KDE crashing. Nvidia and AMD the cause??

[Pepse]

Well, before I elaborate on my problem I will mention I did a search on this site for AMD probs and didn't get any help.

Now, as for KDE crashing. This came about when I downloaded and installed the second to the newest nVidia driver (6106). After installing the driver and rebooting out of init 3 to get back to my desktop it would boot until just before the nVidia splash screen and take me to a konsole. So, I ended up re-installing the default nVidia driver that is on Mandrake 9.1 Powerpack and that allowed me to get back to normal, sort of. As KDE is loading the desktop it crashes; "kcminit" crashed and caused a signal 11 etc., etc., etc.....SIGSEGV. Well just about anywhere I go something crashes. Such as "kppp", "konsole", "ksplash, the list goes on. Also, I have re-downloaded and tried to install the 6106 drivers and also the 6111 drivers with the same results. And I had an extra hdd that I put MDK 9.1 PP on and had the same results. Before I forget, my system is a Tyan KT-A mobo, a 1.33 Gig AMD Athlon sckt "A" CPU, a nVidia GeForce 2 MX 400 64 meg PCI vid card, 512 megs of PC133 SDRAM, Sound Blaster Live 5.1 snd cd, WD 30 gig UDMA hdd, and a 450 watt power supply. Anyway I have had this mobo for about 3 months, before that I had a Tyan Trinity 400 mobo with an Intel 600 meg CPU. That is all that changed 3 months ago. About 5 months ago I had downloaded the latest nVidia driver (5xxx) and had no problems running it. When I started running Linux 3 years ago I remember something about changing a file by adding a line that ended with " =nopentium ". I have been told on another board that the issue with AMD and (apparently) nVidia was resolved and there is no need for that line. I am trying to find out what the full command is and where to edit it in. I mean, unless someone has an idea of something else to try I am willing to try this and see. I can't see where it would hurt. IMHO it's either going to cure it or nothing will change.

Oh yeah, I also went to nVidia and downloaded an older driver, 4363, and had the same problems.

Later. Pepse.

[bsm2001]

Tou might want to look at this thread HERE (http://www.justlinux.com/forum/showthread.php?s=&threadid=131669)

I would also recomend upgrading to the latest MDK 10. Post your XF86Config-4 file.

[Pepse]

As for upgrading to MDK 10.0? I will as soon as I have the extra money to get the 10.0 PowerPack. Now, as for my XF86Config-4?? What you are curious about would be under "Modules". There is NOT a line that says either "Load Glcore" or "Load DRI ". There is a line that says: Load "usr/X11R6/lib/modules/extensions/libglx.so" . I tried to remove it and replace it with: Load "glx" # 3d layer" ; well after not getting anywhere and re=installing the default drivers and when I was back in XF86Config-4 I noticed that the line I tried to remove was still there and also it did add the Load "glx" # 3d layer . And in the Device section is does show "nvidia" . As for the link you provided I went there and the only thing I figure you want me to see is to try the "6111" drivers. As I said, been there, done that, no change.

Pepse.

[cybertron]

I've had the same thing happen when I go from the nvidia drivers to the default Mandrake ones. A bunch of KDE apps crash before and after I log in and eventually it just takes me back to the login screen.

As far as not being able to get into X when you reboot after installing, have you tried installing the drivers again from the command-line when X fails to start? I have to do that almost every time I install new nVidia drivers because the first reboot after I install them X refuses to start with an error message about "No usable screens found."

[Pepse]

So, when I get sent back to the commandline just re-login as root and then cd to the directory with the drivers and do this all again?? Just making sure I follow you on this. I'll have to wait until this evening when I get back from work. I'll check this thread first for an answer.

Later. Pepse.

[cybertron]

Yep, sounds like you follow me. Hope it works for you too.

[Pepse]

Nope, that didn't change anything. But, I am glad I went back to my XF86Config-4 file. I discovered that in the line for modules for glx I had "/glx" , so I removed the "/" and I also removed the line: "/usr/X11R6/lib/modules/extensions" , I was hoping that would help but no luck. So, when "X" crashed and sent me back to the command line I re-did everything and it didn't do any good. Just like before. I ended up re-installing the default drivers.

Later. Pepse.

[XiaoKJ]

I think it has nothing to do with the nvidia drivers cos its already apparently working by giving you X

if you are not convinced, do glxgears in X and if you can get it running, then you shouldn't worry about XF86Config-4 and nvidia drivers, but rather binary files may be loose on your filesystem. Check that you have enough space, and you don't have hard disk errors.

Finally, please use point form next time -- its really hard making my eyes stay on your post...:D

[cybertron]

Originally posted by XiaoKJ
I think it has nothing to do with the nvidia drivers cos its already apparently working by giving you X


But he said X crashes when he tries to start it with the nvidia drivers. I'm guessing he won't be able to run glxgears at all.


Finally, please use point form next time -- its really hard making my eyes stay on your post...:D

Good point.:)

Pepse, can you tell us what error it gives you when you try to startx? When you get dropped to the command-line login, login as a regular user and type "startx". If we know exactly what error message you get it might help.

[XiaoKJ]

Originally posted by cybertron
But he said X crashes when he tries to start it with the nvidia drivers. I'm guessing he won't be able to run glxgears at all.

But then why can he get to click on kppp and the likes in his first post? seems funny... I don't think its nvidia drivers -- REALLY!

Maybe you could try to clarify, pepse?

[cybertron]

Originally posted by XiaoKJ
But then why can he get to click on kppp and the likes in his first post? seems funny... I don't think its nvidia drivers -- REALLY!


That happened after he went back to the default drivers.

So, I ended up re-installing the default nVidia driver that is on Mandrake 9.1 Powerpack and that allowed me to get back to normal, sort of. As KDE is loading the desktop it crashes; "kcminit" crashed and caused a signal 11 etc.

I guess I don't know what is included in the powerpack though. Maybe the nVidia driver is the default one there.

[XiaoKJ]

Originally posted by cybertron
I guess I don't know what is included in the powerpack though. Maybe the nVidia driver is the default one there.

I think maybe mandrake had the nvidia drivers patched to third-party locations, and wouldn't let users use the nvidia default drivers (6111)

[Pepse]

OK, I re-installed the 6111 drivers and of course I had to go back to the default nVidia drivers, which are called 4321-3mdk . Anyway I did do "startx" and here is (hopefully enough info)

Using vt 7
(EE) NVIDIA (0): Failed to initialize the NVIDIA kernel module!
(EE) NVIDIA (0): *** Aborting ***
(EE): Screens found, but none have a usuable configuration.

Fatal server error:
no screens found.

X10: Fatal IO error 104 (Connection reset by peer)
":0.0"
After 0 requests (o known processed) with 0 events remaing.

If it ain't enough I will copy down everything tonight after work and repost it.

Point Form?? I'll use it if I can figure out how. I assume it is a font sizer?? Of course glxgears doesn't work. The error is:

Xlib: extension "GLX" is missing on display ":0.0"
Error: Couldn't get an RGB, Double-buffered visual.

Hope this helps.

Later. Pepse.

[bwkaz]

Originally posted by Pepse
(EE) NVIDIA (0): Failed to initialize the NVIDIA kernel module!
(EE) NVIDIA (0): *** Aborting *** This error is explained in the nVidia README file; you might want to read it.

;)

[Pepse]

OK, I guess I should have looked in the nVidia-readme more thoroughly; but that really wouldn't have crossed my mind. So, from what I see here is that I probably need MDK 10.0? But, if that is true why did I have the same problem when downloading and older driver, such as 4363?? This was after I had probs with the 6111 drivers. Now the problem will be trying to figure out how or what to do to correct this problem, because I have never been to /sbin/lsmod/ or loaded anything through modprobe or insmod. But I will try to figure things out on Fri.

Pepse.

[bwkaz]

Originally posted by Pepse
So, from what I see here is that I probably need MDK 10.0? No.

You need to use the right kernel sources when you compile the driver. ;)

Actually, before you try that, follow some of the other suggestions given in the README and run dmesg after trying to "/sbin/modprobe nvidia" as root. You should have some kind of error in there if one is being produced.

If you don't get any errors, try to startx as your normal user at that point. You probably won't get that error anymore if that happens.

At that point, you'll need to make nvidia modprobe itself automatically when your distro starts. Different distros have different methods of doing this, but most of them have a file something like /etc/sysconfig/modules that lists modules to load at boot time.

[Pepse]

Sorry for being late responding. Been gone from computer since late Friday night. Anyway, I did a "/sbin/modprobe nvidia " and here is (I hope) what you are curious about:

Warning: loading /lib/modules/2.4.21-0.13mdk/kernel/drivers/video/nvidia.o will taint the kernel: non-GPL license-NVIDIA

see http://www.tux.org/lkml/#export-tainted for information about tainted modules

/lib/modules/2.4.21-0.13mdk/kernel/drivers/video/nvidia.o init_module: Device or resource busy

Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters

You may find more information in syslog or the output from dmesg

modprobe: insmod /lib/modules/2.4.21-0.13mdk/kernel/drivers/video/nvidia.o failed

Modprobe: insmod nvidia failed

There you have it. I did "dmesg" and I really don't know what I should be seeing. I pretty much read every line and I don't think there is anything amiss.

Don't ask me to cut-and-paste or whatever because I, to this day, I don't know how to do that. And we're talking early to mid 90's of dealing with windows based apps.

Pepse.

I followed that link that was created with my post. It looks to me that what XiaoKJ said about 3rd party drivers. It don't look good for my sys.

[bwkaz]

Originally posted by Pepse
Warning: loading /lib/modules/2.4.21-0.13mdk/kernel/drivers/video/nvidia.o will taint the kernel: non-GPL license-NVIDIA

see http://www.tux.org/lkml/#export-tainted for information about tainted modules That's no big deal (if you want to se details, check out that URL). It just basically means that the kernel developers will not help you if you think you've found a bug in the kernel, because that "bug" could have easily been caused by the nVidia driver, and they can't get source for it.

/lib/modules/2.4.21-0.13mdk/kernel/drivers/video/nvidia.o init_module: Device or resource busy That one is a problem.

Are you running kernel 2.4.21-0.13mdk (is that what uname -r says)?

Try doing a dmesg | grep NVRM to see what the nvidia module wrote to the kernel log. The only reference to EBUSY (which is the error code that corresponds to "device or resource busy) that I see in the entire 1.0-6111 driver is a check on the return value of the call to request_irq -- if request_irq returns EBUSY, then the nvidia driver will write a line to the kernel logs saying something about the IRQ that it couldn't get.

Does your BIOS have the "Assign IRQ to VGA" option turned on? Does it have "PnP OS" turned off?

Don't ask me to cut-and-paste or whatever because I, to this day, I don't know how to do that. Easy -- select whatever you want to copy, then (while that program is still running) click the middle mouse button inside the textbox in your web browser.

;)

Selecting automatically does a copy, and middle-mouse is paste. Some programs also have a second clipboard that you can use explicit cut/copy/paste actions with (and those that do will all use the same global clipboard), but xterm, for example, won't do that, so if you're copying from xterm into e.g. Firefox, you need to select in xterm and then middle click in Firefox.

But I don't need that, unless you can't figure out anything that the dmesg | grep NVRM is saying. (You have to run that after you try to /sbin/modprobe nvidia, just in case you're wondering... ;))

It looks to me that what XiaoKJ said about 3rd party drivers. No, it doesn't. What XiaoKJ said about third party drivers was basically wrong (sorry XiaoKJ, but the problem has nothing to do with any patches that Mandrake made to the OP's kernel; nVidia's driver is maintained well enough that it can work around all of them -- and besides, even if Mandrake was adding incompatible patches to their kernel, nVidia distributes precompiled kernel modules that work on Mandrake, so they have to be able to compile it and get it to work somehow).

[Pepse]

Yes, kernel 2.4.21-0.13mdk is what I am running. As for my BIOS "Assign IRQ to VGA is enabled" and "PNP" is on. As for the latter when I started with MDK 8.1 I was told to turn off PNP. With 9.0mdk I was told to leave it on. I now have 9.1 PP, and if you want me to turn it off no problem.

Now as for the main event, if I understood you right first I did /sbin/modprobe nvidia then I did "dmesg | grep NVRM " , which says:

NVRM: loading NVIDIA Linux x86 NVIDIA Kernel Module 1.0-6111 Tues Jul 27 07:55:38PDT
NVRM: register chrdev failed

If that don't seem like enough info I am not sure what I might have did wrong. To me that reply doesn't seem like what you are looking for.

As for the font style I use it is the default for this site. I will go to a larger font if I can figure out what it means when I clck on the "size" and I get a box that says: Enter the text to be formatted with the specified font. WHAT??

Later. Pepse.

[bwkaz]

Originally posted by Pepse
As for my BIOS "Assign IRQ to VGA is enabled" and "PNP" is on. As for the latter when I started with MDK 8.1 I was told to turn off PNP. With 9.0mdk I was told to leave it on. By who? (Actually, don't bother answering that, it doesn't really matter.)

Having "PnP OS" turned off will never cause problems with any OS. Having it on sometimes causes problems with some OSes (and on Linux, it seems to depend on the kernel version and any patches the distro may have applied -- not necessarily even in the realm of PnP support).

Turning that setting off means that the BIOS will initialize every PCI device on the system. There's no way this can possibly be bad at all. Turning it on means that the BIOS won't initialize devices that it doesn't think are critical for booting -- and if the OS doesn't know how to initialize those devices, or it doesn't do it properly, or it doesn't do it at all, then that device won't work.

Now as for the main event, if I understood you right first I did /sbin/modprobe nvidia then I did "dmesg | grep NVRM " , which says:

NVRM: loading NVIDIA Linux x86 NVIDIA Kernel Module 1.0-6111 Tues Jul 27 07:55:38PDT
NVRM: register chrdev failed Hmm... that's not actually the place I thought it was erroring yesterday. However, it does help.

That message is printed when NV_REGISTER_CHRDEV returns something less than 0. NV_REGISTER_CHRDEV is defined based on whether the kernel source you compiled against had devfs turned on or not (try grep DEVFS /lib/modules/$(uname -r)/build/.config to find out). If devfs is turned on, then NV_REGISTER_CHRDEV is defined to be devfs_register_chrdev, and if not, it's defined to be just plain old register_chrdev.

If that function fails, then the whole nvidia_init_module call fails with th same error as the registration function returned (which is -EBUSY). I don't have the source for your kernel, so I can't tell what might cause your kernel's registration function to return -EBUSY, either, but my kernel will return -EBUSY when the requested major device number is already in use.

What other drivers do you have running, that might not be observing the nVidia device major number allocation of 195? Anything else from outside the kernel tree?

[Pepse]

OK, I did "grep DEVFS /lib/modules/$(uname -r)/build/.config" and the reply was: grep: /lib/modules/2.4.21-0.13mdk/build/.config: No such file or directory .

So, either it is broke and I should re-load MDK or I somehow didn't do the command right or I should have been within another program or????

Other drivers that would be running?? I have a printer driver, scanner driver, sound card driver. I don't think any of these are what you are asking about, but I don't think there is anything else that would be running that would be walking on the nVidia drivers. I turned PNP off.

Pepse.

[bwkaz]

Originally posted by Pepse
OK, I did "grep DEVFS /lib/modules/$(uname -r)/build/.config" and the reply was: grep: /lib/modules/2.4.21-0.13mdk/build/.config: No such file or directory. Um... are you SURE you have your kernel-source package installed? The one whose version matches your kernel version exactly?

What does ls -l /lib/modules/$(uname -r)/build say -- is the build symlink there? Is it pointing to the root of the proper kernel source directory?

Other drivers that would be running?? I have a printer driver, scanner driver, sound card driver. Which ones? Are any of them closed source like nVidia's?

What is that printer driver?

I don't think there is anything else that would be running that would be walking on the nVidia drivers. It's not that it's specifically being anti-nVidia, necessarily. It's just using a device major number that it shouldn't be using because it's been reserved for nVidia.

[Pepse]

Sorry for the delay in replying; been a crappy week at work. Anyway I did "ls -l /lib/modules/$(uname -r)/build" and the reply was "No such file or diectory." Now the way I see it you are having me check the status of the MDK Kernel?? If so it seems I have a problem, a big problem. As I mentioned in the beginning, about 3 months ago I went from an Intel board to an AMD board (same brand, Tyan), anyway it was acting a bit goofy so I re-installed MDK 9.1PP without really changing anything. It was kinda like installing MDK over MDK. I did it mainly so get the kernel to recognize the AMD CPU. So, if I am right about what is being checked here then it looks like I may have to completely clean up my hda drive.

The printer is an HP Photosmart 7150. The one thing to note is I didn't add or subtract anything after switching to the AMD cpu. The cards are in their proper slots, too.

Later. Pepse.

[bwkaz]

Huh? You don't need to reinstall anything to get the kernel to "recognize" the AMD CPU... the way most distros come configured, their kernel will run on ANY Intel 686 clone (from the Pentium Pro and Pentium 2, up through the Pentium 4, and from the AMD K6-2 (or something early like that) up through the AMD Opteron, since even the Opteron can run plain old i686 programs in its 32-bit compatibility mode).

That's why AMD CPUs are Intel compatible. ;)

I bet you just don't have kernel-source installed (or if you do, its version doesn't match your kernel version). It used to be on Mandrake's CD 3; check around on there, or on Mandrake's FTP site if this isn't the same kernel as your version of Mandrake installed.

[Pepse]

OK, but before I look around on my CD's I remembered something that might be of importance. As I mention in my first post I also took another empty hard drive and installed 9.1PP on it and downloaded the 6111 drivers and had the same problem. So, I think there might be something wrong with my mobo or something on it. What do you think.

Pepse.

[Pepse]

OK, hopefully I am learning something here. I installed the kernel-source, it is the same as what is in my box. If I understand this right I installed this to help my kernel get along with the nvidia kernel? But, what I don't understand is if I had to install the kernel-source then why/how was my box running before I did this?? And whilest I was in MCC package installer I did a search for "kernel" and came up with these 2 items I think my be useful to my box. "kernel-multimedia-2.4.21-0.16mdk-1-1mc" (Description: A preemptible Linux Kernel, which reduces the latency of the kernel), and "kernel-multimedia-source-2.4.21-0.16mdk" (Summary: The source code for the Linux Kernel. Description: The kernel-sources package is specifically for use with the multimedia kernel, which includes preemption and low latency patches.)

True, I have no idea if these would do me any good but I thought I would ask.

One more thing. Is it safe to shut down my box?? I mean by installing the kernel-source I should not have any issues with restarting and running the box, RIGHT??

Pepse.

[cybertron]

Originally posted by Pepse
One more thing. Is it safe to shut down my box?? I mean by installing the kernel-source I should not have any issues with restarting and running the box, RIGHT??

Yes. The kernel source doesn't mess with your current kernel unless you compile it and overwrite your current kernel.

As far as why it worked before, that's probably because there was a precompiled kernel interface for your kernel and that driver version. If the new driver didn't have one of those, then you would need the kernel source to install it.

[bwkaz]

Originally posted by Pepse
I also took another empty hard drive and installed 9.1PP on it and downloaded the 6111 drivers and had the same problem. So, I think there might be something wrong with my mobo or something on it. Not likely. It's more likely that you just didn't install kernel-source on that drive either. ;)

I installed the kernel-source, it is the same as what is in my box. If I understand this right I installed this to help my kernel get along with the nvidia kernel? You have installed kernel-source so that the nVidia driver can use the headers that your kernel-source package contains to adapt itself to your running kernel. It does, after all, become an integral part of the kernel when you modprobe it (just like any other kernel module does), so it needs to know a lot of information about the kernel. That information comes from these header files.

These files are in /lib/modules/<kernel version>/build/include, and the kernel-source package is what creates the "build" part of that path.

But, what I don't understand is if I had to install the kernel-source then why/how was my box running before I did this?? It wasn't running any 3D capable driver, unless that was the nVidia driver, and it had been compiled by somebody else... but beyond that I have no idea.

And whilest I was in MCC package installer I did a search for "kernel" and came up with these 2 items I think my be useful to my box. "kernel-multimedia-2.4.21-0.16mdk-1-1mc" (Description: A preemptible Linux Kernel, which reduces the latency of the kernel), and "kernel-multimedia-source-2.4.21-0.16mdk" (Summary: The source code for the Linux Kernel. Description: The kernel-sources package is specifically for use with the multimedia kernel, which includes preemption and low latency patches.) Neither of those is necessary. However, if you install the one (kernel-multimedia-whatever), you will need to install the other one (-source) before recompiling the nVidia driver (and you will need to recompile the nVidia driver every time you change kernels).

Basically, this kernel just has some patches applied to it to make desktop interactivity better. It won't help getting nVidia installed though.

One more thing. Is it safe to shut down my box?? I mean by installing the kernel-source I should not have any issues with restarting and running the box, RIGHT?? You're installing source code, not a kernel itself. So no, you shouldn't have any issues booting (the BIOS and kernel control that, not the kernel source -- at least it doesn't control it directly, until you use that source to compile an actual kernel).

[cybertron]

Ooh, bwkaz, 9999 posts. Do we get to have a party when you hit 10000?:D

[bwkaz]

Hmm... looks like I did already.

Whee!

http://kadzban.is-a-geek.net/blowhorn.gif

;)

[Pepse]

:confused: Well folks it seems I have a major issue with my system. Okay, as I said I installed the kernel-source and after it was said I could shut down I decided just to "reboot". Well I went to "logout""pepse", and "end session" and the next screen I clicked "reboot" and the next screen was empty. So, I went "back" and relogged in. No problem, so I decided what the heck I'll just shutdown and see what happens. So I went to "logout" and the little green lizard screen came up and I clicked "reboot" and it did. When it booted up I still have the "KDE Crash Handler" problems. So, I decided to once again reload the nvidia 6111 drivers. Same ol', same ol'. No different than before. My box won't take the new drivers, and has a problem running the default drivers. The KDE you-name-it-it-crashes. I would have posted sooner but I had some other issues that I did when editing my XF86Config files. Hey, at least I didn't panic like I use to. Went to visit a relative came home sat down and figured it out.

[bwkaz]

2.4.21 -- did that include the AMD "mem=nopentium" fix? There was an issue with AMD speculative caching and AGP that afffected both older 2.4 Linux kernels and older Win2K kernels. Newer kernels have it fixed, but older kernels required that you boot with mem=nopentium for them to work. Use your bootloader to temporarily edit the boot line; this may be the issue.

<only half-joking> Or just upgrade your kernel to be sure... </only half-joking>

[Pepse]

"Use your bootloader to temporaily edit your bootline, this blah, blah, blah." Ain't never been there. Give me a map and I'll do it.:D All I can say to for my copy of MDK 9.1 PP, is that it came directly from Mandrake in France around the first or second week of Sept., '03. But, even if you think it is new enough give me the commands anyway. It can't hurt.

Oh, 2 half jokings equals a full joking, right??:D :D

Seriously, my version of upgrading my kernel is when I can get 10.0. But in the meantime I am learning more by having this problem and am very grateful for people on this list to help me, and help me learn more. :)

Later. Pepse.

[bwkaz]

Originally posted by Pepse
Ain't never been there. Give me a map and I'll do it.:D That will depend on which bootloader you use. lilo or grub?

I think the Mandrake default is grub, but I'm not sure.

One way to find out might be to look for an /etc/lilo.conf file and a /boot/grub directory. If you have the former, then you can run lilo (that doesn't tell you for sure that you are, though). If you have the latter, then you can run grub (but again, it doesn't tell you that you are for sure).

If you have only one of those, then you're likely running that bootloader. If you have both, then ... I guess try the grub instructions first, or something. ;)

For grub, to add that option to the kernel command line, you have to select your Linux installation and then hit the 'e' key to edit it. Then go down to the long line that looks like a kernel (it won't start with "root", it'll be the other line). Hit 'e' again when that line is highlighted, and you'll get a line where you can edit. Add a space to the end, and then add 'mem=nopentium' (just without the quotes). Then hit enter on that line to save the changes, and hit 'b' to boot.

For lilo in text mode, move the selection at least once, and then move it back to the line you were on. Then just hit the spacebar and type in 'mem=nopentium' (again, without the quotes -- you'll see that what you're typing gets added after whatever text you had selected). Then hit enter to boot.

For lilo in graphical mode, you have to hit some other key combination to get to a point where you can do the same thing as lilo in text mode -- I think it's Ctrl-X. At that point just follow the text lilo instructions (hit space, then type it in, then hit enter).

[cybertron]

Actually, Mandrake defaults to lilo.

And the easiest way I know of to add an option like mem=nopentium to a LILO boot option is just to hit 'esc' and then type the name of the option you want to load, a space, and then the mem=nopentium part. That may override other appends you have configured in the lilo.conf file though, so if there's anything important you probably want to type that too.

[Pepse]

:confused: Either I am dumb and missed something or you guys forgot something. OK, lilo is my bootloader. I've always went with it since I started with Mandrake Linux. Looking at both posts, neither one seems to detail how to get to where I gotta go to edit. I went to a konsole and hit ESC twice and the reply was: Display all 2801 possibilities? y or n. So, I did "Y" and got 2801 items and scrolled down until I seen "lilo" and "lilo-bmp2mdk". So, where ever I was at least showed something about lilo. But, I get the feeling that I should be somewhere else? Like vim or mc. Or maybe I should do something on bootup??

Pepse.

[cybertron]

Whoops, my mistake. You hit escape during boot when it gets to the selection screen (the graphical menu that lists the different options you can boot).

[Pepse]

Okay, let's see if I did this right (I doubt it). I am on my other hdd that has the same issues as my big one. I booted up the box and hit the escape key when it got to the selection screen and it put me in a text mode with linux linux-nonfb failsafe and floppy at the top of the screen. Then it had "boot:" so I typed in lilo and the reply was "no such image [Tab} shows a list." So I hit tab and got the same reply except that it showed "Boot: lilo". So, I hit the space bar and typed "mem=nopentium" after lilo. Hit the enter key and nothing happened so I did ctrl-alt-del and the box rebooted and everything booted up fine. That bothers me because if I did it right then it was too easy. That bothers me. What do you think??

Pepse.

[cybertron]

You'll want to type "linux mem=nopentium" at the Boot: prompt. I'm not sure why lilo was being weird when you hit tab.

[Pepse]

Well, the good news is I was able to add the "mem=nopentium" line to my bootloader (yes it is really easy). The bad news is that nothing else changed. I still have KDE crashing all over (I'm on my bigger hdd, now). After editing the bootloader I went and re-installed the nvidia 6111 drivers and as usual I had to go back to the MDK/Nvidia 4321 drivers. Now, I have a question about what someone on Mandrake's board told me to do when editing XF86Config-4. He said that to make sure there is a line in "Modules" that says: Load "glx" # 3d layer . I did have to put that line in. My question is should 3d layer be all lower case or should it be: # 3D Layer ?? Also, the line I removed said: Load: "/usr/X11R6/lib/modules/extensions/libglx.so" . At least I see that the "mem=nopentium" line didn't cure my problem so there is an error somewhere else. That is why I mention the XF86Config-4 part. And yes, under "Device" is has "nvidia".

Pepse.

[cybertron]

You don't have to worry about anything on the line after a '#' symbol. That's just a comment. Unfortunately, I'm not really sure why it still isn't working. Since the mem=nopentium thing was bwkaz's we'll hope that he has an idea.:)

[Pepse]

I more or less asked for that command in the beginning hoping it would help. So, we will wait and see what bwkaz might think.

Pepse.

[bwkaz]

Originally posted by Pepse
Also, the line I removed said: Load: "/usr/X11R6/lib/modules/extensions/libglx.so" That's good; when you say Load "ZZZ", X looks for a /usr/X11R6/lib/modules/extensions/libZZZ.so file, and then for a /usr/X11R6/lib/modules/extensions/libZZZ.a file. It will load the first one that it finds. (At least, this is true with an X.org downloaded X tree, or an XFree86 downloaded X tree, if you configure them both properly -- some distros may patch them so this behavior is different, or they may not configure them properly. In other words, YMMV, but this is the standard.)

Therefore, Load "glx" should load up the same extension file as the Load line you deleted.

If adding mem=nopentium to your boot line didn't help, then I'm guessing that your kernel did already have that fix. But as for what to try next... hmm, I'm not really sure.

When KDE crashes, does your kernel go with it? (Do you get booted back to an X login or your text console? Or don't you get any screen output after it dies?) If not, then it's (most likely) not the nVidia kernel module itself, but rather either the GLX extension, the OpenGL library, or something like that (something that gets linked into X or one of the KDE programs). I assume by "crash" you mean "segfault", correct?

[Pepse]

OK. Yeah, it is the "SIGSEV" crashes. As for "KDE Crashes" it does it on boot up when it is loading the "periphials"; the one that shows a globe as it is preparing desktop. As for what things are doing that I will say that it is just about anything KDE; like "kcminit", "ksplash", "kppp", "konsole", "kmserver", .......................... the list goes on and on. It has been happening for about, what?, 3 weeks?? The system don't reboot, or freeze, or whatever. Just these annoying little "crash" happenings.

Oh, in reference to GLX?? When I go to a console and try to get GLXGEARS, I get:

Xlib: extension "GLX" missing on display ".0.0."

Error: couldn't get an RGB, double-buffered visual

Later. Pepse.

[bwkaz]

So the glx extension still isn't loading... huh.

<grasping at straws>

Are you sure you've been editing the right config file? Check /var/log/XFree86.0.log if you use XFree86, or some other X.org log file (in /var/log; possibly /var/log/Xorg.log?) if you use X.org. It will say which config file it's loading -- make sure that that config file is set up properly regarding glx, etc.

</grasping at straws>

If you run ldd on each of the crashing executables, are there any libraries in common that they all load? Other than libc and libstdc++; those should be linked to all KDE programs. (Actually, do any other C++ programs work without crashing on this machine?)

[cybertron]

Originally posted by bwkaz
<grasping at straws>

Are you sure you've been editing the right config file? Check /var/log/XFree86.0.log if you use XFree86, or some other X.org log file (in /var/log; possibly /var/log/Xorg.log?) if you use X.org. It will say which config file it's loading -- make sure that that config file is set up properly regarding glx, etc.

</grasping at straws>


Good point. Mandrake actually uses the XF86Config-4 file, and it isn't particularly obvious that you need to use it instead of the regular XF86Config file.

[dwill525]

Sig11 errors ( post #8)

If you are using Mandrake 10, w/Nvidia card you will have to configure your "XF86Config-4" file with kwrite as root

The following two sections should look something like this.

Section "Module"
Load "dbe" # Double-Buffering Extension
Load "v4l" # Video for Linux
Load "extmod"
Load "type1"
Load "freetype"
Load "glx" # 3D layer

Section "Device"
Identifier "device1"
VendorName "nVidia Corporation"
BoardName "NVIDIA GeGorce2 DDR (generic)"
Driver "nvidia"
Option "DPMS"
Option "IgnoreEDID" "1"
Option "NvAGP" "3"

[cybertron]

He's actually running the 9.1 (9.2?) powerpack, but the config should be essentially the same.

[Pepse]

I am running XFree 4.3. And Mandrake 9.1 PP. Now as for the "Modules" section the only difference is the last line. Mine has:

Load: "/usr/X11R6/lib/modules/extensions/libglx.so"

And my "Device" section is the same for the first 5 lines. I don't have an AGP card; and I presume line 6 has something in common with line 7 in reference to AGP.

Now the reason the last line in modules is different is because I am back to the default MDK drivers.

As for crashing. I get a big box on my screen after I exit from any KDE programs. Gnome is not affected. Anyway the top of the box has: X unknown - The KDE Crash Handler. Then below that it has "General" and "Backtrace". In "General" it has a "Short description" (such as) The application unknown (konsole) [or whatever KDE app I exited] crashed and caused the signal 11 SIGSEGV . Then it gives "A short description" such as SIGSEGV is a signal due to a bug in the system. Now when I click on "Backtrace" I get a screen with: (no debugging symbols found)...(no debugging symbols found) [that line is repeated 13 times], then it says

#0 0x40f13677 in waitpid ( ) from /lib/i686/libpthread.so.0

#1 0x40639e7b in KCrash: :defaultCrashHandler (int) ( ) from /usr/lib/libkdecore.so.4

#2 0x410803b in _libc_sigaction ( ) from /lib/i686/libc.so.6

#3 0x4090180b in qt_cleanup( ) ( ) from /usr/lib/qt3/lib/libqt-mt.so.3 .

Other KDE apps that crash have very similar lines in the backtrace area.

Hope all this helps.

Pepse.

[kozaki]

Running MDK-10.0 with xorg-x11-6.7.99.902-0.cvs20040820.1mdk.i586, booting on 2 different kernels.
I have the *same problem* than Pepse after installing xorg-x11 along with Nvidia 6111 driver.

i discovered i somehow can lauch X only with '$ kde', henceforth KDE-3.2 starts, shows same mess error :
The application unknown (konsole) [or whatever KDE app I exited] crashed and caused the signal 11 SIGSEGV .
"A short description" such as SIGSEGV is a signal due to a bug in the system.
Now when I click on "Backtrace" I get a screen with: (no debugging symbols found)...(no debugging symbols found)...

whatever i try t oinstall Nvidia driver with options such as :--add-this-kernel or --extract-only (+ make install) does change nothing.
I eventually reinstalled kernel-sources-2.6.7.14sds.1phoenix & then couldn't get the Nvidia driver installed. Logs for that are :
modprobe: FATAL: Error inserting nvidia (/lib/modules/2.6.7-14sds.1phoenix/kernel/drivers/video/nvidia.ko): Invalid module format
~ kernel: nvidia: no version for "struct_module" found: kernel tainted.
~ kernel : Loading module: nvidia
~ kernel: nvidia: version magic '2.6.7-14sds.1phoenix preempt 586 gcc-3.3' should be '2.6.7-14sds.1phoenix preempt 586 gcc-3.4'
~ kernel: nvidia: no version for "struct_module" found: kernel tainted.
~ kernel: nvidia: module license 'NVIDIA' taints kernel.

So i booted back on kernel-2.6.3-7, reinstaled Nvidia-6111... and still have the same KDE apps krash.

# glxgears -info
Xlib: extension "GLX" missing on display ":0.0".
glxgears: Error: couldn't get an RGB, Double-buffered visual.

System runs apparently fine, but will crash on rare but various conditions (shifting tty, into various applications).

Having read this post, i now cannot believe anymore it may be xorgg-x11 related, but cannot figure what esle to try.

Now i edited my /etc/X11/xorg.conf accordingly to dwill525's and will reboot, then come back here to say what happened.

[kozaki]

This a known bug, Mandrake team workin on it.

More details :
* http://qa.mandrakesoft.com/show_bug.cgi?id=11060
* http://linuxfr.org/forums/14/3277.html (in french, bottom of the page)

" c'est un bug connu et l'équipe de mandrake travail dessus, mais rencontre beaucoup de difficulté avec le driver qui malheureusement est proprio....voili voilou, donc pour ceux qui comme moi n'ont plus l'accélération 3D :c'est normal et c'est en cours de correction."

[Pepse]

Somehow I missed Kozaki's replies, but after going to the english link and reading it all I see that it appears to be an nvidia problem. The one thing mentioned in that link is the problem is with kernel 2.6.x.x. I am running 2.4.21-0.13mdk. Now what I have done since my last post is I reformatted "hda" and of course everything was fine and good again. That is until I re-downloaded the 6111 drivers and installed them, same problem of when it was booting up it took me to a shell. So, I decided the he** with it and re-formatted "hda" . I am definately thinking of switching to ATI. They can't be any worse than nvidia.

Later. Pepse.

[bwkaz]

Actually, they are, at least for the few people that I've talked to that have used both...

There are a lot of people that are using the nVidia drivers with no problems whatsoever. Therefore, your issue must be something unique about either your system (your hardware), your kernel, or your kernel sources.

With ATI, there are a lot of posts that I've seen where people install the ATI drivers, then games don't recognize that hardware OpenGL is actually in use (because for some mysterious reason, the OpenGL renderer string is still "Mesa software renderer" or some such). I don't have any good way to even start to troubleshoot that.

With your issue, the segfaults in KDE, it wouldn't surprise me if that was due to some flaky hardware -- how big is your power supply, and what do you have for a processor, which specific video card do you use, and what speed is/are your hard drive(s)?

[Pepse]

The thought of switching to ATI came from reading a letter in the Linux Format magazine (Sept., 04). A person did mention of having success with the ATI Raedeon Linux drivers for his Raedeon 9200SE. So, I thought maybe it wouldn't hurt. But as I read your whole response it is possible that I could be worse off.

Now as for my box. It has a 450 watt power supply; ATX. The mobo it a Tyan Trinity KT-A, socket "A" with an AMD Athlon 1.33 gig CPU, 512 megs of PC133 Sdram, a Soundblater Live 5.1 sound card, a Western Digital 30 gig HDD (was new about 2 years ago); it is either 4xxxrpm or 5200rpm, LG DVD rom, LG CDRW, an Nvidia GeForce 4 FX 5200 128 meg AGP video card (BUT I just installed that about 10 days ago. I had an Nvidia GeForce 2 64 meg PCI card before that.) Also, I changed mobo's about 10 days ago because the other Tyan Trinity KT-A board had a bad AGP slot. So I bought a NEW one. So, during this whole episode I switched the mobo and video card 10 days ago. Also, to note is that I reformatted "hda" since I started using the new board and AGP card. As it stands right now I can say that GLX Gears works. As we read this I am at 620 FPS if I watch the "gears" turning and 1120 FPS when I am back here writing this; and 1232FPS when I "minimzse" the "gears" box. Oh, Interage is the maker of the card, and PNY made the other one.

An opinion of mine: I didn't really have any problems with updating the Nvidia drivers until I went from a Tyan board with an Intel CPU to this KT-A board with the AMD CPU. I am not sure why, because when I started with Linux 3+ years ago I had a Tyan S1590 mobo with an AMD K6-2 450 meg CPU, then had an FIC mobo with an AMD K6-2 500 CPU. The FIC mobo and Linux didn't get along so I replaced it with the Tyan board with the Intel chip. One other thing (for what it might be worth) the manual that came with the Intel Tyan board mentioned *Unix compatibility, this board doesn't.

Pepse.

Oh yeah, when I reformatted the hdd I installed the kernel drivers.

[bwkaz]

Does it help to revert to the GF2?

[Pepse]

Are you saying that I should try the GF2 again because I got a different ( but same type/model) board?? And regardless of your answer do you think I should load an older driver?? Something that would be pre-MDK 10.0/kernel 2.6.x.x?? Just thinking that it could have an affect on my system. I mean I didn't have any video problems before I downloaded the 6101/6111 drivers. Then too, I didn't have any problems until I went with this AMD'ed CPU board. Which makes me wonder if it is the board or video card or ??????

Pepse.

[bwkaz]

Did I misunderstand what you said? I thought you said that it worked when you had the GF2 card in there. I was thinking that it might be due to power requirements (your PSU can't supply quite enough current to the FX5200). The GF2 should be a little less power hungry (but not by much).

As far as the older driver, it might work. I try to run the latest driver possible myself (that allows me to get the newest card available too, if I have to upgrade ;)), but I do remember some people at nvnews.net having issues with the newest drivers sometimes. Some of the issues seemed to depend on their card's BIOS version, and some of them didn't... anyway, sure, I suppose it's worth a shot.

However, you won't be able to use anything before 1.0-5336 with a 2.6 kernel (unless you patch the driver; check www.minion.de for the patches, if they're even still available)...

[cybertron]

But he's not using the 2.6 kernel because he's got the 9.1 powerpack. Unless he upgraded at some point in the thread and I forgot about it, which is possible considering how long it is now:)

You might actually want to try something earlier than 5336 and see if the 2.6 compatibility of later versions broke the 2.4 compatibility. I haven't heard anything one way or the other one the subject.

[Pepse]

I still have the 2.4.x.xx kernel with Mandrake 9.1 PP. This problem started when I installed the 6101 drivers with the GF2 PCI card and is still a problem with a new KT-A mobo and the GF4 AGP card. In case you forgot, I replaced the mobo because the AGP slot was bad/didn't work. IIRC this board was made in 2001, which also means the mobo BIOS is 2001.

What I will due in the next couple of days (or sooner) is to get out an extra hard drive, put MDK 9.1 PP on it and download the 5336 nvidia drivers and see what happens.

Later. Pepse.

[cybertron]

Originally posted by Pepse
What I will due in the next couple of days (or sooner) is to get out an extra hard drive, put MDK 9.1 PP on it and download the 5336 nvidia drivers and see what happens.
Make sure you get a version < 5336 too, just in case.

[bwkaz]

OK, now that I know what's going on (:p), that sounds like a good way to go also (try 5336, and also probably 4363 or some release around there -- get them from http://www.nvidia.com/object/linux_display_archive.html).

[araneae]

i encountered some problems regarding nvidia driver(gforce2 mx).
i'm using mandrake 10 official.
after i install the nvidia driver(1.0-6111) there no more graphical login. i need to type "startx".
then some crashes.
and the games terminated immediately after i clicked on them. even my xmms is not working.
the logout window only has Logout and Cancel. before this the logout gives option ;
End Session
Restart
Turn-off

but when I log in as root. everything is OKAY. O.K!!! i can play all the games, all 3d stuffs. xmms work well.
but the logout Xwindow still give only 2 option; logout, cancel

I read all the previous post i don't get a clue how to overcome this nvida problem guess this nvidia problem is about driver itself. looking forward for the new driver.

[cybertron]

Did you update your XF86Config-4 file after you installed the drivers?

[araneae]

yes i did the confguration part as mentioned in the read me. and guess what. after all the trouble i managed to install it and it works. no more crashes. no more missing stuff. i don't remember how i did it.i guess it was like this;
init 3
su root
<password>
cd /tmp
sh NVIDIA-Linux-x86-1.0-6111-pkg1.run
startx
edit XFree86Config-4
logout
init 5

[cybertron]

I have often had to install the nVidia drivers twice when I get a new version. I'm not sure why, but especially after the first reboot after the first install X doesn't like to start. Maybe that's all that happened in your case too. Who knows?

Anyway, glad to hear that you got it working:)

[Pepse]

Well, as it stands right now something 's amiss, duncanbojangles. OK, so you read the header, did you read all 5 flippin' pages?? If you did you will note that I really prefer nVidia video cards and I've been a Tyan person for around 6 years. Sorry to be sarky but, I am just running into walls with my situation and I've got people out there trying to help me get this straight. I am sure they got other problems to solve and the sooner I can get this resolved one way or the other the better. And they will have one less troubleshooter to think on.

For those helping me. I took a needed break Saturday; went to a Renaissance Festival, had a d#mn good time. Today, Sunday, I ended up working on a couple whine-doze computers of my kin. Will get that other hard dive going soon.

Later. Pepse.

[duncanbojangles]

Yes, I read all 5 pages, but I felt that I'd restrained myself long enough to be allowed one witty comment. I'm sorry for not being able to help more, and as such, I will remove the comment if possible. I'm sorry for your situation.

[Pepse]

duncanbojangles, you are forgiven. I seen the humor there, but as you see I am frustrated as heck (where ever the hell, heck is :-) ). You could have left it there it wasn't that bad.

Anyway, onto the problem, again. I grabbed a smallhard drive and loaded MDK 9.1 PP on it and got online and downloaded the nVidia 5336 and the 4496 drivers. No different. I changed video cards, no different. I even used a Quantum Fireball hard drive; my main drive is a Western Digital and as memory served I was told a while back that WD hard drives tend to be flakey with Linux. So, it appears that maybe this Tyan board is causing the problem, because as I mentioned before there is no mention in the book about Unix/Linux compatibility like was mentioned in my old Tyan Trinity 400 board. Anybody else have any further thoughts on this??

Later. Pepse.

[Bubba56]

OK, before ya flame me for not knowing what I am talking about, this is more hardware based reply than Linux, but I am TRYING to help if this is still unresolved.

"Now as for my box. It has a 450 watt power supply; ATX. The mobo it a Tyan Trinity KT-A, socket "A" with an AMD Athlon 1.33 gig CPU"

from Tyan's site:

Trinity KT-A S2390B 1600+ to 1700+

http://www.tyan.com/support/html/amd_athlon_duron.html

and from AMD's site:

http://139.95.253.213/SRVS/CGI-BIN/WEBCGI.EXE/,/?St=3,E=0000000000176542427,K=1739,Sxi=12,Case=obj (1224)

Thoroughbred AND Palomino cores for 1600+ (minimum suggested for board in question) = 1400Mhz.

Now this may or may not be relevent to other problems with video, but if you are running at the stock 1333 as you stated and the core voltage, it may be that there is a voltage issue( Palomino 1.75V, Thoroughbred 1.6V) causing some underlying problem that is affecting the AGP slot. I know this a real long shot, but I have seen far odder things happen that are hardware related. Sorry if this is WAY offbase, but I thought I would take a shot at helping.

Or one other possible site ??

http://www-106.ibm.com/developerworks/library/l-hw2.html

Again, just trying to toss some help, if I am way out there, well just ignore me :(

[Pepse]

Bubba56, in reference to the Tyan site, are you saying that I should switch to an AMD XP1600+ or XP1700+?? Your link took me to Tyan's CPU compatiblity page; is that where I was to go?? And if that is what you are saying, then why do you think an XP chip would cure my issue?? As for your link to AMD's site it gave me an error of "Bad input on previous form". But I will look around their Support site and see if there is anything useful to this problem.

Later. Pepse.

[Digit0]

Originally posted by kozaki
This a known bug, Mandrake team workin on it.

More details :
* http://qa.mandrakesoft.com/show_bug.cgi?id=11060
* http://linuxfr.org/forums/14/3277.html (in french, bottom of the page)

" c'est un bug connu et l'équipe de mandrake travail dessus, mais rencontre beaucoup de difficulté avec le driver qui malheureusement est proprio....voili voilou, donc pour ceux qui comme moi n'ont plus l'accélération 3D :c'est normal et c'est en cours de correction."

ASUS A7N8X

Barton 2500 +

MDK 10 final

Kernel 2.6.3-7

Same problem as poster. Downloading the kernel source now and lets see what happens, really odd though It alll of the sudden started doing it.. had made no changes. Also noticed the missing load glx and dri... very strange sir.

[Digit0]

yep, the source fixed all probs... prob updated kernel without noticing...

[Pepse]

I did load the kernel source and it did not help. Also, you are running Mandrake 10.0, I'm running Mandrake 9.1PP. I did post my problem to Tyan and I did get an answer; in effect: " This board was released during a period that Tyan had no experiance with Linux at any level. If boards previous to this one worked with your distribution then that was just pure "dumb luck". Mandrake is one difficult distribution to deal with unless you have some very good Linux knowledge so this distribution is still not one that Tyan directly deals with. We are getting our feet wet with Red Hat, SuSE, Free BSD, Solaris, SCO Linux, and TurboLinux8. We really do not have the resources to help you in regards to this specific product and Linux. No Linux OS was ever supported for this board so we couldn't confirm nor deny any related issues you might find with this particular board. It could very easily be that this particular VIA chipset is just not happy with the drivers that are being fed to it for the Nvidia card under this distribution. The VIA chipsets in and around this period were very picky and problematic with driver solutions so this alone could account for the continued issue with your Nvidia card when used on this specific motherboard." This was quoted from the reply from Joel, tech6 at Tyan. So, it looks like I will be looking for another Tyan board. Financially I can't go too much farther up; probably the Tyan 2495, or maybe something from Asus.

Later. Pepse.

[kozaki]

I'd like to thank Pepse & araneae for posting what they're trying to solve that prob :D

As for me, everything went back to no bug as soon as i went back to kernel 2.6.3-16mdk.

That sounds logical according to MandrakeSoft bug page (that one i mentioned above).

Pepse > hope you succeed ASAP !