Invert screen causes X restart

[Chris Gregan]

<Build: UME Beta 4>
<Environment: Jax10>

Summary:
When inverting the screen X restarts after a short period of inverted screen.

Steps to Reproduce:
1) Browse to preferences
2) Select Screen Resolution
3) Select Inverted in the Rotation drop down
4) Apply

Expected result:
Screen appears rotated 180 degrees

Actual result:
X crashes and restarts

[Steve Kowalik (stevenk)]

Confirmed on McCaslin 20080506, too.

[Steve Kowalik (stevenk)]

Hi Bryce,

I was wondering if you could please look at this bug.

[Bryce Harrington (bryce)]

Is there any error message in your /var/log/Xorg.0.log[.old]?

[Bryce Harrington (bryce)]

Also, please test doing rotation using `xrandr --output <OUTPUT> --rotate inverted and see if you can replicate the effect.

[Steve Kowalik (stevenk)]

There is no obvious error message in Xorg.0.log.old, but I am attaching to see if you can distill something.

[Cherry (sorata118)]

I tried to use xrandr to resize or rotate the screen. It also crashed the X and restarted.

"dmesg |tail -n 30", got information as follow:

[ 3114.628213] [drm] RGB0-1: set mode 640x480 6
[ 3123.082039] [drm] Resetting scheduler.
[ 3123.085995] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 3164.317023] [drm] RGB0-1: set mode 640x480 6
[ 3164.742266] [drm] RGB0-1: set mode 640x480 6
[ 3173.134007] [drm] Resetting scheduler.
[ 3173.137958] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 3233.157089] [drm] RGB0-1: set mode 640x480 6
[ 3233.582307] [drm] RGB0-1: set mode 640x480 6
[ 3242.063864] [drm] Resetting scheduler.
[ 3242.067797] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 3340.322920] [drm] RGB0-1: set mode 640x480 6
[ 3340.748113] [drm] RGB0-1: set mode 640x480 6
[ 3349.121095] [drm] Resetting scheduler.
[ 3349.125032] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 3600.317841] hildon-desktop[7915]: segfault at 00000000 eip 00000000 esp bf8f244c error 4
[ 3602.153705] [drm] RGB0-1: set mode 640x480 6
[ 3602.578824] [drm] RGB0-1: set mode 640x480 6
[ 3611.773029] [drm] Resetting scheduler.
[ 3611.777024] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 3642.210648] hildon-desktop[8073]: segfault at 00000000 eip 00000000 esp bfdda92c error 4
[ 3643.197715] [drm] RGB0-1: set mode 640x480 6
[ 3643.623330] [drm] RGB0-1: set mode 640x480 6
[ 3651.958472] [drm] Resetting scheduler.
[ 3651.962413] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 3696.462931] hildon-desktop[8227]: segfault at 00000000 eip 00000000 esp bf9eed3c error 4
[ 3697.383125] [drm] RGB0-1: set mode 640x480 6
[ 3697.809311] [drm] RGB0-1: set mode 640x480 6
[ 3706.961678] [drm] Resetting scheduler.
[ 3706.965664] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.

[Loïc Minier (lool)]

There's a crash on "xrandr --auto" when the VGA is plugged; it's likely to be a similar issue.

(gdb) bt
#0 0x00000000 in ?? ()
#1 0xb77df8c2 in tasknavigator_item_screen_changed (screen=0x8094100,.
    item=0x84b4000) at tasknavigator-item.c:57
#2 0xb78b0b0c in g_cclosure_marshal_VOID__VOID ()
   from /usr/lib/libgobject-2.0.so.0
#3 0xb78a466c in g_closure_invoke () from /usr/lib/libgobject-2.0.so.0
#4 0xb78b7bcf in ?? () from /usr/lib/libgobject-2.0.so.0
#5 0x084b25a0 in ?? ()
#6 0x00000000 in ?? ()

Bill had a look and will upload a check for some NULL function pointers; thanks Bill!

[Bill Filler (bfiller)]

The following stack trace was provided by lool:

#0 0x00000000 in ?? ()
#1 0xb77df8c2 in tasknavigator_item_screen_changed (screen=0x8094100,
    item=0x84b4000) at tasknavigator-item.c:57
#2 0xb78b0b0c in g_cclosure_marshal_VOID__VOID ()
   from /usr/lib/libgobject-2.0.so.0
#3 0xb78a466c in g_closure_invoke () from /usr/lib/libgobject-2.0.so.0
#4 0xb78b7bcf in ?? () from /usr/lib/libgobject-2.0.so.0
#5 0x084b25a0 in ?? ()
#6 0x00000000 in ?? ()

[Bill Filler (bfiller)]

I fixed the function task_navigator_item_screen_changed() in hildon-desktop and pushed a new version to the UME PPA. I've attached the modified patch here.

[Bill Filler (bfiller)]

please retest and verify the crash is fixed.

[Loïc Minier (lool)]

Fixed with Bill's changes in 1:2.0.11-1~svn15367-0ubuntu3 .

[Cherry (sorata118)]

I upgraded the hildon-desktop to 1:2.0.11-1~svn15367-0ubuntu3. It still crashs.
"dmesg | tail -n 30" got as follow:
[ 59.868209] [drm] LVDSGetPWMMaxBacklight: the max_pwm_blc is 31250.
[ 59.868228] [drm] fb0: psbfb frame buffer device
[ 59.868239] [drm] intel_lvds_prepare
[ 59.868243] [drm] intel_lvds_set_power: 0
[ 59.868246] [drm] intel_lvds_set_backlight: the level is 0
[ 59.868251] [drm] LVDSGetPWMMaxBacklight: the max_pwm_blc is 31250.
[ 59.928962] [drm] LVDS: set mode 1024x600 8
[ 59.929021] [drm] intel_lvds_commit
[ 59.929024] [drm] intel_lvds_set_power: 1
[ 60.368854] [drm] intel_lvds_set_backlight: the level is 100
[ 60.368859] [drm] LVDSGetPWMMaxBacklight: the max_pwm_blc is 31250.
[ 60.519747] [drm] Initialized psb 4.0.0 20080125 on minor 0
[ 60.750180] NET: Registered protocol family 17
[ 72.164868] [drm:psb_watchdog_func] *ERROR* Detected Poulsbo Scheduler Lockup.
[ 73.027909] eth0: no IPv6 routers present
[ 73.188362] [drm] Resetting scheduler.
[ 73.191247] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.
[ 100.562389] hildon-desktop[6274]: segfault at 00000000 eip 00000000 esp bf9137fc error 4
[ 102.138885] [drm] intel_lvds_prepare
[ 102.138893] [drm] intel_lvds_set_power: 0
[ 102.138897] [drm] intel_lvds_set_backlight: the level is 0
[ 102.138902] [drm] LVDSGetPWMMaxBacklight: the max_pwm_blc is 31250.
[ 102.249805] [drm] LVDS: set mode 1024x600 a
[ 102.249868] [drm] intel_lvds_commit
[ 102.249872] [drm] intel_lvds_set_power: 1
[ 102.623801] [drm] intel_lvds_set_backlight: the level is 100
[ 102.623806] [drm] LVDSGetPWMMaxBacklight: the max_pwm_blc is 31250.
[ 111.652483] [drm:psb_watchdog_func] *ERROR* Detected Poulsbo Scheduler Lockup.
[ 112.675848] [drm] Resetting scheduler.
[ 112.678744] [drm:psb_scheduler_reset] *ERROR* Detected Poulsbo rasterizer lockup.

[Loïc Minier (lool)]

It works just fine on the Q1U; at this point this is a Poulsbo driver issue.

[Loïc Minier (lool)]

@Inuka, could you please take a look?

[cathyli (cathy-li)]

I tried in hildon-desktop to 1:2.0.11-1~svn15367-0ubuntu3, and applets 0.63, Invert screen work well in Q1U and CB which connected LCD kit.

[Loïc Minier (lool)]

Cool, closing

[Moblin_cracked (shilpa-jabin)]

Hi, with patch 21_multiple_resolution it crashes on open of screen resolution window. Please look into this. its still a bug.

[Loïc Minier (lool)]

@Moblin_cracked: what's your hildon-desktop version?

I can rotate in all directions and back to normal with either poulsbo or intel on crownbeach and q1u respectively.

[mobishrek (arunsbangari)]

Hi, I have seen this crash with 1:2.0.11-1~svn15367-0ubuntu1.. I am a newbie and have limited knowledge on moblin and linux..I see this crash when i apply 21_multiple_resolution patch and if i take out this patch, the screen resolution applet shows up with rotate option disabled..What command do i use to get the logs..? When i click on the screen resolution applet, X crashes..How to get out from this state? All i have to do is press CTRL+ALT+DEL to restart..

Thanks...

[Loïc Minier (lool)]

The ubuntu-mobile ppa has version 1:2.0.11-1~svn15367-0ubuntu4 which doesn't have this issue.

[mobishrek (arunsbangari)]

Thanks Loïc Minier. I will check out that version and update my comments...

[wilbur (wilbur-unail)]

updating to 2.0.11-1~svn15367-0ubuntu4, x still crash when I use either command xrandr or screen-resolution applet.

a similar bug caused by xrandr(https://bugs.launchpad.net/ubuntu/+source/xrandr/+bug/104978). have a look

any body have same experience? or should I do some config changes?

[wilbur (wilbur-unail)]

my xorg.conf is attached

[Loïc Minier (lool)]

wilbur: this bug was reported on jax hardware and generally poulsbo graphics and hildon-desktop; it's fixed in these configs for us. Your bug is probably a different one, please open a new bug and describe your test environment. If you're using vmware, qemu, or kvm, mention that in the new bug report.