#!/bin/bash
set -e -u

SELF="${0##*/}"

mode='crash_kernel'
address=
port='666'
interface=
driver=
output='/var/lib/crash'
size='128M'

main () {
	parse_args "$@"
	"setup_${mode}"
}

parse_args () {
	local temp
	temp="$(getopt -n "$SELF" -o 'hca:p:i:d:o:s:v' -l 'help,collector,address:,port:,interface:,driver:,output:,size:,verbose' -- "$@")" ||
		die "Terminating"
	eval set -- "$temp"
	while [ -n "$1" ]
	do
		case "$1" in
			-h|--help) usage 0 ;;
			-c|--collector) mode='collector' ; shift ;;
			-a|--address) address="$2" ; shift 2 ;;
			-p|--port) port="$2" ; shift 2 ;;
			-i|--interface) interface="$2" ; shift 2 ;;
			-d|--driver) interface="$2" ; shift 2 ;;
			-o|--output) output="$2" ; shift 2 ;;
			-s|--size) size="$2" ; shift 2 ;;
			-v|--verbose) set -x ; shift ;;
			--) shift ; break ;;
			*) die "Internal error!" ;;
		esac
	done
	CRASHHOSTNAME="${1:-}"
}
usage () {
	echo "Usage: $SELF [options] (--collector | <FQDN-of-crash-capture-host>)"
	echo " Either setup host as crash collector, which installs a service listening for incoming connections,"
	echo " or setup local host to send memory dumps on crashes to given host."
	echo
	echo "Options:"
	echo " -c, --collector"
	echo " -o, --output <output-directory> [default: '${output}']"
	echo
	echo " -a, --address <IP-address>"
	echo " -i, --interface <network-interface>"
	echo " -d, --driver <network-driver>"
	echo " -s, --size <reserved-memory> [default: '${size}']"
	echo
	echo " -p, --port <TCP-port> [default: '${port}']"
	echo " -v, --verbose"
	echo " -h, --help"
	exit ${1:-0}
}

setup_collector () {
	echo "Setting up as crash collector..."
	# Create user and service for collecting data from the network:
	adduser --system crash --home /var/lib/crash

	create_file "${output}/collector.sh" <<__SH__
#!/bin/sh
umask 0177
exec cat >"${output}/core-\$(date +%s).gz"
__SH__
	update-inetd --add "${port}\tstream\ttcp\tnowait\tcrash\t${output}/collector.sh"

	# Make sure the port is allowed by the firewall:
	ucr set "security/packetfilter/tcp/${port}/all=ACCEPT"
	service univention-firewall restart
}

setup_crash_kernel () {
	echo "Setting up crash kernel..."
	detect_hypervisor
	detect_network
	install_kexec
	create_initramfs_hook
	create_initramfs_modules
	create_initramfs_script
	create_initramfs
	setup_grub
	setup_kernel
	setup_boot
	finish
}
detect_hypervisor () {
	local hypervisor="$(cat /sys/hypervisor/type 2>/dev/null || :)"
	echo "Hypervisor: $hypervisor"
	case "$hypervisor" in
	xen) [ -n "${FORCE_XEN:-}" ] || die "Xen is not supported. (Xen implements its own mechanism to capture crashed domains. Support for kexec was only added later. If you want to try anyway, set FORCE_XEN=1)" ;;
	esac
}
detect_network () {
	local brif sys
	declare -a pif=()
	[ -n "$CRASHHOSTNAME" ] || die "Missing address"
	: ${address:=$(dig +short +search "$CRASHHOSTNAME" | tail -n 1)}
	echo " using IP ${address:?Failed to resolve "$CRASHHOSTNAME"}"
	: ${interface:=$(ip route get "$address" | sed -rne 's,.* dev (\w+) +src .*,\1,p')}
	echo " using network interface ${interface:?Network interface not found; use '-i <network-interface>'}"
	sys="/sys/class/net/$interface/brif"
	if [ -d "$sys" ]
	then
		echo " bridge interface $interface detected - looking for physical interfaces..."
		for brif in "$sys"/*
		do
			brif="${brif##*/}"
			[ -e "/sys/class/net/$brif/device/driver" ] || continue
			echo "  found $brif"
			pif+=("$brif")
		done
		case "${#pif[@]}" in
		0) die "No physical interface found; use '-i <network-interface>'" ;;
		1) ;;
		*) die "More than one interface found; use '-i <network-interface>'" ;;
		esac
	else
		pif="$interface"
	fi
	: ${driver:=$(basename "$(readlink "/sys/class/net/$pif/device/driver")")}
	echo " using network driver ${driver:?Network driver not found; use '-d <network-driver>'}"
}
install_kexec () {
	which kexec >/dev/null 2>/dev/null && return
	ucr set repository/online/unmaintained=yes
	apt-get -qq update
	apt-get -qq install kexec-tools
	sed -i -e '/^LOAD_KEXEC=/s/true/false/' /etc/default/kexec
	kexec -u || :
}
create_initramfs_hook () {
	create_file /etc/initramfs-tools/hooks/crash <<'__SH__'
#!/bin/sh
PREREQ=
prereqs () {
echo "$PREREQ"
}
case "$1" in
prereqs)
 prereqs
 exit 0
 ;;
esac
. /usr/share/initramfs-tools/hook-functions
copy_exec /bin/ip
copy_exec /bin/nc
__SH__
}
create_initramfs_modules () {
	grep -qFe "$driver" /etc/initramfs-tools/modules && return
	echo "$driver" >>/etc/initramfs-tools/modules
	update_initramfs=true
}
create_initramfs_script () {
	create_file /etc/initramfs-tools/scripts/crash <<__SH__
#!/bin/sh
exec >/dev/console 2>/dev/console
set -x
trap 'sh </dev/console >/dev/console 2>/dev/console' EXIT
echo "Loading network driver '$driver'..."
modprobe "$driver"
sleep 5
cd /sys/class/net
for iface in *
do
	echo "Trying network interface '\$iface'..."
	[ -e "\$iface/device/driver" ] || continue
	ip addr add "$(ip addr show dev "$interface" | sed -rne 's,^.*inet ([0-9./]+) brd.*,\1,p')" dev "\$iface"
	ip link set "\$iface" up
	ip route add default via "$(ip route list | sed -rne 's,^default via ([0-9./]+) dev .*,\1,p')"
	echo "Dumping crash to '$address'..."
	gzip -c </proc/vmcore | nc "$address" "$port" && {
		echo "Done. Rebooting in 3s..."
		sleep 3
		echo b >/proc/sysrq-trigger
		break
	}
	echo "Failed."
	ip addr flush dev "\$iface"
done
__SH__
}
create_initramfs () {
	"$update_initramfs" || return
	# kernel + initramfs + uncompressed <= 128 MB !
	ucr set initramfs/modules=dep
	update-initramfs -u
}
setup_grub () {
	grub=$(ucr get grub/append)
	case "$grub" in
	*crashkernel=*) ;;
	*) ucr set grub/append="$grub crashkernel=${size}" ;; # 384M-2G:64M,2G-:128M
	esac
	ucr set grub/gfxmode= grub/gfxpayload= grub/bootsplash=nosplash grub/loglevel=7 grub/quiet=no
}
setup_kernel () {
	create_file /etc/sysctl.d/crash.conf 0644 <<__CONF__
# panic when hung task is detected
kernel.hung_task_panic=0
# panic on NMIs from I/O
kernel.panic_on_io_nmi=0
# panic on oops or kernel bug detection
kernel.panic_on_oops=1
# panic on NMIs from memory or unknown
kernel.panic_on_unrecovered_nmi=1
# panic when soft lockups are detected
kernel.softlockup_panic=0
# panic when out-of-memory happens
vm.panic_on_oom=0
__CONF__
}
setup_boot () {
	create_file /etc/rc.local '' FORCE <<'__SH__'
#!/bin/sh -e
#
# rc.local
#
# This script is executed at the end of each multiuser runlevel.
# Make sure that the script will "exit 0" on success or any other
# value on error.
#
# In order to enable or disable this script just change the execution
# bits.
#
# By default this script does nothing.

load_crash () {
	[ "$(cat /sys/kernel/kexec_crash_loaded)" = 0 ] || return
	console=--reset-vga
	args='1 irqpoll maxcpus=1 reset_devices boot=crash debug break=init nosplash'
	set -- $(cat /proc/cmdline)
	while [ $# -ge 1 ]
	do
		case "$1" in
		blacklist=*) args="${args} $1" ;;
		boot=*) ;;
		BOOTIF=*) args="${args} $1" ;;
		BOOT_IMAGE=*) ;;
		break|break=*) ;;
		console=ttyS*) args="${args} $1" console=--console-serial ;;
		console=*) args="${args} $1" ;;
		crashkernel=*) ;;
		drop_capabilities=*) ;;
		fastboot|forcefsck|fsckfix) ;;
		loglevel=*) args="${args} $1" ;;
		netconsole=*) args="${args} $1" ;;
		nfsroot=*|ip=*) ;;
		panic=*) ;;
		quiet|verbose|debug|debug=*) ;;
		resume=*|resume_offset=*|noresume) ;;
		root=*|rootflags=*|rootfstype=*|rootdelay=*) ;;
		ro|rw) ;;
		splash|nosplash) ;;
		*) echo "Unknown arg '$1'" ;;
		esac
		shift
	done

	kexec -p /boot/vmlinuz-`uname -r` --initrd /boot/initrd.img-`uname -r` --append="$args" $console
	cat /sys/kernel/kexec_crash_loaded
}
load_crash
:
__SH__
}
finish () {
	if [ "$(cat /sys/kernel/kexec_crash_size)" = 0 ]
	then
		echo "First 'reboot'; after that test with 'echo c >/proc/sysrq-trigger'"
	else
		if [ "$(cat /sys/kernel/kexec_crash_loaded)" != 0 ]
		then
			kexec -p -u
		fi
		/etc/rc.local
		echo "Test with 'echo c >/proc/sysrq-trigger'"
	fi
}

update_initramfs=false
create_file () {
	[ -z "${3:-}" ] && [ -f "$1" ] && return 0
	cat >"$1"
	chmod "${2:-0755}" "$1"
	case "$1" in
	*/initramfs-tools/*) update_initramfs=true ;;
	esac
}

die () {
	echo "${0##*/}: $*" >&2
	exit 1
}

main "$@"