problem with modules (modutils) and 2.5.31 kernel

I searched for, but could not find references to, a problem I have with 2.5.31 and modutils. (2.5.31 running with Debian distribution on a P3 laptop or on two different P4 workstations, each with a somewhat different kernel configuration -- no further details here, in case this is already a well-known problem.)

insmod, depmod, update-modules all fail with the message:

=> kernel: QM_SYMBOLS: Bad address

That message comes from the routine new_get_kernel_info (in modstat.c, a part of the modutils package 2.4.19-3 under Debian), in response to a call to query_module that is querying the kernel itself:

query_module(NULL, QM_SYMBOLS, syms, bufsize, &ret)

The return code is -1 (error) and errno is the code for EFAULT, indicating a bad address for one of syms or ret.

Is the problem that the modutils version (2.4.19-3 from Debian unstable, in either precompiled or locally built versions) is too old for the 2.5.31 kernel? (I see references to a forthcoming modutils 2.5, but could not locate it.) Or is there a known problem with the kernel itself?

Sorry if this message is a bit naive. I do need the 2.5.31 kernel, because the 2.4.19 (and earlier) kernels cannot correctly assign IRQs and iomem addresses to my Dell machine (problems with handling the 82801DB ICH4 controller resulting in extremely poor IDE disk performance and incorrect USB handling), whereas the 2.5.31 kernel has no problem there.
However, I need to use a couple of modules (they cannot be compiled directly into the kernel) and so need to get around that problem with modutils. Any pointers or patches appreciated!

Bernard Moret

Bernard M.E. Moret	     [email protected]      http://www.cs.unm.edu/~moret
(505) 277-5699 (office)    (505) 277-6927 (FAX)     (505) 277-3112 (department)
Department of Computer Science, University of New Mexico, Albuquerque, NM 87131

Disable CONFIG_PREEMPT and rebuild the kernel.

or, as per a previous thread on the same issue, change line 181 of arch/i386/mm/fault.c from:

if (preempt_count() || !mm)

to:

if (in_interrupt() || !mm)

and rebuild.

Sorry, no time to fool around with diff right now. Hope it was helpful. --hobbs

On Tue Aug 20 2002, [email protected] wrote:

[2.5.31]

Just a loud "me too" here.

# Heinz Diehl, 68259 Mannheim, Germany