2.6.5 Problems:No module symbols loaded - kernel modules not enabled
 

This question has been asked before but I have yet to see it resolved. I did a clean fresh install of Woody 2.4.18 and then dist-upgrade to sid. I installed module-init-tools and kernel-package.

I first tried compiling 2.6.5 from source (from Kernel.org) and had the problem...

I then ran apt-get install kernel-source-2.6.5 to compile the kernel with the Debian patches...

Using menuconfig, I compiled the kernel with all of the options I required compiled in the kernel. I compiled the kernel the "Debian" way with make-kpkg kernel-image and then installed it with dpkg -i newkernelname.deb This all went fine and the /lib/modules/2.6.5 folder with modules was created (Obviously, some things are still compiled in as modules...). I modified Lilo.conf and ran lilo, rebooted and things "appeared" OK.

It just happens that during one of my reboots, there was a fschk forced on one of the partitions and I could see a number of FATAL's trying to load modules during boot (which do not appear in dmesg or kern.log, only the message in the title above)! So there are modules and they are not being loaded!

If I run lsmod while booted into 2.6.5, it shows nothing...
If I run lsmod after booting 2.4.18, it shows a dozen or more modules loading.

It's almost as if the "new" module-init-tools cannot find the path to the new modules?

Does anyone have any thoughts or ideas of what I might try?

the reason you are seeing fatals is because your /etc/rc.d files are telling your system during startup to load certain modules.

If you menuconfiged it, there is a good chance that some things you had as modules, are no longer such, so it depends entirely on what the mesage says.

So there are several causes:
1 - the modules aren't there because they are part of the kernel
2 - the modules aren't there because you didn't make them in any way shape of form *ie the aren't in the kernel either*
3 - software can't start because ... for instance netfilter .... you need to update this post kernel update (i imagine there are a few things that have this need since there were allegedly changes made to the user space)

there are other reasons but i'm throwing a blank
oh yeah
4 - you might have a duplicate entry in your rc.d file which tells it to load a module that's already loaded, for instance some sound modules load other modules automatically (especially if the kernel was told to auto load modules as required)


OK to fix this ...
personally i wouldn't bother if its a home machine, because you know if something isn't working because ... well ... it's not working

IF it is a work server however things are more dire, since you don't want to customers to be the first person to let you know somethings not up

check dmesg | less after boot ... as to what those fatals were, the update the scripts or packages (or both to suit)

i only used once make-kpkg but i think there it might be like
make-kpkg kernel-image modules-image

The FATALS do not show up in dmesg or kern.log so I can't say for sure what is attempting to load... Everything seems to work, however. What does concern me is that the make-kpkg DOES create a folder /lib/modules/2.6.5 and it does contain entrys as wells as a couple of folders beneath it. When I compiled the kernel, I "think" I compiled everyting that I needed into the kernel and not as modules. When I run menuconfig, I have all "*'s" where I require things and no "M's" appear. It may be that the tree is created as a matter of course???

When I boot to the "old" 2.4.18 kernel, there are about a dozen modules loaded and this may be what is trying to load with the "new" 2.6.5 kernel and it can't find them in the new path??? I can look in the /etc/rc.d stuff to see. I can remove them once I find that everything works in 2.6.5 and that it is stable. I have another machine with 2.6.3 on it, and I recall having to remove some stuff in /etc/rc.d there also...

Thanks for the ideas and I will look into it more. I have about two more reboots before an fschk is run on one of my partitions so I will then have the time to see exactly what the FATALs are (I don't know why they do not show up in dmesg or kern.log)

In the overall process of compiling the kernel, at some stage the module structure is generated, that is the directory structure. You also tell it to make the modules and install them. (i can't claim i'm sure WHEN the directory is created, though i would assume on make modules)

Either way, this means that the directories are there, but only those modules that need to be there are there, those that are deemed unnecessary are those that were NOT indicated to be made. The directory structure should still exist.

One of the FATAL messages i get on my home machine is that some sound module couldn't get loaded because it is compiled into to the kernel. The scripts are not varied because of the module install, but then ... some scripts do vary some rc.d scripts ... it is a little bit of a mystery to me to be quite honest.

But if it all works, then all is well ... :)

Re:2.6.5 Problems:No module symbols loaded - kernel modules not enabled
 

Thanks again for the information. I'm only able to work on this pc "occasionally" right now so I will probably try to block out an extended period of time to see if the thing stays stable... If so I can make plans to modify some of the offending scripts. (I think I have two more reboots before an fschk is forced on one of my partitions, so I can then "see" some of the FATALS to see if they are something to worry about!) Everything really important "seems" to be working fine, however!

I still have some other "issues" (the pc won't shutdown properly... it hangs at "acpi_power_off called", I have two host adapters and if I leave my SCSI Zip drive connected it "takes" sda and I can't boot... I can't seem to get the scsihosts boot parameter to work with 2.6.5, I have "two" Tuxes at boot (SMP, 2cpu's?) but a large block of white space to the right of the second Tux... not an issue, but when I "clear" the screen in console mode there is a "phantom" block at the top of the screen that blocks the top of dmesg or a "less" output... just details!!!!)

Thanks! If you have any other thoughts or thoughts on my other issues, I would appreciate it!

You may have read this before but when I get in a bind I follow this steps and it always works.
http://newbiedoc.sourceforge.net/sys...ALL-KERNEL-PKG

the power down thing is always a concern ... if you're lucky it worked before you upgraded, in which case if you grab the original (source) package there should be the configuration file, which is plain text, so look at it (commonly .config) and use those settings in the 2.6 kernel.

if that doesn't work ... lots of experimenting, i find minimalist approach to work best, but ... that's not always the case, depending on your MoBo.

I'll take a look at the original configuration file and compare ACPI and APM and see what's up.

Thanks!

If I may interject, since you have most of the former modules compiled into the kernel, then it stands to reason that you should edit /etc/modules accordingly. You need to comment out the modules that were compiled into the 2.6 kernel. Then run update-modules in a console and then finally reboot linux. The errors should drop dramatically since it won't be looking for phantom(duplicate) modules that were already compiled into the kernel.

I'll check into that. I do think that it is actually "phantom" modules (and nothing to worry about as others have posted). One of the "FATALS" that I can actually read on boot is the AC97 driver for my onboard sound... yet I can read in dmesg where the driver has been loaded by the kernel...

Thanks!

ok kewl, i don't have that scripts, and my modules.conf file is empty (at least on my work machine). I'll have to check what it does at home, but ... i've never had info in that file except when i added sensors. And NVidia does stuff to that file ... never the less, if it works, it's all good.