#!/usr/bin/bash
#
# CDDL HEADER START
#
# The contents of this file are subject to the terms of the
# Common Development and Distribution License, Version 1.0 only
# (the "License").  You may not use this file except in compliance
# with the License.
#
# You can obtain a copy of the license at http://smartos.org/CDDL
#
# See the License for the specific language governing permissions
# and limitations under the License.
#
# When distributing Covered Code, include this CDDL HEADER in each
# file.
#
# If applicable, add the following below this CDDL HEADER, with the
# fields enclosed by brackets "[]" replaced with your own identifying
# information: Portions Copyright [yyyy] [name of copyright owner]
#
# CDDL HEADER END
#
# Copyright 2022 Joyent, Inc.
#

export PS4='[\D{%FT%TZ}] ${BASH_SOURCE}:${LINENO}: ${FUNCNAME[0]:+${FUNCNAME[0]}(): }'
export PATH=/usr/bin:/usr/sbin:$PATH

set -o xtrace

. /lib/svc/share/smf_include.sh
smf_is_globalzone && exit ${SMF_EXIT_OK}

if [ ! -x /usr/sbin/mdata-get ]; then
  echo "Metadata mdata-get tool not found."
  exit ${SMF_EXIT_ERR_FATAL}
fi

function fatal() {
  if [[ -n $1 ]]; then
      echo "FATAL: $*" >&2
  fi
  exit ${SMF_EXIT_ERR_FATAL}
}

# Test if an address looks like an IPv4 address.
function isIPv4() {
  [[ "$1" =~ ^([0-9]{1,3}\.){3}[0-9]{1,3}$ ]]
}

# Test if an address looks like an IPv4 address + CIDR.
function isIPv4AndCIDR() {
  [[ "$1" =~ ^([0-9]{1,3}\.){3}[0-9]{1,3}/[0-9]{1,2}$ ]]
}

# Test if an address looks like an IPv6 address.
function isIPv6() {
  [[ "${1,,}" =~ ^[0-9a-f]*:[0-9a-f]*:[:0-9a-f]*$ ]]
}

# Test if an address looks like an IPv6 address + CIDR.
function isIPv6AndCIDR() {
  [[ "${1,,}" =~ ^[0-9a-f]*:[0-9a-f]*:[:0-9a-f]*/[0-9]{1,3}$ ]]
}

# For old zones that were created prior to OS-2253 and bumping the mdata:fetch
# start timeout, we need to fix this otherwise we could timeout waiting for the
# socket.
cur_timeout=$(svcprop -p start/timeout_seconds svc:/smartdc/mdata:fetch)
if [[ -z ${cur_timeout} || ${cur_timeout} -lt 1800 ]]; then
    # The current manifest has an old timeout value, fix in case we timeout
    # here.  XXX we can still hit OS-2296 here where smf will forget that we
    # set this.
    svccfg -s svc:/smartdc/mdata:fetch 'setprop start/timeout_seconds = 1800'
    svcadm refresh svc:/smartdc/mdata:fetch
fi

# This waits until /.zonecontrol/metadata.sock exists then exits 0
/usr/vm/sbin/filewait /.zonecontrol/metadata.sock

if [[ ! -e /.zonecontrol/metadata.sock ]]; then
    # this is a bug since filewait should not have returned until file existed.
    fatal "missing /.zonecontrol/metadata.sock, Unable to start mdata:fetch"
fi

# Update sysinfo to ensure values that come from metadata are populated.
/usr/bin/sysinfo -fu

echo "Retrieving metadata user-data"
/usr/sbin/mdata-get user-data >/var/db/mdata-user-data.new
case $? in
  0)
    echo "Metadata user-data successfuly retrieved."
    mv /var/db/mdata-user-data{.new,}
    ;;
  1)
    echo "Metadata user-data not defined."
    rm -f /var/db/mdata-user-data{,.new}
    ;;
  *)
    echo "Metadata couldn't be retrieved."
    exit ${SMF_EXIT_ERR_FATAL}
    ;;
esac

echo "Retrieving metadata user-script..."
/usr/sbin/mdata-get user-script >/var/svc/mdata-user-script.new
case $? in
  0)
    echo "Metadata user-script successfuly retrieved."
    mv /var/svc/mdata-user-script{.new,}
    chmod +x /var/svc/mdata-user-script
    ;;
  1)
    echo "Metadata user-script not defined."
    rm -f /var/svc/mdata-user-script{,.new}
    ;;
  *)
    echo "Metadata couldn't be retrieved."
    exit ${SMF_EXIT_ERR_FATAL}
    ;;
esac

echo "Retrieving metadata operator-script..."
/usr/sbin/mdata-get sdc:operator-script >/var/svc/mdata-operator-script.new
case $? in
  0)
    echo "Metadata operator-script successfuly retrieved."
    mv /var/svc/mdata-operator-script{.new,}
    chmod +x /var/svc/mdata-operator-script
    ;;
  1)
    echo "Metadata operator-script not defined."
    rm -f /var/svc/mdata-operator-script{,.new}
    ;;
  *)
    echo "Metadata couldn't be retrieved."
    exit ${SMF_EXIT_ERR_FATAL}
    ;;
esac

echo "Retrieving tmpfs value..."
tmpfs=$(/usr/sbin/mdata-get sdc:tmpfs)
if [[ $? == 0 && -n ${tmpfs} && -f /etc/vfstab ]]; then
  check="swap        -   /tmp                tmpfs";

  if [[ ${tmpfs} == "0" ]]; then
    # When tmpfs is set 0, we remove any entry from /etc/vfstab but cannot
    # adjust the "live" value as current /tmp will be in-use. On reboot the
    # new value will take effect.

    grep -v "^${check}" /etc/vfstab > /etc/vfstab.new \
      && mv /etc/vfstab.new /etc/vfstab
  else
    new="swap        -   /tmp                tmpfs    -   yes    size=${tmpfs}m";
    if ! /usr/bin/grep "^${new}" /etc/vfstab; then
      if ! /usr/bin/grep "^${check}" /etc/vfstab; then
        # no tmpfs line. add it.
        echo "${new}" >> /etc/vfstab
      else
        # existing tmpfs line, but wrong value. fix it.
        /usr/bin/sed -i "" -e "s|^swap.*/tmp.*tmpfs.*$|${new}|" /etc/vfstab
        echo $?
      fi

      if mount | grep "^/tmp"; then
        # Also fix current size, since /etc/vfstab didn't have our correct line
        # but only if we have /tmp mounted at all. If not, we'll have to wait
        # until the next reboot since /tmp will be in-use.
        /usr/sbin/mount -F tmpfs -o remount,size=${tmpfs}m /tmp
      fi

    fi
  fi
fi

#
# If we have NFS volumes, we'll add them to vfstab, and enable the nfs/client
# service so that will mount the volumes for us.
#
echo "Retrieving volume metadata..."

volumes_added=0
while IFS="|" read -r nfsvolume mountpoint name mode type; do

  cat >&2 <<EOF
*** VOLUME ***
NFSVOLUME: ${nfsvolume}
MOUNTPOINT: ${mountpoint}
MODE: ${mode}
NAME: ${name}
TYPE: ${type}
EOF

  if [[ -n ${type} && ${type} != "tritonnfs" ]]; then
      fatal "unsupported volume type: ${type}"
  fi

  # if we don't have volume and mountpoint, we can't add vfstab lines
  if [[ -z ${nfsvolume} || -z ${mountpoint} ]]; then
      fatal "invalid volume specification (need both volume & mountpoint)"
  fi

  if ! grep "^${nfsvolume}[ 	]" /etc/vfstab; then
      if [[ -z ${mode} ]]; then
          mode="rw"
      fi

      if [[ ${mode} != "rw" && ${mode} != "ro" ]]; then
          fatal "invalid volume mode: '${mode}'"
      fi

      line=$(printf "%s - %s nfs - yes %s\n" "${nfsvolume}" "${mountpoint}" "${mode}")
      volumes_added=$((${volumes_added}+1))

      mkdir -p ${mountpoint}
      echo "${line}" >> /etc/vfstab
  fi
done < <(/usr/sbin/mdata-get sdc:volumes \
    | /usr/bin/json \
        -d '|' \
        -a nfsvolume mountpoint name mode type)

if [[ ${volumes_added} -gt 0 ]]; then
    for svc in \
        svc:/network/nfs/nlockmgr:default \
        svc:/network/nfs/status:default \
        svc:/network/rpc/bind:default \
        svc:/network/nfs/client:default \
        ; do

        svcadm enable ${svc} || fatal "Unable to enable ${svc}"
     done
fi

# We use the special sdc:nics value here though this is not an interface for
# use elsewhere.  If this is changed please also update agent.js in the metadata
# agent.
#
# We run this every startup in case nics have changed since last boot. As
# network/physical has an optional_all dependency on this service, we'll have
# had our chance to write these files out before networking comes up. This
# might eventually be replaced by network/physical grabbing data directly.
echo "Retrieving nics data..."
while IFS="|" read -r iface ip ips netmask primary gateway gateways; do

  # if we don't have interface name, we don't know what files to write out.
  if [[ -z ${iface} ]]; then
    continue;
  fi

  # A VM created on an older platform may not have the "ips" or "gateways"
  # properties, so we will need to grab the older "ip" and "gateway" versions
  # instead.
  [[ -z $ips ]] && ips=$ip
  [[ -z $gateways ]] && gateways=$gateway

  # We remove the hostname.netX file first, so we can append to a clean file
  rm -f /etc/hostname.${iface}

  # remove any existing DHCP or addrconf configuration, since we'll create one
  # if it belongs later
  rm -f /etc/dhcp.${iface}
  rm -f /etc/addrconf.${iface}

  OLDIFS=$IFS
  IFS=$','
  for ip in $ips; do
    # so it shows up in the logs
    echo "iface[${iface}] ip[${ip}] netmask[${netmask}]" \
      "primary[${primary}] gateway[${gateway}]"

    [[ -z ${ip} ]] && continue;

    if [[ ${ip} == "dhcp" ]]; then
      touch /etc/dhcp.${iface}
      if hostname=`mdata-get sdc:hostname`; then
          echo "inet ${hostname}" >> /etc/hostname.${iface}
      fi
    elif [[ ${ip} == "addrconf" ]]; then
      touch /etc/addrconf.${iface}
    elif isIPv4 ${ip} && [[ -n ${netmask} ]]; then
      # We're using an older configuration where the routing prefix is
      # specified using a mask instead of CIDR notation. We'll need to
      # invoke ifconfig differently.
      echo "${ip} netmask ${netmask} up" >> /etc/hostname.${iface}
    elif isIPv4AndCIDR ${ip} \
        || isIPv6AndCIDR ${ip} \
        || isIPv6 ${ip}; then
      # Either a routing prefix was specified, or, in the case of IPv6, we
      # won't specify one and fall back on NDP to find it when we configure
      # our interfaces.
      echo "${ip} up" >> /etc/hostname.${iface}
    fi
  done

  if [[ ${primary} == "true" ]]; then
    rm -f /etc/defaultrouter
    for gateway in $gateways; do
      if [[ -n ${gateway} ]] && isIPv4 ${gateway}; then
        echo "${gateway}" >> /etc/defaultrouter
      fi
    done
  fi
  IFS=$OLDIFS

  # XXX we leave old hostname.netX files around and just replace when we have
  # one next.
done < <(/usr/sbin/mdata-get sdc:nics \
    | /usr/bin/json \
        -d '|' \
        -e 'this.ips = this.ips ? this.ips.join(",") : ""' \
        -e 'this.gateways = this.gateways ? this.gateways.join(",") : ""' \
        -a interface ip ips netmask primary gateway gateways)

# rebuild resolv.conf
resolvers=$(mdata-get sdc:resolvers)
resolvers_result=$?

# Determine if resolv.conf is managed for us
maintain_resolvers=$(mdata-get sdc:maintain_resolvers)
maintain_result=$?
if [[ ${maintain_result} != 0 ]]; then
    echo "Error getting maintain_resolvers, code: ${maintain_result}"
    maintain_resolvers="false"
fi

# If this is our first boot, write an initial set of resolvers
if [[ -f /var/svc/provisioning ]]; then
    echo "First boot: writing resolvers"
    maintain_resolvers="true"
fi

if [[ ${resolvers_result} == 0 && ${maintain_resolvers} == "true" ]]; then

    # if dns_domain is missing or broken, we still write out, just w/o search
    search="search $(mdata-get sdc:dns_domain)"
    if [[ $? != 0 ]]; then
        search=
    fi

    if [[ ${resolvers} == "[]" ]]; then
        nameservers=
    else
        nameservers=$(echo ${resolvers} | json -a | sed -e "s/^/nameserver /")
    fi

    rm -f /etc/.resolv.conf.tmp
    if [[ -n ${search} ]]; then
        echo "${search}" > /etc/.resolv.conf.tmp
    fi
    if [[ -n ${nameservers} ]]; then
        echo "${nameservers}" >> /etc/.resolv.conf.tmp
    fi

    cp /etc/.resolv.conf.tmp /etc/resolv.conf \
        && cat /etc/.resolv.conf.tmp >&2 \
        && rm -f /etc/.resolv.conf.tmp
else
    if [[ ${resolvers_result} == 0 ]]; then
        echo "Error getting resolvers, code: ${resolvers_result}"
    fi

    if [[ ${maintain_resolvers} == "true" ]]; then
        echo "Not setting resolvers, maintain_resolvers=${maintain_resolvers}"
    fi
fi


# Fetch routes

# It is possible to specify the same route in several different ways using
# route(1m). We therefore use route(1m) itself to manage adding, deleting
# and determining duplicates.
zone_routes_file=/etc/inet/static_routes
vmadm_routes_file=/etc/inet/static_routes.vmadm
tmpdir=$(mktemp -d /tmp/mdata.XXXXXX)

if [ -z $tmpdir ]; then
    echo "Error creating temporary directory."
    exit ${SMF_EXIT_ERR_FATAL}
fi

# directory structure for the new copy of static_routes.vmadm (the one that
# will replace the current static_routes.vmadm once all of the adds and
# deletes have been applied):
new_root=${tmpdir}/new-routes
new_inet=${new_root}/etc/inet
# directory structure for the previous copy of static_routes.vmadm (used to
# determine routes that have been removed since the last time mdata-fetch
# was run):
old_root=${tmpdir}/old-routes
old_inet=${old_root}/etc/inet
# directory structure for the zone's persistent routes - those not created
# by vmadm (used to determine if vmadm routes are duplicates):
zone_root=${tmpdir}/zone-routes
zone_inet=${zone_root}/etc/inet

mkdir -p $new_inet
mkdir -p ${old_root}/etc/inet
mkdir -p ${zone_root}/etc/inet

if [[ -f ${vmadm_routes_file} ]]; then
    cp $vmadm_routes_file ${old_inet}/static_routes
fi

function route_in_per_zone_file()
{
    cp $zone_routes_file $zone_inet
    output=$(route -pR $zone_root add $* 2>&1)
    [[ $output =~ "entry exists" ]]
}

# If re-running this script after initial boot, network/physical and
# network/routing-setup are already enabled, so apply routing adds and
# deletes manually
if [[ $(/usr/bin/svcs -H -o state network/routing-setup) == "online" ]]; then
    routing_up="true"
fi

while IFS="|" read -r gateway dst linklocal; do
    echo "route: gateway[${gateway}] dst[${dst}] linklocal[${linklocal}]"
    route_type=""
    if [[ ${linklocal} == "true" ]]; then
        route_type="-interface "
    fi
    route_str="${route_type}${dst} ${gateway}"

    if route_in_per_zone_file $route_str; then
        echo "not adding duplicate route: ${route_str}"
        # the zone has also defined this route; do nothing
    else
        echo "adding route to file: ${route_str}"
        route -pR $new_root add $route_str
        if [[ -n "${routing_up}" ]]; then
            route add $route_str
        fi
    fi

    route -pR $old_root delete $route_str
done < <(/usr/sbin/mdata-get sdc:routes \
    | /usr/bin/json -d '|' -a gateway dst linklocal)


# Anything left in the old static_routes file is a delete. Don't delete the
# route from the routing tables if there's a duplicate route in the zone's
# static_routes file
if [[ -f ${old_inet}/static_routes ]]; then
    egrep -v "^(#|$)" ${old_inet}/static_routes | while read -r route_str; do
        if [[ "${route_str}" == "" ]]; then
            continue
        fi

        if route_in_per_zone_file "$route_str"; then
            echo "not deleting duplicate route: ${route_str}"
        else
            if [[ -n "${routing_up}" ]]; then
                route delete $route_str
            fi
        fi
    done
fi

if [[ -f ${new_inet}/static_routes ]]; then
    cp ${new_inet}/static_routes ${vmadm_routes_file}
else
    rm ${vmadm_routes_file}
fi
rm -rf $tmpdir


# Unconditionally enable mdata:execute, so that the last provisioning step
# is always taken (regardless of whether user-script exists or not)
svcadm enable smartdc/mdata:execute

exit ${SMF_EXIT_OK}