Another attempt at S6 init

From: TheOldFellow <theoldfellow_at_gmail.com>
Date: Tue, 21 Apr 2015 15:34:18 +0100

Folks I have decided to have another go at getting an S6 system going. At
the moment I am frustrated by my stage 1 kernel-panicing just after there
is no possibility of help, so I post in the hope that some suggestion may
make sense to my ancient grey cells.

I have built a fresh LFS, it is on a LVM2 partition, so I need an
Initramfs, and this enters the Stage 1 script with a devtmpfs somewhat
populated, and /sys and /proc already mounted. My thinking is that these
mounts are OK, and I don't see any advantage in umounting them, just to
remake /dev in a tmpfs.

So I should need much less than Laurent has in his example. (did I mention
the ancient grey cells?)

########################################
cd /
umask 022

# close stdin
fdclose 0
if { s6-echo "*s6 init stage 1 starts *" }

#close stdout and stderr
fdclose 1 fdclose 2

# somewhere to mount /tmp and /dev/service, but not /dev
if { s6-mount -wt tmpfs -o mode=0755,size=67108864 tmpfs /mnt/tmpfs }

# copy the image of the tmpfs onto it.
if { s6hiercopy /img/tmpfs /mnt/tmpfs }

# connect stdin to /dev/null
redirfd -r 0 /dev/null
# connect stdout to a dangling fifo which will be picked up by the logger
redirfd -wnb 1 /service/s6-svscan-log/fifo # (black magic: doesn't block)
# connect stderr
fdmove -c 2 1
# load the environment variables
s6-envdir /etc/s6-init/env


background
{
  s6-setsid
  redirfd -w 1 /service/s6-svscan-log/fifo # (blocks until the logger reads)
  /etc/s6-init/init-stage2
}
unexport !

# Start stage 2.
s6-svscan -t0 /service
########################################

But somewhere along the line this panics (attempt to close init), and I am
at a loss to know how to debug it. The best suggestion is probably 'use
systemd', but please refrain.

How do you debug this sort of thing?

TOF.
Received on Tue Apr 21 2015 - 14:34:18 UTC