Created attachment 64564 [details]
Linux 3.4.5 Kernel .config
system hangs at boot with linux 3.4.0 to 3.4.5 (Gentoo kernel).
Last message shown at display is:
[drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
I am using the Nouveau graphics driver on an AMD64 system with NV43 [GeForce 6600]
# lspci|grep VGA
01:00.0 VGA compatible controller: NVIDIA Corporation NV43 [GeForce 6600] (rev a2)
*** Bug 51198 has been marked as a duplicate of this bug. ***
I've been seeing this also, with the latest kernel that I could run being 3.3.8. Today I did some more digging, and it appears that the last line is actually a red herring. For some reason not all printk's actually make it to the screen. Perhaps the video card is getting confused by something and stops printing lines at some point.
In any case I found out that my problem actually seems to be listed in this unsolved bug: https://bugzilla.redhat.com/show_bug.cgi?id=832548
I've "fixed" things on my end my by simply skipping all the nouveau perf entries in nouveau_perf_init in nouveau_perf.c. (I added continue right within the while loop). Obviously this doesn't get me any performance levels, but at least things don't crash, and other than that the card seems to work as expected.
I'd be happy to provide more information if needed, just let me know what you need.
I've got the exact same card as the OP:
03:00.0 VGA compatible controller: NVIDIA Corporation NV43 [GeForce 6600] (rev a2)
Does this still happen with linux v3.10-rc2 (latest at the time of writing)?
It would be nice to know what the first commit is that breaks things for you. Please do a git bisect between v3.3 and v3.4 on the drivers/gpu/drm/nouveau directory.
Alternatively (or in addition to), pick a recent kernel, get a crash dump (similar to what the referenced redhat bug has), and provide disassembly of the involved function (gdb nouveau.ko and then disassemble func_name). That way we can tell exactly where it's dying. You might need to use netconsole to get the crash dump, since it dies when initializing the video driver.
No information provided in the bug (e.g. dmesg), request for retest with recent kernel unanswered for months. Closing as invalid.
Unfortunately I don't have much time to play around with this machine and test things, but this bug is still present in linux 3.10.7. I realize that without additional information this is going to be hard to fix. Hopefully I'll be able to provide more information in the future if I find some time for that.
We can't do anything without your VBIOS (/sys/kernel/debug/dri/0/vbios.rom). It'd also be really nice to get a fresh stacktrace from that 3.10 (or newer) kernel, as well as the disassembly I asked for.
Created attachment 85900 [details]
vbios.rom - kernel 3.10.7