<< | Thread Index | >> ]    [ << | Date Index | >> ]

To: cipe-l,AT,inka,DOT,de
Subject: sk_run_filter problem
From: Calin Dumitrescu <calin,AT,microuser,DOT,ro>
Date: Tue, 30 Nov 2004 22:25:19 +0200

I'm trying to load the cipcb module into a 2.4 kernel with CONFIG_FILTER=y (os=Debian Woody), and I'm runnig into the same "unresolved symbol" problem that some people have reported a while back on this list and elsewhere.

# modprobe -dv cipcb
=============================================
Module cipcb
kname cipcb
objkey cipcb
names: cipcb
mode: NORMAL
Module matching cipcb: /lib/modules/2.4.28-dmaide/misc/cipcb.o
=============================================
/sbin/insmod /lib/modules/2.4.28-dmaide/misc/cipcb.o
Using /lib/modules/2.4.28-dmaide/misc/cipcb.o
Symbol version prefix ''
/lib/modules/2.4.28-dmaide/misc/cipcb.o: unresolved symbol sk_run_filter
/lib/modules/2.4.28-dmaide/misc/cipcb.o: insmod /lib/modules/2.4.28-dmaide/misc/cipcb.o failed
/lib/modules/2.4.28-dmaide/misc/cipcb.o: insmod cipcb failed


The kernel is vanilla 2.4.28 plus a simple unrelated patch that enables UltraDMA mode in the SATA driver. You can get the patch from the following address:

http://lkml.org/lkml/2004/9/18/23

With CONFIG_FILTER disabled, the module loads fine. Unfortunately, I cannot leave it disabled because I need to run dhcpd on the host.

I don't think the problem is isolated into CIPE itself, because the kernel-bundled "pppoe" module shows the same symptoms:

# depmod -ae
depmod: *** Unresolved symbols in /lib/modules/2.4.28-dmaide/kernel/drivers/net/pppoe.o
depmod: sk_run_filter
depmod: *** Unresolved symbols in /lib/modules/2.4.28-dmaide/misc/cipcb.o
depmod: sk_run_filter


---
From the little information I have managed to gather from the web, I have two hypotheses:


1. Problem with Debian's "modutils" package (too old, too new ;) , too different). Most if not all of the people who encountered this were using Debian systems.

If this is true, then *why*? I don't like the prospect of grafting a custom modutils into a stable system. Could it be fixed with a patch in CIPE instead? I simply don't understand why these "unresolved symbol" messages occur.

2. Regression in the 2.4.28 kernel or earlier.

Opinions?
Thanks

--
Calin


<< | Thread Index | >> ]    [ << | Date Index | >> ]