It seemed I'm not the only person having this problem, so I decided to file a bug... Basically the nvidia 6106 drivers run into an issue with USB 2.0 and kernel 2.6.7... Nvidia calls it a usb/kernel bug, but USB 2.0 in kernel 2.6.7 (not 2.6.6 & earlier) will disable IRQ 10 (in my case), which sends my nvidia card into a lockup frenzy when starting X with the 6106 drivers.... Leaves my system at a black screen when starting X; the system doesn't *seem* to accept keyboard input other than ctrl-alt-del which does a successful shutdown/reboot. I've already mentioned this to cyfred, I'm not sure if it affects all systems, or certain hardware. I'm on an nforce2 (DFI NF2 Ultra Infinity) more info on post #43 here: http://www.nvnews.net/vbulletin/showthread.php?t=30970&page=3&pp=15 and it seems another user may have hit this same problem here: http://forums.gentoo.org/viewtopic.php?p=1353336 I haven't managed to track down any proper fixes yet, problem disappears with kernels < 2.6.7, or disabling USB 2.0 in kernel. I also haven't tried kernel 2.6.8, since as of this post its still in 'rc' status. Reproducible: Always Steps to Reproduce: 1. kernel 2.6.7, USB 2.0 enabled 2. emerge & set up X with nvidia 6106 3. start X Portage 2.0.51_pre13 (default-x86-2004.0, gcc-3.3.3, glibc-2.3.4.20040619-r0, 2.6.7-hb3 i686 AMD Athlon(tm)) ACCEPT_KEYWORDS="x86 ~x86" CFLAGS="-march=athlon-xp -O2 -pipe -mcpu=athlon-xp -fomit-frame-pointer"
I'll try and reproduce this in 2.6.8 in a little bit, however there aint much we can do to nvidias stuff to fix it if there really is a problem in there drivers.
If you can show that it's a problem in their drivers and not something else, there's an nvidia guy we can CC.
Well I cant reproduce it on 2.6.8-rc2 (or rc1 I think but didnt test it too much) I guess try one of them out and report back
same results with: nvidia 6106 + 1) development-sources-2.6.6 (modules or inkernel USB) 2) development-sources-2.6.7 (modules or inkernel USB) 3) development-sources 2.6.8-rc2 (modules or inkernel USB) Also, it doesn't matter (in my case) whether I enable EHCI or OHCI or both, I can only start X with nvidia 6106 with ZERO USB support as far as I can tell so far. I also played around with various USB options for several hours, no luck.
UPDATE: I managed to get this working with OHCI (USB 1.1) with kernel 2.6.8-rc2 some things that may have affected it: (these are working settings) CONFIG_SYSVIPC=y CONFIG_X86_UP_APIC is not set CONFIG_X86_MCE is not set CONFIG_SCSI=y (was =m) CONFIG_IEE1394 is not set CONFIG_AGP=y CONFIG_AGP_NVIDIA=y (was previously attempting to use NvAGP) CONFIG_USB_DEVICEFS is not set CONFIG_USB_DYNAMIC_MINORS is not set CONFIG_USB_EHCI_HCD is not set (tried this both ways, =y/m doesn't work) I think it may have had to do with NvAGP vs agpgart, or APIC. (I've had previous apic issues on this nforce2) I may try to track down a specific working set of options, and try to track down what may be breaking it. It think I should be able to enable ieee1394, I disabled it 'just in case'
If it wasnt working with OHCI thats more likely a kernel problem with how you set it up... (likely not definately); However nvidia have said its a USB2.0 problem and basically after looking at the ehci (USB2.0 driver) code diffs b/n 2.6.6 and 2.6.7 id say it has something to do with the ability to put the usb port / device to sleep. Im not exactly sure how the interupts are disabled but it looks to me that what needs to happen is a interupt check done in the nvidia kernel module, and if the interupt is disabled it needs to explicitly re-enable it and lock it enabled... or perhaps use software interupts instead of hardware ones in that case, this might actually be better cause the previous would probably have an (bad?) effect on USB2.0 data.
*** Bug 58551 has been marked as a duplicate of this bug. ***
Please test this with 6111 which just went into cvs. Report back on success or failures. Thanks.
Using gentoo-dev-sources-r13 and the 1.0.6111 nvidia-kernel driver produces the same issue. startx will cause a black screen everytime if ehci is compiled into the kernel either as a module or directly into the kernel.
Check X isnt eating up your CPU like 99% usage or something (best way to ssh in to your machine when the lock up occurs)... If X is eating up CPU usage try disabling APIC (either by recomipiling your kernel, or specifying "noapic" on the kernel boot line)... If you aren't using APIC let me know that aswell...
X is eating up 99.9% of the CPU when I type 'startx'. Disabling apic using the kernel parameter 'noapic' does not solve this issue. Still get a black screen and X eating up all the cpu.
OK well there isnt much we can do with this atm.. marking upstream Summarily: Dont use echi if you have this problem, or drop back to an older driver.
This issue seems to be resolved in gentoo-dev-sources-2.6.8-r2.
hollywoodb: could you confirm that this is fixed?
Moving to x11-drivers.
Marking this as FIXED since someone reported it as such and no one else is still complaining.