On Mon, 2005-12-19 at 07:31 -0800, Steven Ringwald wrote: > Bob Chiodini wrote: > Sorry for the confusion. I have been trying with two different drives. > One is a USB key-fob thing, which comes up as a SCSI drive (sda), and > the other is a CF card plugged into an IDE convertor (this comes up as > hda). When I couldn't get the USB fob to work, I tried the CF card, > since the IDE modules should already be in the kernel, and therefore > should be slightly simpler. > > >Using the linuxrc script above, what happens if you cd /sysroot, then > >run pivot_root . initrd? /sysroot/initrd must exist. > > > > > > Yes. It does. What happens is that it returns EINVAL, because it is a > CPIO-style image, not a ramfs. Have you tried building it into a compressed ramfs? Does the 2.6 kernel support ramfs? Might be an option. > > >When linuxrc exits the kernel should run init. Is busybox compiled for > >x86_64 and statically linked? > > > > > Yes. I went so far as to go to the source and get the svn copy/build it. > Ldd reports "not a dynamic executable" and file says ELF 64-bit LSB > executable. > That seemed to work a little better, though pivot_root/switch_root > threw up on my shoes. I was able to get it to chroot into the target > filesystem, though I cannot seem to run init. Okay that explains the switchroot. If you use nash what error does switchroot produce? I also noticed that switch_root, from busybox has an open bug: http://bugs.busybox.net/view.php?id=543 Further searching led to this: http://lkml.org/lkml/2005/11/15/384 Which might be relevant. From the email in this thread it sounds like chroot . is what you need. What happens if you do something like this: --- at the end of linuxrc echo 0x0100 > /proc/sys/kernel/real-root-dev echo Starting system cd $SYSROOT chroot . cd / Letting linuxrc just exit. Bob...
