On Wed, Jun 25, 2003 at 07:31:40AM -0700, Pat LaVarre wrote: > > From: Jan Hudec <bulb () ucw ! cz> > > ... > > cramfs images ... > > should be possible to mount them on loopback. > > Elaborate please? I see: Ouch... I seems to be kind of mess there and I don't have an initrd image handy (and even that - it might differ in debian and redhat). Kernel doc says, that initrd is gzip compressed image of any filesystem. mkinitrd doc says, that it creates cramfs images and kernel does not support them, but Debian kernels do. Mainline kernels definitely do support cramfs. > $ sudo mount /boot/initrd-2.4.20-pel.img /mnt/loop0 -o > loop=/dev/loop0 Using -o loop suffices - it will pick the first loop device it can. > > From: Jan Hudec <bulb () ucw ! cz> > > ... > > The first says to give correct device - eg. > > /dev/hda2 - that does not happen here. > > Agreed. > > > > another is to add a patch (posted to l-k ... > > > years ago) which added kernel support for this, > > > and > > > > The second says there should be a patch - it is > > not in the mainline kernel (it would work > > otherwise) and it's not in RedHat kernel only > > (mainline kernel would not work with RedHat setup). > > Lost me sorry. > > I can accept on your authority that I don't need a > patch, but I don't yet see how you discovered that. > > After considerable study, I tentatively conclude, here > we mean to be arguing specifically from the evidence > of having made a kernel from kernel.org with a .config > from Red Hat and then seeing that kernel boot with > root=LABEL=/ no worries, for example: > > title Red Hat Linux (2.4.21-rh-pel) > root (hd0,0) > kernel /vmlinuz-2.4.21-rh-pel ro root=LABEL=/ > initrd /initrd-2.4.21-rh-pel.img > > Since the defconfig fails and the Red Hat config > works, I figure there is some minimal change to the > .config that matters. As yet I have no idea what we > think that minimal change might be. initrd, cramfs? (some other fs?) By the way, as you speak about a "defconfig", have you carefuly inspected each and every option if you need it or not? > > From: Jan Hudec <bulb () ucw ! cz> > > ... > > make install has only one way - via the > > /sbin/installkernel script. > > Great hint, thanks. Now in 2.5.73 I notice: > > $ make help | tr -s '[:blank:]' ' ' > ... > Architecture specific targets (i386): > ... > install - Install kernel using > (your) ~/bin/installkernel or > (distribution) /sbin/installkernel or > install to $(INSTALL_PATH) and run lilo > ... > $ Yes, it is that way a loooong time, just there was no help target. > Back in 2.4.20 I see: > > $ grep mkinitrd /sbin/installkernel > /sbin/new-kernel-pkg -v --mkinitrd --depmod --install > $KERNEL_VERSION > $ ... well, a RedHat's way of installing kernel - easy to use as long as you stick to the original config, but you are on your own if you don't. > Possibly that -v in /sbin/installkernel I added myself > in a past life. Because of that -v, I see: Verbosity options don't matter ;-) Out of curiosity you might want to study what the new-kernel-pkg actualy does, but you don't need to - > # make install > ... > creating initrd /boot/initrd-2.4.20-pel.img using > 2.4.20-pel > ... > # > > I think that "creating ..." claim from > /sbin/new-kernel-pkg says I then executed the command: > > /sbin/mkinitrd -f /boot/initrd-2.4.20-pel.img > 2.4.20-pel > > Adding -v to that /sbin/mkinitrd invocation yields: > > Using modules: > Using loopback device /dev/loop0 > /sbin/nash -> /tmp/initrd.CKgyYL/bin/nash > /sbin/insmod.static -> /tmp/initrd.CKgyYL/bin/insmod > > To me this also suggested `man nash`. > > Pat LaVarre ------------------------------------------------------------------------------- Jan 'Bulb' Hudec <bulb@ucw.cz> -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
