LFS on OLD machine

Ken Moffat ken at kenmoffat.uklinux.net
Sun Feb 15 05:17:36 PST 2004


On Sat, 14 Feb 2004 tbdewey at earthlink.net wrote:

> Well, here I am again so soon.  I'm going on with BLFS, but I'm
> posting here because it seems that LFS is where the problem lies.
> While trying to make the rescue disk in Chapter 3 of BLFS 5.0, I
> have the following problem:
>
> losetup /dev/loop1 /tmp/rfloppy
> gives, "/dev/loop1: No such device or address."
>
> For good measure
> mount -o loop /tmp/rfloppy /mnt/loop1
> gives, "Could not find any loop device. Maybe this kernel does not
> know about the loop device? (If so, recompile or 'modprobe loop.)"
>
> modprobe loop
> gives, "Can't locate module loop."
>
> I've compiled the kernel three times (with GCC-2.95.3) and the last
> two times I am certain I had selected support for loop devices,
> ramdisk and initial ramdisk in the menu.config.  I've recompiled and
> reinstalled modutils. I've read the documentation in the kernel
> directory and googled a few hundred pages about this and no one
> seems to report much trouble with it.
>
Tom,

 a selection of thoughts.  I'm not up to speed on the rescue disk in
BLFS, but I'm used to making mistakes in my builds :)

 Did you select it as a module or built in ?  I imagine this stuff needs
to be built in, at least in the recovery kernel.  Assuming you've kept
the .config (perhaps copied to another directory ad renamed to e.g.
config-2.4.24-attempt3 to preserve it from `make mrproper'), have a look
to see what it says.  Sometimes (depends on what options you change) you
have to copy out the .config after make menuconfig, make mrproper, copy
back the config to .config, make oldconfig, then make dep etc.

 Maybe you've built the kernel several times without changing the
EXTRAVERSION in the Makefile, and perhaps forgot to copy one of the
bzImages, or ended up with modules that don't belong to the kernel of
the same name.

> One question keeps nagging at me, and that is, the estimated
> build time for installing Linux-2.4.22 in Chapter 8 of LFS 5.0 is 4.2
> SBU. On this machine one SBU equals about an hour. But it has
> never taken more than an hour to compile the kernel.
>

 SBUs are to be treated with caution.  For me, kernels seem to be
between 1.5 and 4.5 SBUs, depending on the machine, which compiler and
binutils you use to create the static binutils, what options you put
into the kernel.

> The messages that pass on the screen during "make modules"
> and "make modules install" do not mention loop.
>
> There are 8 loop devices in /dev.
>
> I have the files:
> /usr/include/linux/loop.h
> /usr/src/linux-2.4.22/include/linux/loop.h
> /usr/src/linux-2.4.22/include/linux/modules/loop.stamp
> /usr/src/linux-2.4.22/include/linux/modules/loop.ver
> /usr/src/linux-2.4.22/include/config/blk/device/loop.h
> /usr/src/linux-2.4.22/drivers/block/loop.c
> /usr/src/linux-2.4.22/drivers/block/loop.o
>
> Insmod /usr/src/linux-2.4.22/drivers/block  loop.o
> gives, "Couldn't find the kernel version the module was compiled
> for."
>

 No, you don't modprobe or insmod from the kernel source tree!  At the
end of make modules_install the Makefile runs depmod to sort these
things out.  Modules get installed below
/lib/modules/linux-2.4.${VERSION}${EXTRAVERSION}

 - if you're using modules, you may need to put the pathnames in
/etc/modules.conf (symptom: modprobe cannot find the module, but works
if you give it the full path as well as the filename).

> I feel that I'm not in Kansas anymore.
>
> Again, thanks for everything.
>
> tom
>
> P.S. this is being written with vi on the LFS 5.0 system, so
> something works, anyway.
>

Ken
-- 
Brighton tops UK Jedi league
http://www.theregister.co.uk/content/28/35186.html




More information about the lfs-support mailing list