Nvidia drivers incompatible with Ndiswrapper

[bjrosen]

I've been running the FC3 with a custom 2.6.11.5 kernel and the 7167 drivers, everything was fine. I installed ndiswrapper and then rebooted and the Nvidia driver wouldn't load. I tried reinstalling the Nvidia drivers, didn't help. I also did a fresh make install and make modules_install and then reinstalling the Nvidia drivers, still wouldn't work. I was able to go back to an older 2.6.10 kernel (on which I'd never installed ndiswrapper) and after reinstalling the Nvidia drivers I was able to load them.
Why are the Nvidia drivers sensitive to ndiswrapper?, they are completely on related subs-systems.

[Lithorus]

It might just aswell be the other way around, ndiswrapper incomptaible with the nvidia drivers. Perhaps you could post the nvidia bugreport which is supposed to help people who want to help, so we have somewhere to start?

Also things like how you installed the ndiswrapper would help, since it oftens involves kernel source patches etc.

[bjrosen]

There were no kernel patches, all I did was follow the standard procedure for installing ndiswrapper from source, i.e. make install,

http://ndiswrapper.sourceforge.net/p...p/Installation

I haven't been able to get the Nvidia drivers working with 2.6.11.5 on that machine since I did the ndiswrapper install. I did a make uninstall on ndiswrapper, rebuilt 2.6.11.5 and reinstalled it, cleaned out the 2.6.11.5 directories and did another reinstall. Nothing seems to work. The Nvidia drivers still work with the old 2.6.10 kernel and they worked with 2.6.11.5 before I did the ndiswrapper install. I haven't a clue as to what ndiswrapper did and why it only effects 2.6.11.5.

[Lithorus]

Yet STILL a nvidia-installer log would help a great deal...
See http://www.nvnews.net/vbulletin/showthread.php?t=46678 for intructions

[lightsout]

They are not inherenitly incompatabile - I'm using ndiswrapper 1.1 with the Nvidia 7167 drivers (graphics & 1.031 nvsound drivers too). I've been using them for several previous versions without issue.

What are the error messages you get? When you do "modprobe ndiswrapper" or "modprobe nvidia"!

[Adam Benjamin]

I am seeing an "incompatibility" between ndiswrappers and my nvidia driver. The problem is I have a wireless card (hence the ndiswrappers) and (of course) an nvidia card. If the both are active, then I see the following in the message log:

Jul 23 18:10:42 panda kernel: irq 177: nobody cared (try booting with the "irqpoll" option)
Jul 23 18:10:42 panda kernel: <c044af26> __report_bad_irq+0x36/0x7d <c044b0f4> note_interrupt+0x187/0x1b7
Jul 23 18:10:42 panda kernel: <c044aa1e> handle_IRQ_event+0x23/0x4c <c044aafa> __do_IRQ+0xb3/0xe5
Jul 23 18:10:42 panda kernel: <c0406486> do_IRQ+0x67/0x84
Jul 23 18:10:42 panda kernel: =======================
Jul 23 18:10:42 panda kernel: <c040487e> common_interrupt+0x1a/0x20
Jul 23 18:10:42 panda kernel: handlers:
Jul 23 18:10:42 panda kernel: [<f8e9ff4b>] (ndis_isr_shared+0x0/0x86 [ndiswrapper])
Jul 23 18:10:42 panda kernel: Disabling IRQ #177
Jul 23 18:12:10 panda kernel: ndiswrapper: device wlan0 removed

Needless to say, now that I'm aware of the conflict, I can run either the ndiswarpper or the nvidia drivers independantly just fine. The thing is I'd love to have both co-existing nicely. I don't have sufficient knowledge to diagnose IRQ related issues without guidance, although I'm an experienced linux admin so I should be able to handle any reasonable suggestions about things to try.

I'm running on Fedora Core 5, with the Livna provided packages ndiswrapper v1.18, release 1.lvn5, as well as the nvidia bundling of version 8762 of the nvidia drivers. The nvidia-bug-report.log is attached in a zipped form.

Any help would be, of course, much appreciated.

Thanks,

Adam Benjamin

[netllama]

This looks more like a BIOS problem. Does the error change depending on whether ndiswrapper or the nvidia kernel module is loaded first?

Have you verified that you're using the latest BIOS?

1.18 is not the latest ndiswrapper version (1.22 was released this morning).

Thanks,
-Lonni

[Adam Benjamin]

Thanks for your fast response...

The error doesn't get invoked right away. Once both devices are up and running, I'll start to do something over the wireless and all of a sudden I get the Disabling IRQ message. It's the same IRQ every time. (So at least it's consistent. :/ ) I didn't see any change in behavior depending on which one was loaded first.

I took your recommendation of updating the bios (hadn't realized there would be one... I've had the laptop for about three weeks now.) No change there.

I see the ndiswrapper update you're referring to. I will try it out just as soon as it is available (as an RPM) via the Livna site. (Saves me having to package it up myself.) So unless I forget, I'll report back on my success in a couple of days.

Just to remind you, both the Nvidia graphics drivers and the ndiswrappers work just fine independent of each other. It's simply when they try and co-exist the the problem crops up. Fortunately for me it's unlikely I will need wireless and OpenGL at the same time, but it would be nice not to have to fuss with it. And I'm in no way suggesting that it's the Nvidia driver that's at fault... but until I know more, I can say they're not playing well together.

Thanks for any help you can offer,

Adam

[netllama]

The issue here is that both drivers are using the same IRQ. Neither driver can control which IRQ it is assigned. That is handled by the kernel to some limited extent, and the BIOS. The problem you're reporting isn't a driver bug, its a BIOS bug, I'm afraid.

[Adam Benjamin]

Just to follow-up...

I finally had to roll my own ndiswrapper and nvidia rpms (the folks at Livna seem to be in flux.) Am now running under the latest kernel (2.6.17-1.2174_FC5smp,) ndiswrappers (v1.23) and nvidia drivers (8762.) Still no dice.

So, before I start seeking support from Dell directly, is there any liklihood that my problem is in any way driver or kernel related? Phrased another way - even if the drivers are not at fault, is there anything I can do at the kernel level - or am I really stuck with lobbying for a bios fix?

I suppose part of my thoughts are along the lines: it works under Windows XP... so why can't it work happily under Linux? My apologies for not knowing more about this stuff already... I know lots of theory about how interrupts work, but not to the practical level of how Linux implements them, and hence what controls I do or don't have, etc.

Thanks for your assistance,

Adam

[netllama]

This can't be fixed from within a driver, as neither driver controls which IRQ it is assigned.

-Lonni

[pgs]

Originally Posted by Adam Benjamin

Just to follow-up...

So, before I start seeking support from Dell directly, is there any liklihood that my problem is in any way driver or kernel related? Phrased another way - even if the drivers are not at fault, is there anything I can do at the kernel level - or am I really stuck with lobbying for a bios fix?

Adam

You may try to boot with "pci=routeirq" option.
This could re-route the irq assignements differently and change
the situation (note: change does not mean improve).

If it, by chance, improves the things, please report to the proper
people on the linux kernel side.

bye