> > I think that you are horribly wrong, because I have at no point used > > initramfs-tools (nor did I mention it). Yes, I am using a (slightly > > larger) initramfs to boot the system, but this is generated by combining > > cdebootstrap with cpio. > hmm is that code free?
Well depends. Actually that code is mostly contained in my shell history. > i've had zero idea that cdebootstrap provides a mount nor does udev, > so i'm still puzzled with your bug report. As I pointed out in my first mail the init script of udev (/etc/init.d/udev) tries to mount --bind and --move stuff. And this fails on rootfs[1], so either the kernel or udev can be blamed and udev is easier to blame. > please define it or bring a testcase up? I have attached a script to construct a testcase. Please create an empty directory, read the script before executing it and then execute it from this empty directory as the user root. It will create a file initramfs.cpio.gz which can be used with qemu to demonstrate the behaviour: qemu -m 512 -kernel /vmlinuz -initrd initramfs.cpio.gz -append "ide0=noprobe ide1=noprobe ide2=noprobe ide3=noprobe ide4=noprobe" /dev/zero Apart from a missing fstab and a missing hostname the system will complain about some mount failed and this happend inside /etc/init.d/udev. After logging in, you can see, that no udevd has been started. Helmut
build.sh
Description: Bourne shell script
