Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

Reply
 
Thread Tools
Old 12-23-03, 09:01 AM   #13
ringerc
Registered User
 
Join Date: Dec 2003
Location: Perth, Western Australia
Posts: 5
Default

Arrggh, only single attachments allowed. Sorry, I'm used to email, or the "attach after post" of bugzilla.

Here's the dmesg.
Attached Files
File Type: txt dmesg.txt (14.9 KB, 265 views)
ringerc is offline   Reply With Quote
Old 12-23-03, 09:10 AM   #14
ringerc
Registered User
 
Join Date: Dec 2003
Location: Perth, Western Australia
Posts: 5
Angry

.... and XFree86.0.log. I'm glad to see "log" in the list of accepted filenames ;-)

*sigh*

Sorry for the spam, folks. To save on it, I've appended my XF86Config to the attached file. I should've probably appended the files or zipped them, but didn't think of it. Hopefully this'll help get the problem _fixed_ though.

Note the
(WW) NVIDIA(0): Failed to verify AGP usage
reported by many.

Also, I can't remember, but ... have all the users having problems been using VIA chipsets? Anybody been using a custom kernel with things like CONFIG_HIMEM disabled (or CONFIG_X86_UP_APIC and maybe CONFIG_X86_UP_IOAPIC enabled) ?
Attached Files
File Type: txt xfree86_log_and_config.txt (41.6 KB, 204 views)
ringerc is offline   Reply With Quote
Old 12-23-03, 09:20 AM   #15
zander
NVIDIA Corporation
 
zander's Avatar
 
Join Date: Aug 2002
Posts: 3,740
Default

@bwkaz: Yes, the patches were updated yesterday to address this problem (and 2.4.23 incompatibility).

@ringerc: Judging from the Oops, you are seeing the same problem.
zander is offline   Reply With Quote
Old 12-23-03, 10:16 AM   #16
ringerc
Registered User
 
Join Date: Dec 2003
Location: Perth, Western Australia
Posts: 5
Thumbs up FC1 - works4me with patch, tweaks

@zander: Indeed - that seems to have done the trick, with a few tweaks. I'd have tried it earlier if I'd properly clicked that the patch wasn't 2.6 only (blame it on it being post-midnight here). Thanks for the quick reply.

For anybody else needing to use Fedora Core 1 with this driver and the default FC kernel, I needed to perform a few tweaks.

First: a Red Hat patch to the kernel has added a typedef for irqreturn_t to the default kernel includes; as such, you'll need to comment out

typedef void irqreturn_t;

from nv-linux.h (after applying the 2.6 patch) to get the drivers to compile. For those totally unfamiliar with C (though I'm pretty close to that state...), replacing the line with:

/* typedef void irqreturn_t; */

will do the trick nicely.

As the Nvidia driver build process needs a valid .config you'll need to:

cp /boot/config-`uname -r` /usr/src/linux-`uname -r`/.config

if you're using a FC packaged kernel. Naturally this won't do you much good unless you have the kernel-source RPM for your kernel image installed.

You'll also want to

export CC=gcc32

before running

make
make install


to make sure you build the interface with the compiler RH used to build the fedora kernel. Once I got all that sorted, it worked a treat (at least with NvAGP=1). RenderAccel is still happy. Performance seems back to normal:

12949 frames in 5.0 seconds = 2589.800 FPS

BTW, my previously posted info contained some stupid errors. Most notably:
Load "dri"
wasn't commented out in my XF86Config, render acceleration was enabled, the logo was supressed, and the latter two options were in the Monitor section not the Device section. Fixing this made _absolutely_ no difference to the problems reported, but it's worth noting.

Thanks to all the NVidia driver developers, by the way, for the RENDER acceleration. I will buy nothing but NVidia for quite some time, simply because the drivers are vastly superior in quality to any other vendor or OSS driver. RENDER accel is just the most notable example - it at least doubles the performance of many of the common desktop apps I use.

(lack of message preview is driving me nuts)

Last edited by ringerc; 12-23-03 at 10:50 AM.
ringerc is offline   Reply With Quote
Old 12-23-03, 10:43 AM   #17
zander
NVIDIA Corporation
 
zander's Avatar
 
Join Date: Aug 2002
Posts: 3,740
Default

Yes, the Linux 2.6 patches test for Linux 2.4.23 (or higher) and typedef irqreturn_t for older releases; it is unfortunate that conflicting changes such as this one frequently aren't easily identifyable.
zander is offline   Reply With Quote
Old 12-23-03, 11:21 AM   #18
Papol
Registered User
 
Join Date: Dec 2003
Posts: 31
Default

By the way: It appears that copying nvida-settings from :

~./NVIDIA-Linux-x86-1.0-4620-pkg0/usr/bin

of the 4620 drivers will work after some fashion with the 5328 drivers as well.

I realize it doesn't sound like a *good* idea and it probably isn't It just was something I tried.
Papol is offline   Reply With Quote
Old 12-23-03, 11:41 AM   #19
ringerc
Registered User
 
Join Date: Dec 2003
Location: Perth, Western Australia
Posts: 5
Default

@prismo: works4me , and there are some cool things that can be done with it, too. Here's a pointless and impractical example:

brainf__ck.py
Code:
#!/usr/bin/env python

import os
import sys

ascend = range(50,200)
descend = range(50,200)
descend.reverse()

for redgamma in ascend:
        curgamma = str(redgamma / 50.0)
        result = os.system("/usr/bin/nvidia-settings -a localhost:0.0/RedGamma=%s" % curgamma)
for bluegamma in ascend:
        curgamma = str(bluegamma / 50.0)
        result = os.system("/usr/bin/nvidia-settings -a localhost:0.0/BlueGamma=%s" % curgamma)
for redgamma in descend:
        curgamma = str(redgamma / 50.0)
        result = os.system("/usr/bin/nvidia-settings -a localhost:0.0/RedGamma=%s" % curgamma)
for bluegamma in descend:
        curgamma = str(bluegamma / 50.0)
        result = os.system("/usr/bin/nvidia-settings -a localhost:0.0/BlueGamma=%s" % curgamma)

result = os.system("/usr/bin/nvidia-settings -a localhost:0.0/RedGamma=1")
result = os.system("/usr/bin/nvidia-settings -a localhost:0.0/BlueGamma=1")
More useful things, like a toolbar button to toggle vsync, are trivial. I _love_ this tool.
ringerc is offline   Reply With Quote
Old 12-23-03, 12:13 PM   #20
Azureflare
Registered User
 
Join Date: Dec 2003
Posts: 2
Default Weird Nvidia stuff

OK I am running Mandrake 9.2 on an Athlon 2200+ with a shuttle mobo (not sure of version ATM). The stock NVIDIA 5328 .run file doesn't work for me (It complained about files existing like /usr/X11R6/lib/libGL.so and /usr/X11R6/lib/libGL.so.1 which never existed and do not exist now; it installed, but it ran glxgears at 5 fps which is unacceptable; also getting similar errors in syslog that other people were getting), so I Tried the patches on minion.de. At first it didn't work; I did the thing with commenting out the irq line in nv-linux.h, and the kernel module was built, but it said that the kernel module wasn't created! So I took a peek in the .manifest, and also in the usr/src/nv, and also at the README. I realized that the minion.de patched .manifest expects nvidia.ko, which was NOT being created by the make process! So, I changed references to nvidia.ko to nvidia.o, and it was able to install fine. I'm not sure if this will cause any problems, what is this mysterious nvidia.ko file? I have never had an nvidia.ko file on any Mandrake system that I have used. Anyone know more about this?
Azureflare is offline   Reply With Quote

Old 12-23-03, 12:28 PM   #21
Papol
Registered User
 
Join Date: Dec 2003
Posts: 31
Default Re: Weird Nvidia stuff

Quote:
Originally posted by Azureflare
OK I am running Mandrake 9.2 on an Athlon 2200+ with a shuttle mobo (not sure of version ATM). The stock NVIDIA 5328 .run file doesn't work for me (It complained about files existing like /usr/X11R6/lib/libGL.so and /usr/X11R6/lib/libGL.so.1 which never existed and do not exist now; it installed, but it ran glxgears at 5 fps which is unacceptable; also getting similar errors in syslog that other people were getting), so I Tried the patches on minion.de. At first it didn't work; I did the thing with commenting out the irq line in nv-linux.h, and the kernel module was built, but it said that the kernel module wasn't created! So I took a peek in the .manifest, and also in the usr/src/nv, and also at the README. I realized that the minion.de patched .manifest expects nvidia.ko, which was NOT being created by the make process! So, I changed references to nvidia.ko to nvidia.o, and it was able to install fine. I'm not sure if this will cause any problems, what is this mysterious nvidia.ko file? I have never had an nvidia.ko file on any Mandrake system that I have used. Anyone know more about this?
.ko is the newer 2.6.x kernel convention for modules. It sounds as if you made a regular .o file, renamed it to a .ko and the 2.4 modutils (like modprobe) didn't care about the suffix. (or are you running a 2.6 kernel?)
Papol is offline   Reply With Quote
Old 12-23-03, 12:30 PM   #22
zander
NVIDIA Corporation
 
zander's Avatar
 
Join Date: Aug 2002
Posts: 3,740
Default

Right, I should've mentioned this; Linux 2.6 kernel modules have the .ko suffix by convention, while it is .o on Linux 2.4 and earlier kernels. The Linux 2.6 installer doesn't try to be smart, yet, and simply assumes Linux 2.6; it will consequently fail on Linux 2.4. I agree that my earlier statements were confusing with respect to this.
zander is offline   Reply With Quote
Old 12-23-03, 12:38 PM   #23
Papol
Registered User
 
Join Date: Dec 2003
Posts: 31
Default

Zander,

I just realized there's a new installer on the ftp nvidia site:

ftp://download.nvidia.com/XFree86/nv...r-1.0.5.tar.gz

if you're interested. Your old patch applies in hunks but isn't 100% clean.
Papol is offline   Reply With Quote
Old 12-23-03, 01:14 PM   #24
Azureflare
Registered User
 
Join Date: Dec 2003
Posts: 2
Default NVIDIA drivers

Oh, great! I'm glad to hear that it isn't a problem. Yeah I'm not running 2.6, I'm running 2.4.22 right now. Anyway, thanks for all the assistance! I'm glad it's working now =)
Azureflare is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


Similar Threads
Thread Thread Starter Forum Replies Last Post
CPUMark99 - how do you compare fuelrod Benchmarking And Overclocking 66 07-19-11 08:32 AM
Athlon XP 2000 wont get recognized properly...!! demonized CPUs, Motherboards And Memory 9 10-10-02 05:01 PM
AMD 2400+ and 2600+ Benchmark Extrapolations savyj CPUs, Motherboards And Memory 2 08-17-02 09:32 PM
Athlon 2400+ and 2600+? 333 MHz Front-side bus?? PaiN Rumor Mill 26 08-16-02 10:49 AM
Help with an athlon XP.. demonized CPUs, Motherboards And Memory 23 08-07-02 09:37 AM

All times are GMT -5. The time now is 11:59 AM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.