I have a Dell Inpsiron E1705 laptop running KDE on Gentoo Linux ~x86. I have been encountering an intermittent issue with compositing that occurs (when it occurs) immediately after a KDM logon. KDE basically freezes. I can see the desktop fully loaded, but if I go back and forth between the x server and a terminal via Ctrl+Alt+Fx key presses, the screen is black. Shortly after this occurs, KDE reports "Compositing was too slow and has been suspended". There is nothing obvious about this in the Xorg.0.log and this occurs with both the 195.24.15 drivers and the 256.29 drivers on both xorg server 1.7.7 and 1.8.1. Since I had been using xorg server 1.7.x with 190.53 before these issues started, I went back to it and my issues appear to have went away because I cannot get KDE to produce an "Intermittent Issue with Compositing with latest Nvidia Drivers", no matter how many times I reboot. This issue appears to be independent of kernel version, as I have used both kernel 126.96.36.199 and kernel 2.6.34 over the course of troubleshooting it. These are completely unpatched kernels from kernel.org.
When this issue does occur, I can restart the x server via a terminal, login to KDE from KDE and the issue will occur again. This occurs repeatedly every time I repeat that procedure until I reboot the system. This issue has also manifested itself in fast user switching.
I have been racking my brains out trying to nail down the cause and I have a thread chronicling my issues at the kde forums:
As far as I can tell, this is an nvidia driver issue. I have been doing google searches galore to try to find other people with the same issue and I have had difficulty finding anyone reporting anything identical to this.
I read the "If you have a stability problem, PLEASE read this first" thread and I have a few things to add to the above.
- My system's BIOS is up to date.
- Dell does not provide graphics card BIOS upgrades, but my graphics card recently died on me, so Dell replaced it, which could have resulted in a BIOS update anyway.
- My kernel is compiled without HPET.
- The only message that appears when doing "dmesg | grep NVRM" is the "loading NVIDIA UNIX x86 Kernel Module 190.53" message.
- I do have a Linux SMP kernel and I was able to observe the following:
# dmesg | grep MM
[ 0.103116] PCI: MMCONFIG for domain 0000 [bus 00-3f] at [mem 0xf0000000-0xf3ffffff] (base 0xf0000000)
[ 0.103227] PCI: MMCONFIG at [mem 0xf0000000-0xf3ffffff] reserved in E820
[ 0.103324] PCI: Using MMCONFIG for extended config space
[ 0.216135] hpet0: at MMIO 0xfed00000, IRQs 2, 8, 0
I reverted my system back to xorg-server 1.8.1 and nvidia-drivers 256.29 to try the recommendation of booting the kernel with pci=nommconf, but I encountered the issue immediately afterward on first boot. This time KDE did not seem to recover gracefully (which also happens some times, but is rare), and after becoming impatient, I forced a reboot, which did not encounter the issue.
I have run nvidia-bug-report.sh and attached the output file. When I encounter the issue again, I will run nvidia-bug-report.sh again and attach another output file. Unfortunately, I do not know how to pass "-- -logverbose 6" to the x server with KDM in use. Also, I am removing pci-nommconf from my grub.conf file.