Bug 8722

Summary:	do_wait: drmWaitVBlank returned -1, IRQs don't seem to be working correctly.
Product:	Mesa	Reporter:	Shuang He <shuang.he>
Component:	Drivers/DRI/i810	Assignee:	Default DRI bug account <dri-devel>
Status:	CLOSED NOTABUG	QA Contact:
Severity:	normal
Priority:	high	CC:	gordon.jin, thomas
Version:	git
Hardware:	x86 (IA32)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:
Attachments:	xorg conf file X server log

Description Shuang He 2006-10-22 20:09:50 UTC

--Architecture(ia32,ia32e,compatiblity): ia32
--2D driver: git(5ac927d6f05ba6f9608d7e758cbac66e8b3a9db1)
--Xorg: 7.1
--Mesa: CVS tip
--OS: FC5
--Kernel: 2.6.17

We got the following report when running glxinfo, glxgears or glean:
do_wait: drmWaitVBlank returned -1, IRQs don't seem to be working correctly.
Try running with LIBGL_THROTTLE_REFRESH and LIBL_SYNC_REFRESH unset.


this issue also happens on 855gm 865g 915gm 945gm platforms

Comment 1 Shuang He 2006-10-22 20:10:29 UTC

Created attachment 7485 [details]
xorg conf file

Comment 2 Shuang He 2006-10-22 20:11:06 UTC

Created attachment 7486 [details]
X server log

Comment 3 Michel Dänzer 2006-10-23 00:56:16 UTC

Which version of the i915 DRM is that? Can you also attach the DRM related
kernel output?

Note that this should only be an issue if the 3D driver is configured for
sync-to-vblank, either via the mentioned environment variables or via the
driconf setting vblank_mode.

Comment 4 Shuang He 2006-10-23 01:12:06 UTC

We didn't update the drm and i915 module for a long time, the related kernel output
[drm] Initialized drm 1.1.0 20060810
ACPI: PCI Interrupt 0000:00:02.0[A] -> GSI 16 (level, low) -> IRQ 16
PCI: Setting latency timer of device 0000:00:02.0 to 64
[drm] Initialized i915-mm 1.7.0 20060929 on minor 0:


And, we didn't got those environment variables set when we run the glxinfo. But
still get those ouput

Comment 5 Michel Dänzer 2006-10-23 08:39:41 UTC

(In reply to comment #4)
> We didn't update the drm and i915 module for a long time, the related kernel
output
> [drm] Initialized drm 1.1.0 20060810
> ACPI: PCI Interrupt 0000:00:02.0[A] -> GSI 16 (level, low) -> IRQ 16
> PCI: Setting latency timer of device 0000:00:02.0 to 64
> [drm] Initialized i915-mm 1.7.0 20060929 on minor 0:

Thanks. Could you try current git master? I don't really expect it to make a
difference though.

Also, what does

grep i915 /proc/interrupts

say?

> And, we didn't got those environment variables set when we run the glxinfo. 

No vblank_mode entry in /etc/drirc or ~/.drirc either?

Comment 6 Thomas Hellström 2006-10-23 11:35:59 UTC

Michel, I get this error when doing a VT switch with the mesa texmem-0-3 branch
as well.

Comment 7 Shuang He 2006-10-23 22:25:10 UTC

>> We didn't update the drm and i915 module for a long time, the related kernel
output
>> [drm] Initialized drm 1.1.0 20060810
>> ACPI: PCI Interrupt 0000:00:02.0[A] -> GSI 16 (level, low) -> IRQ 16
>> PCI: Setting latency timer of device 0000:00:02.0 to 64
>> [drm] Initialized i915-mm 1.7.0 20060929 on minor 0:

>Thanks. Could you try current git master? I don't really expect it to make a
>difference though.
it works! with the current git master, it won't report these information
Thanks for the help

>Also, what does

>grep i915 /proc/interrupts

>say?
This got nothing

>> And, we didn't got those environment variables set when we run the glxinfo. 

>No vblank_mode entry in /etc/drirc or ~/.drirc either?

Yeah, we don't use those files at all.

Comment 8 Shuang He 2006-10-23 22:32:09 UTC

So this issue is caused by the kernel module out of date.

Comment 9 Michel Dänzer 2006-10-24 05:49:45 UTC

(In reply to comment #6)
> Michel, I get this error when doing a VT switch with the mesa texmem-0-3 branch
> as well.

I can't reproduce it. Are you using current drm git master? Also, do you mean
you get this every time after the VT switch, or only if you switch while the
client is running?

Comment 10 Thomas Hellström 2006-10-24 07:03:41 UTC

Yes, with a client running:

Wit Alan's xorg repo, latest git drm and texmem-0-3-branch:

Launch glxgears,
Switch to a text terminal
Switch back. Glxgears should've died because of the vt drilock problem, but just
before that the vblank message is printed out.

Comment 11 Adam Jackson 2009-08-24 12:24:44 UTC

Mass version move, cvs -> git

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.