#!/bin/ksh93
#
#
# This file and its contents are supplied under the terms of the
# Common Development and Distribution License ("CDDL"), version 1.0.
# You may only use this file in accordance with the terms of version
# 1.0 of the CDDL.
#
# A full copy of the text of the CDDL should have accompanied this
# source.  A copy of the CDDL is also available via the Internet at
# http://www.illumos.org/license/CDDL.
#

#
# Copyright 2020 Joyent, Inc.
#

# Traditionally, when a Triton compute node boots, the network is not
# configured until after the local filesystems are mounted (in fact
# network/physical:default depends upon /system/filesystem/smartdc).  Most
# obviously, in the case of a head node or a standalone SmartOS install, the
# network configuration is stored on the local zpool, so the network cannot
# be configured until this happens.
#
# For Triton compute nodes with encrypted zpools, we must enable the admin
# network before the local zpool filesytems are online -- we have to be able
# to communicate to the head node services to obtain the pin to unlock the
# local zpool.  For PXE booted Triton compute nodes, we therefore configure
# the admin network sooner via the network/early-admin:default service.  When
# network/physical:default runs, it will skip the configuration of the admin
# network and configuring the remaining interfaces.
#
# If we are not a Triton compute node, we exit successfully almost immediately
# without configuring the admin network.  When the network/physical:default
# service runs, it will configure all the network interfaces as it traditionally
# has

PATH=/bin:/usr/bin:/sbin:/usr/sbin

. /lib/svc/share/smf_include.sh
. /lib/sdc/config.sh
. /lib/sdc/network.sh

PS4='+ [$LINENO] '

set -o xtrace

if ! smf_is_globalzone; then
    echo "Non-global zone; no action required; exiting"
    exit $SMT_EXIT_OK
fi

# XXX Until OS-8367 is sorted out, immediately enable the varpd service.
# SMF will properly cope with things if it fails (and blocks both
# network/physical and zones).
svcadm enable varpd

if ! boot_file_config_enabled; then
    echo "Boot-time networking files not present; exiting"
    exit $SMF_EXIT_OK
fi

function fatal
{
    # XXX: For SMF methods, does it matter/better to redirect to stderr?
    echo "Error: $*" >&2
    exit $SMF_EXIT_ERR_FATAL
}

if ! boot_file_config_valid; then
    echo "ERROR: boot-time network config file incorrect" >&2
    exit $SMF_EXIT_ERR_CONFIG
fi

unset aggrs
unset tags
typeset -A tagv
typeset -A aggr_links aggr_mode
typeset -A ip ip6 netmask gateway gateway6 mtu mac
/usr/lib/sdc/net-boot-config | while IFS="=" read var value; do
    if [[ "$var" =~ _nic$ ]]; then
        name=${var%_nic}
        tags+=("${name}")
        tagv[$name]="$value"
    elif [[ "$var" =~ _aggr$ ]]; then
        name=${var%_aggr}
        aggrs+=(${name})
        aggr_links[$name]="${value//,/ }"
    elif [[ "$var" =~ _lacp_mode$ ]]; then
        name=${var%_lacp_mode}
        aggr_mode[$name]="$value"
    elif [[ "$var" =~ _mtu$ ]]; then
        (( value < 1500 || value > 65535 )) &&
            fatal "ERROR: $var MTU value \'$value\' is not in" \
            "range [1500, 65535]"

        name=${var%_mtu}
        mtu[$name]="$value"
    elif [[ "$var" =~ _ip$ ]]; then
        name=${var%_ip}
        ip[$name]="$value"
    elif [[ "$var" =~ _ip6$ ]]; then
        name=${var%_ip6}
        ip6[$name]="$value"
    elif [[ "$var" =~ _netmask$ ]]; then
        name=${var%_netmask}
        netmask[$name]="$value"
    elif [[ "$var" =~ _gateway$ ]]; then
        name=${var%_gateway}
        gateway[$name]="$value"
    elif [[ "$var" =~ _gateway6$ ]]; then
        name=${var%_gateway6}
        gateway6[$name]="$value"
    elif [[ "$var" =~ _mac$ ]]; then
        name=${var%_mac}
        mac[$name]="$value"
    elif [[ "$var" == "dns_resolvers" ]]; then
        dns_resolvers=(${value//,/ })
    fi

    eval "CONFIG_$var"="$value"
done

# This must happen befor any other dladm commands (which includes log_if_state)
# or else dladm commands can fail
dladm init-phys

log_if_state before

typeset -A mac_to_link
out=$(dladm show-phys -mpo link,address)
(( $? == 0 )) || fatal "dladm show-phys failed"
while IFS=: read link addr; do
    macaddr=$(normalize_mac $addr)
    mac_to_link["$macaddr"]="$link"
done <<< "$out"

ADMIN_NIC_TAG=${CONFIG_admin_tag:-"admin"}
[[ -n "${tagv[$ADMIN_NIC_TAG]}" ]] || \
    fatal "ERROR: admin nic tag '$ADMIN_NIC_TAG' not present"

# Set $nic to the link that has the admin nic tag. For unfortunate
# historic reasons, for aggrs, the MAC address and link name of an aggr
# are the same value by net-boot-config (and the source of many errors).
# If ${aggr_links[$nic]} is non-empty, it means $nic isn't a MAC address
# but an aggr name (and we're done). Otherwise we must map the MAC address
# from $tagv back to the link.
nic="${tagv[$ADMIN_NIC_TAG]}"
if [[ -z "${aggr_links[$nic]}" ]]; then
    if !valid_mac "$nic"; then
        fatal "ERROR: admin mac address $nic not found on system"
    fi

    _nic=${mac_to_link[$nic]}
    if [[ -z "$_nic" ]]; then
        fatal "ERROR: Invalid value of ${ADMIN_NIC_TAG}_nic ($nic)"
    fi
    nic=$_nic
fi

# If there are other nic tags configured on the same link as
# the admin tag, find the largest MTU to use to set the
# datalink MTU
dlmtu="${mtu[$ADMIN_NIC_TAG]}"

for tag in ${tags[@]}; do
    tagmac="${tagv[$tag]}"

    # If the 'mac' for the nic tag is actually an aggr, the
    # tag will appear in $aggr_links (and we want to use that as
    # the link name).  If it is not an aggr, we need to map the
    # MAC address to the link name so we can check if $tag
    # resides on the same link as the admin interface
    if [[ -n "${aggr_links[$tagmac]}" ]]; then
        link="$tagmac"
    else
        link="${mac_to_link[$tagmac]}"
    fi

    if [[ "$link" == "$nic" && $dlmtu -lt ${mtu[$tag]} ]]; then
        dlmtu=${mtu[$tag]}
    fi
done


if [[ -n "${aggr_links[$nic]}" ]]; then
    mode=${aggr_mode[$nic]:-"off"}

    for l in ${aggr_links[$nic]}; do
        [[ -n "${mac_to_link[$l]}" ]] || fatal "MAC '$l' not present"
        link="${mac_to_link[$l]}"
        if [[ -n "$dlmtu" ]]; then
            dladm set-linkprop -p mtu=$dlmtu $link || \
                fatal "ERROR: Failed to set mtu on $link to $dlmtu"
        fi
        links+="${link} "
    done
    links="${links% }"

    echo "Creating aggr: $nic (mode=$mode, links=${links})"
    dladm create-aggr -l ${links// / -l } -L $mode $nic
fi

if [[ -n "$dlmtu" ]]; then
    dladm set-linkprop -p mtu=$dlmtu $nic || \
        fatal "ERROR: Failed to set mtu on aggr $nic to $dlmtu"
fi

driver=${nic%%+([0-9])}
get_link_state $nic
if [[ "$link_state" == "down" ]]; then
    echo "admin nic '${nic}' is down: unplumbing"
    /sbin/ifconfig $nic down unplumb
    wait_for_nic_state $nic "unknown"
fi

# There's some sort of race condition in the bnx driver: if the plumb
# command comes too soon after the unplumb, the interface can come up
# in a state where it never fires interrupts for link state changes.
if [[ "$driver" == "bnx" ]]; then
    sleep 5
fi

/sbin/ifconfig $nic plumb mtu ${mtu[$ADMIN_NIC_TAG]}
wait_for_nic_state $nic "up"

if [[ -n "${ip[$ADMIN_NIC_TAG]}" ]]; then
    /sbin/ifconfig $nic inet ${ip[$ADMIN_NIC_TAG]} \
        netmask ${netmask[$ADMIN_NIC_TAG]:-"+"} up
    [[ -n "${gateway[$ADMIN_NIC_TAG]}" ]] && \
        /usr/sbin/route add default ${gateway[$ADMIN_NIC_TAG]}
fi

if [[ -n "${ip6[$ADMIN_NIC_TAG]}" ]]; then
    /sbin/ifconfig $nic inet6 plumb mtu ${mtu[$ADMIN_NIC_TAG]}
    [[ "${ip6[$ADMIN_NIC_TAG]}" != "addrconf" ]] && \
        /sbin/ifconfig $nic inet6 addif ${ip6[$ADMIN_NIC_TAG]} preferred up
    [[ -n "${gateway6[$ADMIN_NIC_TAG]}" ]] && \
        /usr/sbin/route add -inet6 default ${gateway6[$ADMIN_NIC_TAG]}
fi

# Add just the routes reachable through the admin network -- usually these are
# only present with rack aware networking (RAN)
/usr/lib/sdc/net-boot-config --routes | while read dst gw; do
    if ! ip_in_net $gw ${ip[$ADMIN_NIC_TAG]} ${netmask[$ADMIN_NIC_TAG]}; then
        continue
    fi
    route add "$dst" "$gw"
done

if [[ -n "${CONFIG_dns_domain}" && -n ${dns_resolvers[0]} ]]; then
    echo "search ${CONFIG_dns_domain}" > /etc/resolv.conf
    for serv in ${dns_resolvers[@]}; do
        echo "nameserver $serv" >> /etc/resolv.conf
    done
fi

touch /etc/svc/volatile/.early_admin_setup

log_if_state after