I have a 780G chipset (HD3200), using xf86-video-ati 6.14.0 with Xserver 1.9, kernel 2.6.37 and 2.6.38-rc4. Looking in my kern.log, I see regularly this error message:
[drm:radeon_crtc_page_flip] *ERROR* failed to reserve new rbo buffer before flip
Just a me too on this, also seeing the same on my ArchLinux 64bit installation with an Mobility Radeon X1400 PCIE and DRM Testing with KMS PageFlip enabled.
ATI Dri version: 7.10.0.git20110206
Kernel version: 2.6.37-rc8-34105-g2b2fd60-dirty
i see those errors appear in my dmesg occasionally, and from what it appears the 3D performance goes down a bit sometimes after they have appeared. noticed by wobbly windows in compiz not being silk smooth after a couple of occurances.
When testing with latest drm and kernel, I can't find the error anymore.
I'm experiencing this error on my updated Debian Testing with an ATI R100 (RV200 4C57) and:
I'm using no manual xorg.conf.
(In reply to comment #3)
> I'm experiencing this error on my updated Debian Testing with an ATI R100
> (RV200 4C57) [...]
Are there also 'reserve failed' messages before the messages referenced in the original report?
Do you notice any problem other than the messages?
Created attachment 56520 [details]
Dmesg with "failed to reserve"
(In reply to comment #4)
> Are there also 'reserve failed' messages before the messages referenced in the
> original report?
No, there aren't. Is the only type of "reserve failed" I got.
> Do you notice any problem other than the messages?
Mhm.. seems not. I've noticed anything related with radeon but, you can check in the attachment of my dmesg.
Created attachment 56529 [details] [review]
Add return value to error message
Please apply this debugging patch and post the value(s) of r when the message occurs again.
After last kernel upgrade ( 3.2.0-1-486 with Debian patches ) I don't get the error messages anymore.
Would you like me to check the value of r with your patch using the previous kernel?
(In reply to comment #8)
> After last kernel upgrade ( 3.2.0-1-486 with Debian patches ) I don't get the error messages anymore.
Great, let's assume for now it's been fixed.
> Would you like me to check the value of r with your patch using the previous kernel?
I don't think that's necessary. If you ever see the problem again with the newer kernel, it would be interesting to know the value of r though.