LinuxQuestions.org - kernel panic: no init found

This is what I've learned since posting the previous message. I haven't tested the linux dhcp server with this method yet, but it works with using windows dns.

PXE Boot Linux+Windows Implementation
The goal is to boot a minimized operating system on a diskless server. For this experiment. I have chosen to boot a minimal installation of RedHat Enterprise EL 3 over the network from a windows server.
Install On Linux Server:
Install 'syslinux' - Inlucdes the pxelinux and memdisk utilities.

Install On Linux Or Windows Server:
Install 'dhcpd' - Assigns client an ip address used for the PXE boot
Install 'tftpd' - Trivial ftp server. The client connects to this to download the image files.

Install TFTP Service
Install the tftp service on the windows or linux server of your choice. Create the directories /tftproot (place your images here) and /tftproot/pxelinux.cfg (place your pxelinux config files here). The initrd (initial ram disk image)is a basic root file system needed to boot the os. The initrd image is stored in ram and disposed when the real root file system (usually either C:\ or /dev/hda1) is ready to mount.

Install SYSLINUX On Linux Server
On a Linux box install PXELINUX. Copy the pxelinux.0 file to your tftp server root directory (usually /tftproot) and the ‘default’ file to the pxelinux.cfg sub-directory in /tftproot.

Edit /tftproot/pxelinux.cfg/default
The default pxelinux config file defines your boot menu choices,the kernels and initrd images for each boot configuration.

default 0
prompt 1
timeout 60

label localboot
LOCALBOOT 0

label 0
kernel vmlinuz-2.4.21-15.ELsmp
append ramdisk_size=819200 initrd=rootfs0.img.gz root=/dev/ram0

label 1
kernel vmlinuz-2.4.21-15.ELsmp
append ramdisk_size=819200 initrd=rootfs1.img.gz root=/dev/ram0

NIC Configuration
Configure the network interface card to boot over the network using PXE. This is done in the bios if the nic supports it, otherwise you will need a boot floppy. For diskless servers set the default boot device to the network interface card.

Windows DHCP Scope Options

DHCP Server (LINUX) Setup your /etc/dhcpd.conf

allow booting;
option subnet-mask 255.255.255.0;

group {
next-server 192.168.168.12; #IP address of the tftp server
#PXE boot image filename located in /tftpboot, comes with syslinux distribution.
filename "pxelinux.0"; }

subnet 192.168.168.0 netmask 255.255.255.0 {
default-lease-time 600;
max-lease-time 7200;
option subnet-mask 255.255.255.0;
option broadcast-address 192.168.168.255;
option domain-name-servers 192.168.168.12;

#entry for dhcp server. dhcpd complains without it.
host client-1 {
hardware ethernet 00:02:B3:EE:F8:FC; #mac address of dhcpd server
fixed-address 192.168.168.12; #ip address of dhcp server
filename "/tftpboot/client01";
}

#The client you want to boot over the network
host client-2 {
hardware ethernet 00:09:6B:FF:15:DE;
fixed-address 192.168.168.101;
filename "pxelinux.0";
}
}

Creating the rootfs#.img file:

The image file must fit into memory so start with a minimal installation of linux and instsall only the software you are going to use. The minimal installation of RedHat Enterprise EL3 take about 700MB.

PXELINUX doesn't limit the size of an initrd or memdisk; in practice, they
need to fit into memory uncompressed. If you're using a Linux kernel,
it needs to specifically fit into the "kernel" part of the memory space, which
is usually just under 1 GB. There must be enough memory for the compressed and uncompressed initrd image.

dd if=/dev/zero of=/rootfs1.img bs=1024 count=819200

mkfs –t ext2 –i 1024 –b 1024 –F /rootfs1.img

Copy Files To Your New Image File

mount –o loop /rootfs1.img /rootmount

cp –a /boot /rootmount
cp –a /initrd /rootmount
cp –a /dev /rootmount
cp –a /usr /rootmount
cp –a /lib /rootmount
cp –a /var /rootmount
cp –a /bin /rootmount
cp –a /sbin /rootmount
cp –a /etc /rootmount
cp –a /home /rootmount
cp –a /root /rootmount
mkdir /rootmount/tmp
mkdir /rootmount/proc

Edit /fstab

Change fstab to point to the new root (/dev/ram or /dev/ram0) and remove the swap partition since we’ll be using ram instead of disk anyway:

Old:

LABEL=/ / ext3 defaults 1 1
LABEL=/boot /boot ext3 defaults 1 2
none /dev/pts devpts gid=5,mode=620 0 0
none /proc proc defaults 0 0
none /dev/shm tmpfs defaults 0 0
/dev/hda3 swap swap defaults 0 0

New:

Change so root points to /dev/ram or /dev/ram0 and remove any lines that point to a hard drive

/dev/ram0 / ext2 defaults 1 1
none /dev/pts devpts gid=5,mode=620 0 0
none /proc proc defaults 0 0
none /dev/shm tmpfs defaults 0 0

Compress File System Image

Make sure your image file is smaller than your amount of installed ram. Allow at least 256MB room for other programs to run in ram.

‘umount /rootmount’

‘gzip -c -9 rootfs1.img > rootfs1.img.gz’

The server must have enough ram to fit both the compressed and uncompressed image into memory.

Quote:

Originally posted by klaxyn
I have a similar problem only I'm booting over the network using PXE. The linux client downloads the vmlinuz and initrd image files from the PXE server - downloads rootfs.gz which is the root filesystem image. The clients are diskless so it loads the image into ram and mounts it.

default pxe config:

KERNEL vmlinuz
APPEND initrd=rootfs.gz root=/dev/ram

When the client boots it reports a no init found error and panics. What is the syntax for using init= ?? I've tried setting the value to /sbin/init, rootfs.gz, initrd.img with no luck. I had this problem before after upgrading my kernel on another server but it was just missing the /initrd diretory. I created an emtpy /initrd and it booted from then on, but that did help here. Also some of my other servers have files in /initrd and other don't but they both seem to boot ok. I'm not at work now and the clients are on a priv net so I'll post some logging and config info later when I have access to it :D - Thanks for any ideas you may have.