web page -> X server crash/blank, no display until reboot [Q965]

[Mike Coleman]

Binary package hint: xorg

In Ubuntu 8.04, with 1:7.3+10ubuntu10, on an Intel 965Q display, I'm getting a crash or black screen when I click on an MP3 file link in Firefox (3 beta 5). Once the screen is blank, it appears that the machine is still running, and I *think* a C-A-Backspace is resetting the X server, but it won't come back. Nothing I do short of a reboot will restore the display.

I'm operating this display rotated left 90 degrees, which is a feature that only really started working in Ubuntu 8, so perhaps that's involved.

Attached is an X server log that is (I believe) of the attempted restart, which fails. The notable lines at the end are

(WW) intel(0): PRB0_CTL (0x0001f801) indicates ring buffer enabled
(WW) intel(0): PRB0_HEAD (0x0100895c) and PRB0_TAIL (0x00008988) indicate ring buffer not flushed
(WW) intel(0): Existing errors found in hardware state.

Apparently this stymies the X server. There is also a suspicious line in /var/log/messages at the time of the blank/crash

May 1 13:45:03 delta kernel: [791787.781107] mtrr: type mismatch for c0000000,10000000 old: write-back new: write-combining

[unggnu (unggnu)]

Could you please recheck if the problem appears without rotation too? I don't think that the mp3 file causes this but Totem instead. Could you try to disable Visualization in Totem (start Totem - Edit - Preferences - Display - disable visual .." and try it again with rotation of course?

[unggnu (unggnu)]

Btw. is it possible to ping your computer after the crash?

[Mike Coleman (tutufan)]

The problem seems not to occur without rotation. Disabling visualization appears to have no effect (the display still blanks). The machine is pingable after the display blanks--I suspect that it's still running, except for the X display being unusable.

When I clicked on the link without rotation, a totem plugin started up, but I don't think the mp3 itself began playing, and I saw no graphics other than the minimal UI (this was prior to disabling visualization). I think it was probably just sitting there waiting for me to press the "play" button. So, it seems like visualiation is not the cause, but rather whatever basic X things the totem plugin does. (?)

More background: Rotation on this display didn't work correctly until this version of Ubuntu. Previously it'd sort of work, but there was lots of screen crap and the server would crash after being used for a bit.

Also, if I'm on a rotated display and I use the "Screen Resolution" widget to try to unrotate it, the display goes into a flashing unusable state until reboot. It does appear, though, that I can log out of a rotated display and into an unrotated display with no trouble.

[Mike Coleman (tutufan)]

Playing with it some more, I'm able to reproduce the problem with the totem application by itself, with a local mp3 file. It seems to hang the rotated X server *some* of the time when visualization is turned on and never when visualization is turned off.

Without rotation, it does not hang, but the "goom" visualization does not show anything, whereas the monoscope visualization does, if that matters. Either seems to be able to crash the rotated server.

[unggnu (unggnu)]

Could you please recheck it with the current upstream -intel driver. If it doesn't work with it we can report the issue upstream. The easiest way is to use the Debian sid driver in Hardy I guess. You have to remove the old i810 driver because the new one doesn't work parallel.
sudo apt-get remove xserver-xorg-video-i810 xserver-xorg-video-all
Download the current driver from http://mirrors.kernel.org/debian/pool/main/x/xserver-xorg-video-intel/xserver-xorg-video-intel_2.3.1-1_i386.deb and install it.
After that restart X and check if your problem is gone. If you have made changes to your xorg.conf generate a new one please (sudo dpkg-reconfigure xserver-xorg) or just remove it to be sure.

If you want to reset everything do the follow commands.
sudo apt-get remove xserver-xorg-video-intel
sudo apt-get install xserver-xorg-video-all

Afterwards you will have the standard hardy driver again.

Btw. I guess every OpenGL app like glxgears could possibly crash X while rotated.

[Mike Coleman (tutufan)]

I finally got to try this, though it was on a fresh installation of Ubuntu 8.04 LTS, rather than the original machine. Unfortunately, with this new intel package (2.3.1-1), the server hangs up right after being rotated (the screen is drawn and the cursor "works", and the cursor bitmap changes as it's moved around the screen, but beyond that, everything's frozen).

So, at this point, this seems to be a larger bug, which is that the Intel X server isn't stable under rotation. It mostly was for the version I had (which I unfortunately didn't make not of, probably 2.2.X), but it's not in 2.3.1.

(Aside: I tried installing Debian 4.0 on this machine, but X won't even start up because it doesn't like my screen size (1920x1200), so Ubuntu still seems to be ahead in that regard.)

[Florian Kisser (flok)]

An hour ago my xserver crashed, while I myself was doing nothing, but I guess that in the background the running firefix instance reloaded a page containing flash. From a friend I know how easily flash can crash the xserver... That's the only explanation I can give.

However, this bug is the only one with a similar log to mine, so I decided to post it. I hope it might help.

I have an Intel GMA950 with driver version 2.3.1 from the debian repos, since the ubuntu version used to crash in xv, because of stolen memory. And Adobe Flash 10, because version 9 crashed firefox all the time.

[unggnu (unggnu)]

I guess it is the best to get a full backtrace. See http://wiki.ubuntu.com/X/Backtracing for directions on how to do this.

[unggnu (unggnu)]

Does X crashes if your resize the glxgears window?

[Mike Coleman (tutufan)]

Unfortunately I have no way to reproduce this at the moment, as
display rotation no longer works under the current version. (The
server doesn't crash--it just isn't usable.)

[Florian Kisser (flok)]

I didn't install the dbg symbols of version 2.3.1, sry.
I just grabbed version 2.3.2 together with dbg from the debian repos. If the bug occures again, I'll post a full bt.
But I hope I won't have to ;)

[Antti Kaihola (akaihola)]

I'm experiencing the same issue using the 2.2.1 version of the Intel driver on a Dell Inspiron 500m. Everything works in normal orientation, but when I rotate left, Totem (and the Firefox plugin) hangs the system when it's trying to display either a video or a visualizer.

I tried to change the video output from v4l2 to v4l with gstreamer-properties, but it didn't make any difference.

I'm currently running a multi-head setup, but the problem occurs also when using only the laptop's own TFT.

Here's output from lspci and xdpyinfo:

# lspci -vvvv
00:02.1 Display controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)
 Subsystem: Dell Unknown device 0153
 Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
 Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR-
 Latency: 0
 Region 0: Memory at e8000000 (32-bit, prefetchable) [size=128M]
 Region 1: Memory at faf00000 (32-bit, non-prefetchable) [size=512K]
 Capabilities: [d0] Power Management version 1
  Flags: PMEClk- DSI+ D1+ D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
  Status: D0 PME-Enable- DSel=0 DScale=0 PME-

$ xdpyinfo
name of display: :0.0
version number: 11.0
vendor string: The X.Org Foundation
vendor release number: 10400090
X.Org version: 1.4.0.90
maximum request size: 16777212 bytes
motion buffer size: 256
bitmap unit, bit order, padding: 32, LSBFirst, 32
image byte order: LSBFirst
number of supported pixmap formats: 7
supported pixmap formats:
    depth 1, bits_per_pixel 1, scanline_pad 32
    depth 4, bits_per_pixel 8, scanline_pad 32
    depth 8, bits_per_pixel 8, scanline_pad 32
    depth 15, bits_per_pixel 16, scanline_pad 32
    depth 16, bits_per_pixel 16, scanline_pad 32
    depth 24, bits_per_pixel 32, scanline_pad 32
    depth 32, bits_per_pixel 32, scanline_pad 32
keycode range: minimum 8, maximum 255
focus: window 0x3a00022, revert to Parent
number of extensions: 33
    BIG-REQUESTS
    Composite
    DAMAGE
    DOUBLE-BUFFER
    DPMS
    Extended-Visual-Information
    GLX
    MIT-SCREEN-SAVER
    MIT-SHM
    MIT-SUNDRY-NONSTANDARD
    RANDR
    RECORD
    RENDER
    SECURITY
    SGI-GLX
    SHAPE
    SYNC
    TOG-CUP
    X-Resource
    XAccessControlExtension
    XC-APPGROUP
    XC-MISC
    XFIXES
    XFree86-Bigfont
    XFree86-DGA
    XFree86-DRI
    XFree86-Misc
    XFree86-VidModeExtension
    XINERAMA
    XInputExtension
    XKEYBOARD
    XTEST
    XVideo
default screen number: 0
number of screens: 1

screen #0:
  dimensions: 2024x1400 pixels (535x370 millimeters)
  resolution: 96x96 dots per inch
  depths (7): 24, 1, 4, 8, 15, 16, 32
  root window id: 0x5a
  depth of root window: 24 planes
  number of colormaps: minimum 1, maximum 1
  default colormap: 0x20
  default number of colormap cells: 256
  preallocated pixels: black 0, white 16777215
  options: backing-store NO, save-unders NO
  largest cursor: 64x64
  current input event mask: 0xfa8033
    KeyPressMask KeyReleaseMask EnterWindowMask
    LeaveWindowMask ExposureMask Struct...

[Antti Kaihola (akaihola)]

The closed bug #219846 seems to me to describe this same issue.

[Antti Kaihola (akaihola)]

Sorry, I was mixing up things when switching the video *input* between v4l and v4l2. Changing the video *output* to "X Windows System (no Xv)" stops the system from crashing, but video playback performance is naturally poor.

[Antti Kaihola (akaihola)]

An interesting note: when using Xv in my multi-head setup, Totem works fine as long as it sits on the non-rotated external display. As soon as I drag or open it on the rotated laptop screen, the system crashes.

And to be exact, the whole system doesn't seem to crash – I'm still able to shut down cleanly by pressing the laptop's power button.

[Bryce Harrington (bryce)]

Forwarding this bug from a Ubuntu reporter:
https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/225400

Starting in around 2.3.1, the X server freezes when rotating 90 degrees on this hardware. In 2.2.x it would rotate but was not very stable and would freeze after certain actions. Currently it freezes immediately on rotation. (the screen is drawn and the cursor "works", and the cursor bitmap changes as it's moved around the screen, but beyond that, everything's frozen).

PCI ID:
(II) PCI: 00:02:0: chip 8086,2992 card 1028,01da rev 02 class 03,00,00 hdr 80

Old Xorg.0.log from 2.2.x with xserver 1:7.3+10ubuntu10:
http://launchpadlibrarian.net/14071327/Xorg.0.log.old

[Mike Coleman (tutufan)]

(I'm the original reporter--to Ubuntu--of the bug.)

[drieschner (drieschner)]

Rotation 90° does work alright if I change, before logging in (in GNOME), the session such that no start scripts are executed. It does not seem that anything is missing afterwards.
Michael

[Mike Coleman (tutufan)]

Michael, is this a suggested fix, or just an observation? What could be in the startup scripts that could hang the X server so badly, I wonder?

[drieschner (drieschner)]

It is an obvservation. I have no idea why this comes out -- I am a Newcomer to unbuntu!
Michael

[drieschner (drieschner)]

<email address hidden> schrieb:
> http://bugs.freedesktop.org/show_bug.cgi?id=17930
>
>
>
>
>
> --- Comment #3 from Mike Coleman <email address hidden> 2008-10-06 12:43:52 PST ---
> Michael, is this a suggested fix, or just an observation? What could be in the
> startup scripts that could hang the X server so badly, I wonder?
>
>
>
It is an obvservation. I have no idea why this comes out -- I am a
Newcomer to unbuntu!
Michael

[Bryce Harrington (bryce)]

Hi Michael,

I've forwarded your bug upstream to https://bugs.freedesktop.org/show_bug.cgi?id=17930. Could you please subscribe to that bug in case they need additional information or wish you to test something? Thanks ahead of time.

@AnttiKaihola - You have very different hardware from Michael (an i855), and your symptoms are quite different - your issue seems to get triggered by the visualizer, whereas Michael's issue seems insensitive to it. So please submit your issue as a separate bug.

[Mike Coleman (tutufan)]

Done. Thanks. --Mike

On Mon, Oct 6, 2008 at 12:59 PM, Bryce Harrington
<email address hidden> wrote:
> Hi Michael,
>
> I've forwarded your bug upstream to
> https://bugs.freedesktop.org/show_bug.cgi?id=17930. Could you please
> subscribe to that bug in case they need additional information or wish
> you to test something? Thanks ahead of time.
>
> @AnttiKaihola - You have very different hardware from Michael (an i855),
> and your symptoms are quite different - your issue seems to get
> triggered by the visualizer, whereas Michael's issue seems insensitive
> to it. So please submit your issue as a separate bug.
>
> --
> web page -> X server crash/blank, no display until reboot [Q965]
> https://bugs.launchpad.net/bugs/225400
> You received this bug notification because you are a direct subscriber
> of the bug.
>

[Carl Worth (cworth)]

Hi Bryce!

I knew we'd have to meet here sooner or later.

I tried to take a look at this bug and here's some of what I found:

* With fairly current, (master within the last couple of weeks at least), X server and xf86-video-intel, I tried various combination of rotation and applications. I rotated both before and after starting GNOME. I ran totem with visualization as well as mplayer rendering with xv, (since the original referenced bug seems to point to problems with xv). And everything worked fine for me---I couldn't find any problems.

* I talked with Keith Packard, and he says that all the rotation code takes place in generic parts of the X server, and the driver should be totally unaware of things. That makes the "starting in around 2.3.1" part of the bug report suspicious. Did the bug really change with xf86-video-intel 2.3.1 or did other things change at the same time?

Since I can't reproduce the bug yet, I'll need some more information from those who can. Specifically, if you can take a broken setup and make it work by switching only the Intel driver, (back to 2.2, say), then that would be very useful information.

Also, if you could update to 2.4 (or even a recent 2.4.9x release candidate for our imminent 2.5 release), and report if things start working for you, that would be great.

Thanks,

-Carl

[Bryce Harrington (bryce)]

Heya Carl,

@Mike and Michael, can both of you please try what Carl suggests and re-test for this bug against more recent code? You can get an ISO of intrepid-beta at http://cdimage.ubuntu.com/releases/8.10/beta/.

[Mike Coleman (tutufan)]

I just updated to the next Ubuntu beta, which is at

ii xserver-xorg-video-intel 2:2.4.1-1ubuntu6 X.Org X server -- Intel i8xx, i9xx display d

First off, basic rotation seems to work fine, and furthermore the Ubuntu "visual effects" also seem to work while rotated, which has not worked for me up to this point. So in this sense, this version is the best ever. (Hooray!)

3D acceleration (still) does not work correctly, though. If "visual effects" is not enabled, then the GL window is either black or frozen, except that if I grab the window and move it around, it seems to work perfectly until I stop. So I guess there's some sort of updating problem. (The behavior is slightly different between glxgears vs the Ubuntu screensaver preview window. The former starts in an almost working state, but with a black bar across the top 20% of the window; then, when I touch it, it's frozen while still and works while the window is moving. The latter starts black, and then likewise once touched is frozen unless the window is in motion (i.e., being moved with the mouse).

If "visual effects" is enabled, the effect looks more complex. Things only work while the window is being dragged, as above, but the rendering is not correct in this case, and leaves strips of rendered material outside the dragged window.

[Carl Worth (cworth)]

(In reply to comment #8)
> I just updated to the next Ubuntu beta, which is at
>
> ii xserver-xorg-video-intel 2:2.4.1-1ubuntu6
> X.Org X server -- Intel i8xx, i9xx display d

Thanks for trying with newer packages, Mike!

> First off, basic rotation seems to work fine, and furthermore the Ubuntu
> "visual effects" also seem to work while rotated, which has not worked for me
> up to this point. So in this sense, this version is the best ever. (Hooray!)

Excellent news.

This particular bug report is about the rotation. So since everything is working correctly, I'll go ahead and mark this bug as fixed. If you do see other issues with rotation, please feel free to reopen the bug.

> 3D acceleration (still) does not work correctly, though.

This is in general, right? Not only when rotated? If so, then it would be most helpful if you could open separate bug reports for each separate issue. (For example, I'm not the right person to help with any 3D issues at all.)

Thanks again for your testing and bug reporting. We really appreciate it!

-Carl

[Mike Coleman (tutufan)]

>> 3D acceleration (still) does not work correctly, though.

> This is in general, right? Not only when rotated? If so, then it would be most
> helpful if you could open separate bug reports for each separate issue. (For
> example, I'm not the right person to help with any 3D issues at all.)

I hadn't tried until you asked, but 3D seems to work fine when the display is not rotated, at least when "visual effects" are not enabled. (It'd be more difficult for me to check the "visual effects" enabled case, but I can if it actually matters here.)

So, yes, I think there is a still a rotation bug. Whether it's the same bug I will leave as a question to the philosophers. :-)

Thanks for your work on this--even in its current state this definitely made my day!

[Bryce Harrington (bryce)]

[Closing as per discussion on the upstream bug report that the crash/black screen issue while rotated is resolved with the 2.4.1 driver in Intrepid. Mike found some other issues but will be reporting those separately.]

[Gordon Jin (gordon-jin)]

(In reply to comment #10)
> So, yes, I think there is a still a rotation bug. Whether it's the same bug I
> will leave as a question to the philosophers. :-)

This should be a different bug, so please file a new bug.
You could also try http://bugs.freedesktop.org/show_bug.cgi?id=17967#c9 to see if vblank_mode impacts your 3d issue.