Weird problems since installing beta driver

[nageroc]

I have a laptop (Gateway MX3410) with a GeForce Go 6100 card and I have been experiencing some weird problem since installing the nVidia beta driver (1.0-9626). I am using the beta driver because the earlier drivers don't work with my chipset. Prior to this, I was using the 'nv' driver and not experiencing these problems.

The worst problem is the key repeat. I will be typing away when all of a sudden every key press results in a handful to dozens of characters being entered. This doesn't happen all the time but when it does, it is impossible to work. The problem comes and goes.

TThis is a samplee of wwhhaaaattt happppens when thiss prooblemm occcuuurs....

Sometimes it's worse. Typing a long message such as this one takes a long time, and I am a fast typer. Argh...I just lost a couple of paragraphs because the backspace key went nuts and gobbled them up. :-(

Occasionally the opposite occurs. For example, in a konsole window I press and hold the enter key and linefeeds are printed every few seconds as opposed to two or three per second.

Next is the mouse scroll wheel. Usually when I scroll a firefox window, it's smooth and follows my motions. Other times I scroll, wait a few seconds and then the window follows. Speaking of firefox, I notice as I type this that the animated smilies next to the text box are very slow and often come to a stop. Other times the animation is normal.

Another weird thing...I have a clock docked in my KDE panel. I've set it to display the seconds and flash the colons. At times, the colons will flash very quickly and the clock will skip ahead or back a few seconds. Once while this was happening, I ran date in a while loop once per second. There was no indication that my clock was skewed although occasionally the loop would sleep for longer than the one second specified.

One last weirdness before I go nuts trying to type this. My OpenGL screen saver (Euphoria) is at times very choppy. Other times it is as smooth as silk.

Not all these problems happen at the same time. They come and go. But it really makes this notebook unusable. Help!

I have attached an nVidia bug report.


Chris

[netllama]

All of this behavior sounds like a CPU frequency scaling bug. If you disable CPU frequency scaling and/or boot with the irq=poll kernel parameter, does that have any impact?

Thanks,
Lonni

[nageroc]

Hi Lonni,

Originally Posted by netllama

All of this behavior sounds like a CPU frequency scaling bug. If you disable CPU frequency scaling and/or boot with the irq=poll kernel parameter, does that have any impact?

Thanks for the quick response. I have set "irq=poll" in grub.conf and so far so good, but I need to let it run for a while before I can say with certainty that this has cured the problem.

CPU frequency scaling is compiled into my kernel. Googling the subject, I came across the following:

Some of the amd64 laptops require that you use ACPI to detect the available frequency states. Does your kernel have the options

CONFIG_X86_POWERNOW_K8
CONFIG_X86_POWERNOW_K8_ACPI

These are compiled as modules but not loaded. I think I will experiment with this seeing as how your assessment is looking more and more likely. Up 30 minutes now and still stable! That annoying little banana never looked so good!

Any idea why this would exhibit itself when running the nVidia driver? I did not notice this problem when running with the 'nv' driver, and I haven't experienced the annoying key repeating problem when working from a vty.

Now if only someone can help me solve my ALSA problems...


Thanks again,

Chris

[nageroc]

Spoke too ssooonn...thheeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee eeeeeeeeeppprrooobbbblllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllleemmm ppeerrrsiistsss.


:-(

[nageroc]

I have rebuilt my kernel, turning off CPU frequency scaling. Let's see if that helps.

UPDATE: Nope. :-(


Chris

[nageroc]

Lonni,

Well, I have switched back to the 'nv' driver, but if you have any other ideas, I'm willing to try them.

UPDATE: Okay, I have ruled out the nvidia driver. I am seeing the same problem with 'nv'. :-( Dang...this really wasn't a problem before installing the beta drivers. I guess I'll look through my emerge logs to see what may have been introduced since then.


Chris

[nageroc]

I found this article on Gentoo Forums:

http://forums.gentoo.org/viewtopic-t...epetition.html

The last note is interesting; problems similar to mine seem to be fixed after changing the time source to tsc. Looking through my logs, I found that my time source changed from pmtmr to tsc on Sept. 20. This corresponds to a kernel update.

I tried changing my time source back by passing the "clock=pmtmr" parameter to the kernel, but that time source is no longer compiled it. It did, however, fall back to pit.

This somewhat agrees with my memory of the problem starting around the time I installed the beta nVidia driver, about a month earlier (Aug. 25).

I'm not going to jinx myself and say that the problem has been resolved, but it's looking good so far.


Chris

[nageroc]

Originally Posted by nageroc

I'm not going to jinx myself and say that the problem has been resolved, but it's looking good so far.

The problem has not exhibited itself since changing the time source, so I think it is safe to say without jinxing myself that it has been resolved. Some research has indicated that the tsc clock source is not reliable when run on a system with multiple cores or where the CPU can change frequency.

Lonni, your suggestion may have worked if CPU Frequency Scaling was the only factor, but my notebook has a dual-core Turion.


Thanks again for the help,

Chris