Also sprach Kurt Wall on Sunday 14 April 2002 02:58: > > mtrr: no MTRR for dc000000,2000000 found Thanks Kurt, I've browsed the net a bit and found this one to be a bug which has been often reported on various list, including those where the kernel celebrities talk to each other. But I haven't seen a fix far and wide ... > This is the Memory Type Range Register (hence, MTRR). In your kernel > config file, set > > CONFIG_MTRR=y > > and recompile. I had MTRR enabled in my config. Now I just did the opposite, disabled it and recompiled. I found no substantial difference in performance and behaviour (of course the error message in the kernel logs disappeared, but there is a warning in /var/log/XFree86.0.log saying: 'System lacks support for changing MTRRs'). Direct rendering is still enabled. > > [drm:r128_cce_indirect] *ERROR* process 687 using buffer owned by 0 > > This means that the process numbered 687 is trying to access a buffer > owned by the UID 0, which is root. The error is coming out of the DRM > code ($KERNEL/drivers/char/drm), specifically r128_state.c. To the > degree I understand the code, and I confess my understanding of it > is limited, this error occurs in three places. In general, > the process is trying to flush some data to keep texture memory > intact, but doesn't have root privileges and is causing an error. > > I'd definitely browse the archives of the Xpert mailing list and see > if this has come up before (I'm sure it has) and how to fix it. This has come up on a lot of lists as well (Xpert, dri-users, dri-devel ...), but I found (or recognized, you know, I'm not a coder) no real fix. I downloaded and installed the latest r128 driver from dri CVS, but the messages remain, and the somewhat distorted kdm screen (often reported) as well. I think I'll wait for better days to come. Klaus
