bloody driver refuses to initialize (rm_init_adapter() failed)

[meem]

folks,

like several others, i am completely unable to get the nvidia driver to work on my rh8 system -- instead, i see this every time in my /var/log/messages:

Apr 8 19:31:45 raining kernel: 0: nvidia: loading NVIDIA Linux x86 nvidia.o Kernel Module 1.0-4349 Thu Mar 27 19:00:02 PST 2003
Apr 8 19:31:45 raining kernel: 0: NVRM: rm_init_adapter failed

this is with a geforce4 420mx card, running on an hp pavilion 734n (which has an athlon 2400+).

pursuant to the other threads on this topic, i have already tried the following:

* setting the "NvAGP" option to 0
* using an older driver (tried 4349, 4191, and 2960 -- all have the same problem, though the error message differs slightly). i've also recompiled it by hand, with similar effect.
* checking the BIOS settings (the "PnP aware OS" option is off; there is no obvious control for the IRQ assignment).
* reducing the amount of physical memory on the machine from 512mb to 256mb.
* checking the module list in my XF86Config file; it appears correct (in particular, GLcore and dri have been removed, and glx has been added).

the failure mode is really obnoxious -- it appears to somehow corrupt the console subsystem such that all of the characters show up as "graphics" characters -- and the only remedy seems to be a reboot.

i work on kernels for a living, and i'd be perfectly happy to investigate further, but it seems that the problem is in nv-kernel.c, for which the source code is not provided -- grrr!

i hope someone has some thoughts on this, because if not i plan to take a hammer to the card and buy something else ... this frustration is certainly not worth a few hundred bucks.

ps. i'll be sure to post pictures of the mutilated card if i do go postal on it :-)

[bwkaz]

Want to post the bottom 40 lines or so of /var/log/messages?

Does it work with version 3123?

Which manufacturer made your nVidia card?

Along the same lines, who made your motherboard? Which chipset is it using?

[meem]

hey there,

my /var/log/messages has no other relevant messages besides what i posted.

i have not tried 3123 -- i will but honestly i think there's a better chance of me getting hit by lightning than that working.

i'm not sure where to get the manufacturer code from (the card came with the machine and has no obvious markings) -- lspci reports:

01:00.0 VGA compatible controller: nVidia Corporation NV17 [GeForce4 MX420] (rev a3)

as for the motherboard, it appears to be an asus with a via chipset -- here's some (slightly dated) information:

http://www.hp.com/cposupport/personal_computing/support_doc/bph07422.html

i cannot find any more recent information on hp's website :-(

[bwkaz]

Hmm... I wish I knew what conditions caused the "rm_init_adapter failed" message. Unfortunately, I don't have the source either. :(

Anyone know?

Nothing seems to be wrong with the chipset or anything, it's a KM-133A, on (like you said) an Asus A7V-VM.

[Andy Mecham]

You should be able to get more information with modprobe nvidia cur_debuglevel=0 Can you post the output of /var/log/messages & XFree86.0.log with that option?

--andy

[bwkaz]

Andy -- is that a new option? Would it help to isolate issues when people get the "could not initialize kernel driver" message, as well?

[Andy Mecham]

bwkaz: In some circumstances, it might; most of the "failed to initialize" messages are install problems, so it might not do much good for those.

--andy

[meem]

andy,

i tried cur_debuglevel=0, but it doesn't seem to work:

# modprobe nvidia cur_debuglevel=0
Warning: loading /lib/modules/2.4.18-27.8.0/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
Warning: ignoring cur_debuglevel=0, no such parameter in this module
Module nvidia loaded, with warnings

... i do see the symbol in the driver though:

# nm /lib/modules/2.4.18-27.8.0/kernel/drivers/video/nvidia.o | grep debug
000006a0 D cur_debuglevel

what am i doing wrong?

[Andy Mecham]

Eh, might be easier just to edit os-interface.c and change the value manually.

--andy