[i915] Crash when mouse at top of screen on 915GM

[peridot]

Binary package hint: xserver-xorg-video-intel

I am using X normally for a variable period of time - sometimes a day or more, sometimes minutes - when suddenly the screen goes blank, the backlight goes out and then comes back on (as if the X server were restarting) but then nothing appears on the screen. I cannot switch to a text terminal with C-A-F1, and C-A-Backspace does nothing. C-A-Del makes the system reboot, and very late in the process I see the splash screen signaling a reboot.

The bug occurs randomly, but I think that every time it has occurred, my mouse cursor has been at the top of the screen, just at the top of a firefox titlebar. (I have no panel across the top.) I normally have compiz turned off, but the bug also occurs when I have turned compiz on. (The attached log is from a crash with compiz active.)

Using xserver-xorg-intel 2.2.1-1ubuntu12 on 8.04; my xorg.conf contains nothing machine-specific. This did not occur with Gutsy's standard X server.

This bug resembles bug #204035 and #176377, but they are marked as closed and people encountering the same problem are urged to submit new bugs.

[peridot (peridot-faceted)]

I have now seen the bug occur when the mouse is at the top of the screen in a non-firefox window. A cursor change occurs there, since I have no panel at the top of my screen. Could this bug be related to bug #153466?

[Bryce Harrington (bryce)]

Please collect a full backtrace when the crash occurs - see http://wiki.ubuntu.com/X/Backtracing for directions.

This looks interesting:

(WW) intel(0): ESR is 0x00000010, page table error
(WW) intel(0): PGTBL_ER is 0x00000010, display A pte
(WW) intel(0): PRB0_CTL (0x0001f001) indicates ring buffer enabled
(WW) intel(0): PRB0_HEAD (0xaba111e4) and PRB0_TAIL (0x000111f0) indicate ring buffer not flushed
(WW) intel(0): Existing errors found in hardware state.

[peridot (peridot-faceted)]

Regrettably I don't have another machine to log into this laptop with, so I can't produce a backtrace the way described in that page. Can I persuade the X server to dump core somehow?

I am trying a git bisection, but since the bug is intermittent, it's a slow process. So far no crashes after about a week's use of 2:2.1.99+git20071114.58150932-1.

[peridot (peridot-faceted)]

Aha. It crashed and I got a core file. I'm afraid it's with version 2.3.1-bwh3, since I needed panel fitting. I'll try to get a core from hardy's version, but this crash appeared to be the same bug: X server dies with the mouse at the top of the screen, machine appears still to be working but console is unresponsive to everything but Control-Alt-Delete. I've also (in the past) seen it happen at the *bottom* of the screen. I'll point out that since I use vertical panels on the sides of my screen, the mouse pointer normally changes at the top and bottom of the screen; in these cases the X server dies before I see the cursor change.

I'm keeping the core file in case it's of any use.

[Antony Jones (wrh)]

This sounds like my problem too. My panel is at the top of the screen, and about twice a day, when I try to restore firefox from a minimized state by clicking it, X crashes to the text screen, but with the cursor still painted on there?

This is from Xorg.0.log.old after a reboot:

Warning: LookupDrawable()/SecurityLookupDrawable() are deprecated. Please convert your driver/module to use dixLookupDrawable().
(II) XAA: Evicting pixmaps

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x7e) [0x80c780e]
1: [0xb7eeb420]
2: /usr/lib/xorg/modules//libxaa.so [0xb759f5b9]
3: /usr/bin/X [0x8173013]
4: /usr/bin/X(CompositePicture+0x150) [0x815a1f0]
5: /usr/bin/X [0x81601df]
6: /usr/bin/X [0x815d085]
7: /usr/bin/X [0x81506ee]
8: /usr/bin/X(Dispatch+0x2cf) [0x808d8df]
9: /usr/bin/X(main+0x48b) [0x807471b]
10: /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe0) [0xb7c80450]
11: /usr/bin/X(FontFileCompleteXLFD+0x201) [0x8073a91]

Fatal server error:
Caught signal 11. Server aborting

(II) AIGLX: Suspending AIGLX clients for VT switch

[Bryce Harrington (bryce)]

Created an attachment (id=19461)
backtrace

I'm forwarding this bug from a Ubuntu reporter:
https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/228431

Problem:
Intermittent lockup when mouse cursor is at top (sometimes maybe bottom?) of screen. Issue appears to have started somewhere between -intel 2.1.99+git20071114.58150932-1 and the 2.2.1 release, and still exists as of at least 2.3.1.

From the backtrace (attached), perhaps it's a VT switching issue?

Original Report:
"I am using X normally for a variable period of time - sometimes a day or more, sometimes minutes - when suddenly the screen goes blank, the backlight goes out and then comes back on (as if the X server were restarting) but then nothing appears on the screen. I cannot switch to a text terminal with C-A-F1, and C-A-Backspace does nothing. C-A-Del makes the system reboot, and very late in the process I see the splash screen signaling a reboot.

The bug occurs randomly, but I think that every time it has occurred, my mouse cursor has been at the top of the screen, just at the top of a firefox titlebar. (I have no panel across the top.) I normally have compiz turned off, but the bug also occurs when I have turned compiz on. (The attached log is from a crash with compiz active.)

Using xserver-xorg-intel 2.2.1-1ubuntu12 on 8.04; my xorg.conf contains nothing machine-specific. This did not occur with Gutsy's standard X server.

[In another example, using 2.3.1-xx] X server dies with the mouse at the top of the screen, machine appears still to be working but console is unresponsive to everything but Control-Alt-Delete. I've also (in the past) seen it happen at the *bottom* of the screen. I'll point out that since I use vertical panels on the sides of my screen, the mouse pointer normally changes at the top and bottom of the screen; in these cases the X server dies before I see the cursor change."

Tested:
Gutsy (xserver 7.2-xx; -intel 2.1.1-xx) - Works
Hardy (xserver 7.3-xx; -intel 2.2.1-xx) - Fails
      (xserver 7.3-xx; -intel 2.1.99+git20071114.58150932-1) - Works
      (xserver 7.3-xx; -intel 2.3.1-xx) - Fails

lspci:
(II) PCI: 00:02:0: chip 8086,2592 card 1043,1882 rev 03 class 03,00,00 hdr 80

[Bryce Harrington (bryce)]

Created an attachment (id=19462)
Xorg.0.log

[Gordon Jin (gordon-jin)]

Is it possible for the users to try a newer driver (2.4.* or better 2.5.*)?

[peridot (peridot-faceted)]

Version 2.4.1 is part of the (currently beta) next version of Ubuntu; I plan to upgrade sometime after it is released (supposedly in October). If there are stable backports I could try those (I'd be very pleased to see this bug go away). I'm afraid it occurs rarely enough that it would take several weeks of use to gain some confidence that the bug had disappeared in any case. (Confirmation that the bug is present would of course probably be faster.)

[Bryce Harrington (bryce)]

Hi peridot,

I've forwarded this bug upstream to https://bugs.freedesktop.org/show_bug.cgi?id=17959. Could you please subscribe to that bug, in case they need additional info or wish you to test something? Thanks ahead of time.

Thanks for the backtrace btw. I don't understand exactly what's going on, but it sort of looks like there's a VT switch trying to occur, that fails. Do you have your system configured to do something (like switch desktops) when your mouse moves to the screen edge?

Also, please could you test this against the Intrepid-Beta (or a daily) livecd, to see if it's still occurring with the latest version of the driver and xserver? You can obtain an ISO of it from http://cdimage.ubuntu.com/. Unless they can reproduce the problem, upstream will most likely wish for you to do this (or may request that you test against current git head).

[peridot (peridot-faceted)]

The VT switch may be a red herring - when the screen goes black I have a habit of hitting Ctrl-Alt-Fn to try to switch to a text terminal. It doesn't work when I've triggered this bug, and X is still crashed if I don't do it. My system is not configured to do anything when I hit a screen edge (apart from change the mouse cursor shape, since the window titlebar is right at the top of the screen).

I'll see what I can do with the beta, but it's a bit awkward, as I have my fingers trained to avoid bumping the top or bottom of the screen and in any case the bug is intermittent, so I go weeks without triggering it. Once 8.10 is released I'll probably upgrade to it (hopefully it won't be as big a disaster as upgrading to 8.04).

[Antony Jones (wrh)]

This bug is no longer occuring for me - no crashes for a long time on 8.04 and not had any yet on 8.10 even with dual monitors enabled which used to cause the crashes to become much more frequent.

[Jesse Barnes (jbarnes-virtuousgeek)]

This sounds a lot like the DSPARB related problem we fixed awhile back. I'll mark it fixed, but please re-open if the reporter can confirm with 2.5.x or 2.6.x.

[Bryce Harrington (bryce)]

peridot, upstream thinks this bug is fixed now and closed it. Antony's last comment seems to confirm this, so I'm closing this bug as well. However, if you can reproduce this problem with the 2.5.x driver (available now in Jaunty, or from the XorgEdger's page), please feel free to reopen this bug and the upstream bug.