Re: [Etherboot-users] /sbin/init hangs on boot

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Hi!
Thanks for replying!

When I say PXE I mean that I am using newer motherboards that come with
netboot built in the BIOS; I say Etherboot I mean I older motherboards
with an EPROM extending the BIOS features (I designed an ISA card that
works as a BIOS extension).
My card is working fine with LTSP and Etherboot (it starts the NIC,
downloads the kernel and load the OS) but when I tried to netboot the
way I said before, even using Etherboot from floppy, the thinclient
starts the NIC, download de ELF (or NBI), run the "init=/rd/bin/init0"
script and hangs in the 2 last lines of init0:
+++++
exec /sbin/init '#' <<<<<<<< 
exec /etc/rc.sysinit
+++++

My thinclient is an old Pentium.

When I use a newer PC, an HP Compaq whose BIOS has PXE, my
implementation works fine.

It follows my config files: RAMDISK, pxelinux.cfg/default, kernel
.config, my RAMDISK tree and my mkelf-linux command line.

Regards,
Tiago Alves  

To create my RAMDISK I use "makerd":

-------------------------
#!/bin/sh
# makerd --- creates an initrd ramdisk file from an existing directory tree.
# Copyright  2001 Jamie Zawinski <jw...@dn...>
#
# Permission to use, copy, modify, distribute, and sell this software
and its
# documentation for any purpose is hereby granted without fee, provided that
# the above copyright notice appear in all copies and that both that
# copyright notice and this permission notice appear in supporting
# documentation.  No representations are made about the suitability of this
# software for any purpose.  It is provided "as is" without express or 
# implied warranty.
#
# Created by jwz, 23-Dec-2000

verbose=
dir=
target=

# Increase this if you get "No space left on device" errors.
IMAGESIZE=4096
#2048

make_ramdisk() {

  IMAGE=$target.tmp
  dd if=/dev/zero of=$IMAGE bs=1k count=$IMAGESIZE 2> /dev/null

  for devnum in 0 1 2 3 4 5 6 7 8; do
    if losetup /dev/loop$devnum $IMAGE 2>/dev/null ; then break; fi
  done

  if [ "$devnum" = "8" ]; then
    rm -rf $IMAGE
    echo "All of your loopback devices are in use!" >&2
    exit 1
  fi

  LODEV=/dev/loop$devnum

  if [ -n "$verbose" ]; then
    echo "Using loopback device $LODEV"
  fi

  echo y | mke2fs $LODEV $IMAGESIZE >/dev/null 2>/dev/null

  MNTPOINT=$target.mnt
  rm -rf $MNTPOINT
  mkdir $MNTPOINT

  mount -t ext2 $LODEV $MNTPOINT || {
    echo "Can't get a loopback device"
    exit 1
  }

  rmdir $MNTPOINT/lost+found

  cpio_v=""
  if [ -n "$verbose" ]; then
    echo "Copying $dir..."
    dash_v="-v"
  fi

  cwd=`pwd`
  ( cd $dir; find . -depth -print | 
    cpio $dash_v --quiet -pdm $cwd/$MNTPOINT ) \
    2>&1 | sed "s@$cwd/$MNTPOINT/\./@/@g" | sed 's@^@    @'

  umount $MNTPOINT
  losetup -d $LODEV

  ( gzip $dash_v -9 < $IMAGE > $target ) 2>&1 | sed 's@^@Compression:@'
  rmdir $MNTPOINT
  rm -rf $IMAGE

  if [ -n "$verbose" ]; then
    echo "Wrote $target:"
    ls -ldGF $target
  fi
}

usage() {
  echo "usage: `basename $0` [-v] input-directory output-ramdisk-file"
  exit 1
}

main() {
  while [ $# -gt 0 ]; do
    case $1 in
     -v)
       verbose=1
       ;;
     *)
       if [ -z "$dir" ]; then
         dir=$1
       elif [ -z "$target" ]; then
         target=$1
       else
         usage
       fi
       ;;
    esac
    shift
  done

  if [ -z "$dir" -o -z "$target" ]; then
    usage
  fi

  make_ramdisk
}

main "$@"
exit 0
-------------------------

My PXE config is (this works fine):

pxelinux.cfg/default:
------------------------
prompt 0
label linux
  kernel kiosk.vmlinuz
  append initrd=kiosk2.rd root=/dev/ram0 init=/rd/bin/init0 ID=01
SERVER=192.168.100.1 vga=795 ip=dhcp
------------------------

My mkelf-linux command line is:

------------------------
mkelf-linux --ip=dhcp --append="root=/dev/ram0 init=/rd/bin/init0 rw
ID=02 SERVER=192.168.100.1" kiosk.vmlinuz21 kiosk3.rd > kiosk.vmlinuz2
------------------------

My init0 is (where the boot hangs, look at the 2 last lines):

---------------------
#!/rd/bin/sh
#
# /sbin/init for diskless workstations
# Copyright (c) 2000, 2003, 2004 by Jamie Zawinski <jw...@dn...>
#
# Permission to use, copy, modify, distribute, and sell this software
and its
# documentation for any purpose is hereby granted without fee, provided that
# the above copyright notice appear in all copies and that both that
# copyright notice and this permission notice appear in supporting
# documentation.  No representations are made about the suitability of this
# software for any purpose.  It is provided "as is" without express or 
# implied warranty.
#
# Created: 23-Dec-00, Jamie Zawinski <jw...@dn...>
#
# This file is run by the kernel from a ramdisk file system.  The kernel
# should invoke this as the first thing.  It mounts stuff, then runs init.
#
# The root file system on this ramdisk will remain the kiosk's root
# directory (read-only.)  This script NFS mounts a few directories to
# fill in the dangling symlinks, then invokes init to finish booting.
#
# expects SERVER=<ip> and ID=<n> to be passed on the kernel command line.
# SERVER is IP of the NFS server to use; ID is the name of the sub-directory
# to mount as this particular kiosk's writable (non-shared) file system.

# The programs in /ro/bin are symlinks to busybox, an all-in-one shell
# program.

# First, make sure the boot ramdisk (root file system) is writable.
# (/proc needs to be mounted first, in order for this to work)

/rd/bin/mount -t proc proc /proc
/rd/bin/mount -t rootfs rootfs -o remount,rw /

# first mount various vital directories from the NFS server.
# /ro is the root directory shared between the server and clients..
# /ro2 is the root directory shared between all clients (the pieces
# where the clients differ from the server.)
# /rw is the per-client writable directory; each client has its own.
#

/rd/bin/mount -t nfs     -o nolock $SERVER:/opt/kiosk        /ro
/rd/bin/mount -t nfs    -o nolock $SERVER:/home/guest/$ID	/rw

# We needed a handful of devices in /dev even before mounting anything.
# Now that the mounts are up, we can use the server's fully-populated dev.
# but devices are still open in our /dev, so just rename it out of the
# way, and then turn /dev into a symlink to the server's version.
#

 /rd/bin/mv /dev rd/dev
 /rd/bin/mkdir /dev
 /rd/bin/mount -t nfs    -o nolock $SERVER:/home/guest/$ID/dev /dev

# Now invoke "init" and complete the boot-up with server-loaded files.
# The argument "#" is what the kernel passes as the runlevel to tell
# init to read the "initdefault" entry in /etc/inittab.
#

exec /sbin/init '#' # <<<<<<<<<<<<<< HERE IT HANGS :(
exec /etc/rc.sysinit
---------------------

My rd_root look like this:

----------------------------
total 36
lrwxrwxrwx  1 root root    7 Nov  7 13:40 bin -> /ro/bin
lrwxrwxrwx  1 root root    8 Nov  7 13:40 boot -> /ro/boot
drwxr-xr-x  4 root root 4096 Nov  3 13:34 dev
lrwxrwxrwx  1 root root    8 Nov  7 13:40 etc -> /rw/etc/
-rwxr-xr-x  1 root root 5769 Oct 28 14:18 fastboot
lrwxrwxrwx  1 root root    9 Nov  7 13:40 home -> /rw/home/
lrwxrwxrwx  1 root root    7 Nov  7 13:40 lib -> /ro/lib
lrwxrwxrwx  1 root root   10 Nov  7 13:40 media -> /rw/media/
drwxr-xr-x  2 root root 4096 Nov 11 10:41 mnt
lrwxrwxrwx  1 root root    8 Nov  7 13:40 opt -> /rw/opt/
drwxr-xr-x  2 root root 4096 Oct 20 19:37 proc
drwxr-xr-x  3 root root 4096 Oct 20 19:37 rd
drwxr-xr-x  2 root root 4096 Oct 20 19:37 ro
lrwxrwxrwx  1 root root    8 Nov  7 13:40 root -> /rw/root
drwxr-xr-x  2 root root 4096 Oct 20 19:37 rw
lrwxrwxrwx  1 root root    8 Nov  7 13:40 sbin -> /ro/sbin
drwxr-xr-x  2 root root 4096 Oct 28 18:18 sys
lrwxrwxrwx  1 root root    8 Nov  7 13:40 tmp -> /rw/tmp/
lrwxrwxrwx  1 root root    7 Nov  7 13:40 usr -> /ro/usr
lrwxrwxrwx  1 root root    7 Nov  7 13:40 var -> /rw/var
----------------------------

> Tiago Alves da Fonseca wrote:
> > Hi!
> > I am working this last month with thinclients. It worked fine with PXE
> > and Etherboot running LTSP. Based on a solution described in DNA Lounge
> > (www.dnalounge.com -> kiosk project), I tried to run thinclients that
> > boot from the network and execute all aplications locally. When I tried
> > to load all the software via PXE, the machine booted correctly and all
> 
> define "via PXE", cuz to me that is the same as "using Etherboot".
> 
> > aplications runned fine but when I tried to implement the same solution
> > using Etherboot (using mkelf-linux and the same ramdisk) the computer
> > entered in a loop when the init script started the /sbin/init.
> 
> What is the first and last line of the loop?  (as in, how far did it
get and where 
> did it loop back to?)
> 
> Post your pxe config file.  mainly the kernel and append.
> 
> Here is what I use for a knoppix based TS:
> label knoppix37
>          kernel knoppix/vmlinuz
>          APPEND nfsdir=192.168.1.7:/mnt/knoppix37 nodhcp lang=us 
> ramdisk_size=100000 init=/etc/init apm=power-off nomce vga=791 
> initrd=knoppix/miniroot.gz BOOT_IMAGE=knoppix noacpi
> 
> wag#1: you need a bigger ramdisk_size=100000
> 
> ^Carl K
> 
> 

-- 
NeoMail - Webmail that doesn't suck... as much.
http://www.neomail.org

> Tiago Alves da Fonseca wrote:
> > Hi!
> > I am working this last month with thinclients. It worked fine with PXE
> > and Etherboot running LTSP. Based on a solution described in DNA Lounge
> > (www.dnalounge.com -> kiosk project), I tried to run thinclients that
> > boot from the network and execute all aplications locally. When I tried
> > to load all the software via PXE, the machine booted correctly and all
> 
> define "via PXE", cuz to me that is the same as "using Etherboot".
> 
> > aplications runned fine but when I tried to implement the same solution
> > using Etherboot (using mkelf-linux and the same ramdisk) the computer
> > entered in a loop when the init script started the /sbin/init.
> 
> What is the first and last line of the loop?  (as in, how far did it
get and where 
> did it loop back to?)
> 
> Post your pxe config file.  mainly the kernel and append.
> 
> Here is what I use for a knoppix based TS:
> label knoppix37
>          kernel knoppix/vmlinuz
>          APPEND nfsdir=192.168.1.7:/mnt/knoppix37 nodhcp lang=us 
> ramdisk_size=100000 init=/etc/init apm=power-off nomce vga=791 
> initrd=knoppix/miniroot.gz BOOT_IMAGE=knoppix noacpi
> 
> wag#1: you need a bigger ramdisk_size=100000
> 
> ^Carl K
> 
> 

-- 
NeoMail - Webmail that doesn't suck... as much.
http://www.neomail.org