[ILUG] Re: etch and kernel modules

[Niall O Broin]

On 22 Nov 2006, at 11:09, martin f krafft wrote:

>> Yes, it's a little easier to deal with, though I guess most
>> distros now have cramfs in the kernel. At boot time, where do the
>> files in an  initram fs file GO?
>
> As you postulated correctly: a tmpfs.

And as I could have read in the mkinitramfs man page - sorry :-(

>> These are also for hardware features which I simply don't ever intend
>> to use on a server in a data centre but these modules can't be  
>> unloaded.
>> An attempt to rmmod them results in a hanging unkillable process.
>
> Add lines like
>
>   install psmouse /bin/true
>
> to /etc/modprobe.d/local-disabled-modules to blacklist them. Then
> remake the initramfs.

Excellent - that's exactly what I want. However, it brings up another
question. In the interests of experimentation, I tried to alter an
existing initramfs. It's a gzipped cpio, how hard could it be? So I
unzipped it to a directory, removed what I didn't want, made a  
gzipped cpio, and it wouldn't boot :-( Then I RTFS and made the cpio  
correctly in newc format and it booted.

What you have suggested above will remove modules but won't do  
anything about local-scripts. Is there a kosher way of disabling some  
of those, or do I need to do what I have done, extract them by brute  
force?

> I suggest to keep a backup in case the boot does not work

Always a good plan :-)

> I am curious why raid456 and raid10 get loaded. Could you post your
> initramfs somewhere for me to download?

It's from a 2.6.18-1-xen-k7 kernel from backports.org and I've put a
copy of it at http://magicgoeshere.com/files/initrd.img-2.6.18-1-xen-k7

Note that this issue of unwanted raid modules isn't unique to that  
kernel though. I was running 2.6.16 on the box and had the same issue  
with RAID modules (except that raid4, raid5 and raid6 were separate  
modules in that kernel)

>> Well, with Debian you'd have a pretty good chance of getting it
>> out before the next release ;-)
>
> Old habits die hard, huh? If you didn't notice, then we're close to
> releasing etch 18 months after sarge. It took me a year to write the
> book for sarge, which took 3 years to complete. A *lot* changed
> between sarge and etch. It pains me that I cannot release an updated
> edition with etch, but it's just not economically feasible, and
> I did spend most of the time since sarge with severe health
> problems, not unrelated to typing for many hours every day for 12
> months in a row.

Ouchie. Thanks for the book BTW, and extra thanks now that I know how
much it cost you.


Niall