On Thu, Jun 10, 2004 at 13:30:20 +0300, Micha Feigin wrote: > On Thu, Jun 10, 2004 at 08:57:56AM +0200, Christophe Lucas wrote: > > l x (whereisit28@yahoo.com) wrote: > > > Hi, > > > > > > I'm working on a module and it cause segmentation > > > fault. After insmod my_mod.o, I can see it's in > > > initialization stage by using lsmod. I tried to rmmod > > > my_mod but got a message saying the module is busy. > > > How can I rmmod the defect module without having to > > > reboot? > > > > Sorry for this quick answer, but reboot and fix your module where it > > crashes. This is the best solution to have a clean remove of a module. > > > > Kernel 2.6 has a force flag for rmmod. AFAIK kernel 2.4 doesn't, but if > you are causing a segmentation fault in kernel space then you probably > have no choice other then rebooting. > > Like Christophe, I don't see any way around rebooting and fixing the > module. > > Two solutions that can help you with debuging are kgdb and kdb > (depending on the availability of serial console) and possibly working > on the module under a uml kernel where you can just kill the kernel > (possible only if its not a device driver). There is one more possibility -- the user-mode-linux project. It can run a linux kernel in a user-land process. You can attach normal gdb to that process and do debugging. Obviously you don't have access to hardware there, but everything else works just as usual. And I believe it's even possible to simulate the hardware, at least a little, so you can debug the OOPSes and only debug interaction with real hardware on live kernel. http://user-mode-linux.sourceforge.net ------------------------------------------------------------------------------- Jan 'Bulb' Hudec <bulb@ucw.cz>
Attachment:
signature.asc
Description: Digital signature
