i915_driver_irq_handler], pipe A underrun
[ 1301.357843] [drm:i915_driver_irq_handler], pipe A underrun
[ 1301.374604] [drm:i915_driver_irq_handler], pipe A underrun
[ 1301.376460] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:12:VGA-1] probed modes :
[ 1301.376467] [drm:drm_mode_debug_printmodeline], Modeline 9:"1920x1080" 60 148500 1920 2008 2052 2200 1080 1084 1089 1125 0x48 0x5
[ 1301.376475] [drm:drm_mode_debug_printmodeline], Modeline 26:"1280x1024" 75 135000 1280 1296 1440 1688 1024 1025 1028 1066 0x40 0x5
[ 1301.376483] [drm:drm_mode_debug_printmodeline], Modeline 21:"1280x1024" 60 108000 1280 1328 1440 1688 1024 1025 1028 1066 0x40 0x5
[ 1301.376490] [drm:drm_mode_debug_printmodeline], Modeline 20:"1152x864" 75 108000 1152 1216 1344 1600 864 865 868 900 0x40 0x5
[ 1301.376498] [drm:drm_mode_debug_printmodeline], Modeline 27:"1024x768" 75 78800 1024 1040 1136 1312 768 769 772 800 0x40 0x5
[ 1301.376506] [drm:drm_mode_debug_printmodeline], Modeline 28:"1024x768" 60 65000 1024 1048 1184 1344 768 771 777 806 0x40 0xa
[ 1301.376513] [drm:drm_mode_debug_printmodeline], Modeline 29:"800x600" 75 49500 800 816 896 1056 600 601 604 625 0x40 0x5
[ 1301.376520] [drm:drm_mode_debug_printmodeline], Modeline 22:"800x600" 60 40000 800 840 968 1056 600 601 605 628 0x40 0x5
[ 1301.376528] [drm:drm_mode_debug_printmodeline], Modeline 23:"640x480" 75 31500 640 656 720 840 480 481 484 500 0x40 0xa
[ 1301.376535] [drm:drm_mode_debug_printmodeline], Modeline 24:"640x480" 60 25200 640 656 752 800 480 490 492 525 0x40 0xa
[ 1301.376542] [drm:drm_mode_debug_printmodeline], Modeline 25:"720x400" 70 28320 720 738 846 900 400 412 414 449 0x40 0x6
[ 1301.376587] [drm:drm_mode_getconnector], [CONNECTOR:12:?]
[ 1301.377085] [drm:drm_mode_getconnector], [CONNECTOR:14:?]
[ 1301.377093] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:14:DP-1]
[ 1301.377598] [drm:intel_dp_aux_ch], dp_aux_ch timeout status 0x51450064
[ 1301.377603] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:14:DP-1] disconnected
[ 1301.377610] [drm:drm_mode_getconnector], [CONNECTOR:14:?]
[ 1301.377614] [drm:drm_helper_probe_single_connector_modes], [CONNECTOR:14:DP-1]
[ 1301.378117] [drm:intel_dp_aux_ch], dp_aux_c...

Alright, after poking around a bit I believe this is the same thing as bug #828623, which is reported upstream so I'll dupe to that.