X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=initscripts%2Fpl_netinit;h=9ebc18202a7fdd27b6ae8874106713229de6af74;hb=07fee163dd47c168fe628ae417f3c1a6abb89a7c;hp=b652674e8431cf80894f280a02f8d274593e9510;hpb=bc41f5d2730be50cd9ba29d4b7ba028ae8f972db;p=bootcd.git diff --git a/initscripts/pl_netinit b/initscripts/pl_netinit index b652674..9ebc182 100755 --- a/initscripts/pl_netinit +++ b/initscripts/pl_netinit @@ -1,4 +1,4 @@ -#!/bin/sh +#!/bin/bash #-*-shell-script-*- set -x @@ -20,7 +20,7 @@ CD_NET_CONF_BOOT=/usr/boot/$NEW_NODE_CONF_NAME # the other location of cd-based network configuration file CD_NET_CONF_ROOT=/usr/$NEW_NODE_CONF_NAME -# if all other network configuration file sources +# if all other network configuration file sources # don't exist, fall back to this one (always on the cd) FALLBACK_NET_CONF=/usr/boot/default-node.txt @@ -58,51 +58,22 @@ IFCONFIG_OUTPUT=/tmp/ifconfig DEFAULT_NET_CONF=0 -function net_init_failed() { - echo - echo $(date "+%H:%M:%S") " pl_netinit: network initialization failed with interface $ETH_DEVICE" - echo - echo For forensics - echo - echo ========== lspci beg - /sbin/lspci -n | /bin/grep "Class 0200" - echo ========== lspci end - echo - echo ========== ifconfig beg - /sbin/ifconfig - echo ========== ifconfig end - echo ========== ip address show beg - ip address show - echo ========== ip address show end - echo - echo $(date "+%H:%M:%S") " pl_netinit: network initialization failed with interface $ETH_DEVICE" - echo $(date "+%H:%M:%S") " pl_netinit: shutting down machine in two hours" - /bin/sleep 2h - /sbin/shutdown -h now +function net-init-failed() { + verbose-message "pl_netinit: network initialization failed with interface $ETH_DEVICE" + verbose-forensics "failed to configure $ETH_DEVICE" + hang-and-shutdown "net-init-failed" exit 1 } -# Function for checking the IP address to see if its sensible. -function check_ip() { - case "$*" in - "" | *[!0-9.]* | *[!0-9]) return 1 ;; - esac - local IFS=. - set -- $* - [ $# -eq 4 ] && - [ ${1:-666} -le 255 ] && [ ${2:-666} -le 255 ] && - [ ${3:-666} -le 255 ] && [ ${4:-666} -le 255 ] -} - # find and parse a node network configuration file. return 0 if not found, -# return 1 if found and parsed. if this is the case, DEFAULT_NET_CONF will +# return 1 if found and parsed. if this is the case, DEFAULT_NET_CONF will # be set to 1. For any found configuration file, $USED_NET_CONF will # contain the validated contents -function find_node_config() { +function find-node-config() { /bin/rm -f $TMP_OLD_FLOPPY_CONF_FILE 2>&1 > /dev/null - echo $(date "+%H:%M:%S") " pl_netinit: looking for node configuration file on floppy" - + verbose-message "pl_netinit: looking for node configuration file on floppy" + /bin/mount -o ro -t $NODE_CONF_DEVICE_FS_TYPES /dev/fd0 \ $CONF_DEVICE_MOUNT_POINT 2>&1 > /dev/null if [[ $? -eq 0 ]]; then @@ -110,7 +81,7 @@ function find_node_config() { # 1. check for new named file first on the floppy disk if [ -r "$CONF_DEVICE_MOUNT_POINT/$NEW_NODE_CONF_NAME" ]; then conf_file="$CONF_DEVICE_MOUNT_POINT/$NEW_NODE_CONF_NAME" - echo $(date "+%H:%M:%S") " pl_netinit: found node configuration file $conf_file" + verbose-message "pl_netinit: found node configuration file $conf_file" /etc/init.d/pl_validateconf < $conf_file > $USED_NET_CONF /bin/umount $CONF_DEVICE_MOUNT_POINT return 1 @@ -122,22 +93,22 @@ function find_node_config() { elif [ -r "$CONF_DEVICE_MOUNT_POINT/$OLD_NODE_CONF_NAME" ]; then conf_file="$CONF_DEVICE_MOUNT_POINT/$OLD_NODE_CONF_NAME" /bin/cp -f $conf_file $TMP_OLD_FLOPPY_CONF_FILE - echo $(date "+%H:%M:%S") " pl_netinit: found old named configuration file, checking later." + verbose-message "pl_netinit: found old named configuration file, checking later." else - echo $(date "+%H:%M:%S") " pl_netinit: floppy mounted, but no configuration file." + verbose-message "pl_netinit: floppy mounted, but no configuration file." fi /bin/umount $CONF_DEVICE_MOUNT_POINT else - echo $(date "+%H:%M:%S") " pl_netinit: no floppy could be mounted, continuing search." + verbose-message "pl_netinit: no floppy could be mounted, continuing search." fi - # 2. check for a new named file on removable flash devices (those + # 2. check for a new named file on removable flash devices (those # that start with sd*, because usb_storage emulates scsi devices). # to prevent checking normal scsi disks, also make sure # /sys/block//removable is set to 1 - echo $(date "+%H:%M:%S") " pl_netinit: looking for node configuration file on flash based devices" + verbose-message "pl_netinit: looking for node configuration file on flash based devices" # make the sd* hd* expansion fail to an empty string if there are no sd # devices @@ -153,20 +124,20 @@ function find_node_config() { for partition in $partitions ; do check_dev=/dev/$partition - echo $(date "+%H:%M:%S") " pl_netinit: looking for node configuration file on device $check_dev" + verbose-message "pl_netinit: looking for node configuration file on device $check_dev" /bin/mount -o ro -t $NODE_CONF_DEVICE_FS_TYPES $check_dev \ $CONF_DEVICE_MOUNT_POINT 2>&1 > /dev/null if [[ $? -eq 0 ]]; then if [ -r "$CONF_DEVICE_MOUNT_POINT/$NEW_NODE_CONF_NAME" ]; then conf_file="$CONF_DEVICE_MOUNT_POINT/$NEW_NODE_CONF_NAME" - echo $(date "+%H:%M:%S") " pl_netinit: found node configuration file $conf_file" + verbose-message "pl_netinit: found node configuration file $conf_file" /etc/init.d/pl_validateconf < $conf_file > $USED_NET_CONF - echo $(date "+%H:%M:%S") " pl_netinit: found configuration" + verbose-message "pl_netinit: found configuration" /bin/umount $CONF_DEVICE_MOUNT_POINT return 1 fi - - echo $(date "+%H:%M:%S") " pl_netinit: not found" + + verbose-message "pl_netinit: ERROR - not found" /bin/umount $CONF_DEVICE_MOUNT_POINT fi @@ -180,27 +151,27 @@ function find_node_config() { # the file $TMP_OLD_FLOPPY_CONF_FILE will be readable. if [ -r "$TMP_OLD_FLOPPY_CONF_FILE" ]; then conf_file=$TMP_OLD_FLOPPY_CONF_FILE - echo $(date "+%H:%M:%S") " pl_netinit: found node configuration file $conf_file" + verbose-message "pl_netinit: found node configuration file $conf_file" /etc/init.d/pl_validateconf < $conf_file > $USED_NET_CONF return 1 fi # 4. check for plnode.txt on the cd at /usr/boot - echo $(date "+%H:%M:%S") " pl_netinit: looking for network configuration on cd in /usr/boot" + verbose-message "pl_netinit: looking for network configuration on cd in /usr/boot" if [ -r "$CD_NET_CONF_BOOT" ]; then - - echo $(date "+%H:%M:%S") " pl_netinit: found cd configuration file $CD_NET_BOOT_CONF" + + verbose-message "pl_netinit: found cd configuration file $CD_NET_BOOT_CONF" /etc/init.d/pl_validateconf < $CD_NET_CONF_BOOT > $USED_NET_CONF return 1 fi - + # 5. check for plnode.txt on the cd at /usr - echo $(date "+%H:%M:%S") " pl_netinit: looking for network configuration on cd in /usr" + verbose-message "pl_netinit: looking for network configuration on cd in /usr" if [ -r "$CD_NET_CONF_ROOT" ]; then - - echo $(date "+%H:%M:%S") " pl_netinit: found cd configuration file $CD_NET_CONF_ROOT" + + verbose-message "pl_netinit: found cd configuration file $CD_NET_CONF_ROOT" /etc/init.d/pl_validateconf < $CD_NET_CONF_ROOT > $USED_NET_CONF return 1 fi @@ -210,9 +181,9 @@ function find_node_config() { # builtin default. this can't be used to install a machine, but # will at least let it download and run the boot manager, which # can inform the users appropriately. - echo $(date "+%H:%M:%S") " pl_netinit: using default network configuration" + verbose-message "pl_netinit: using default network configuration" if [ -r "$FALLBACK_NET_CONF" ]; then - echo $(date "+%H:%M:%S") " pl_netinit: found cd default configuration file $FALLBACK_NET_CONF" + verbose-message "pl_netinit: found cd default configuration file $FALLBACK_NET_CONF" /etc/init.d/pl_validateconf < $FALLBACK_NET_CONF > $USED_NET_CONF DEFAULT_NET_CONF=1 return 1 @@ -222,25 +193,26 @@ function find_node_config() { } -echo $(date "+%H:%M:%S") " pl_netinit: bringing loopback network device up" -/sbin/ifconfig lo 127.0.0.1 up +verbose-message "pl_netinit: bringing loopback network device up" +ip address add dev lo 127.0.0.1/8 +ip link set lo up -find_node_config +find-node-config if [ $? -eq 0 ]; then # no network configuration file found. this should not happen as the # default cd image has a backup one. halt. - echo $(date "+%H:%M:%S") " pl_netinit: unable to find even a default network configuration" - echo $(date "+%H:%M:%S") " pl_netinit: file, this cd may be corrupt." - net_init_failed + verbose-message "pl_netinit: ERROR - unable to find even a default network configuration" + verbose-message "pl_netinit: file, this cd may be corrupt." + net-init-failed fi # load the configuration file. if it was a default one (not user specified), -# then remove the saved copy from /tmp, but continue on. since a network +# then remove the saved copy from /tmp, but continue on. since a network # configuration file is required and boot scripts only know about this location # they will fail (as they should) - but the network will be up if dhcp is # available -echo $(date "+%H:%M:%S") " pl_netinit: loading network configuration" +verbose-message "pl_netinit: loading network configuration" . $USED_NET_CONF if [[ $DEFAULT_NET_CONF -eq 1 ]]; then @@ -274,7 +246,7 @@ if [[ -n "$NET_DEVICE" ]]; then dev_address=$(cat $device/address | tr A-Z a-z) if [ "$device" == "$NET_DEVICE" -o "$dev_address" == "$NET_DEVICE" ]; then ETH_DEVICE=$device - echo $(date "+%H:%M:%S") " pl_netinit: found device $ETH_DEVICE with mac address $dev_address" + verbose-message "pl_netinit: found device $ETH_DEVICE with mac address $dev_address" break fi done @@ -294,79 +266,81 @@ fi # still nothing? fail the boot. if [[ -z "$ETH_DEVICE" ]]; then - echo $(date "+%H:%M:%S") " pl_netinit: unable to find a usable device, check to make sure" - echo $(date "+%H:%M:%S") " pl_netinit: the NET_DEVICE field in the configuration file" - echo $(date "+%H:%M:%S") " pl_netinit: corresponds with a network adapter on this system" - net_init_failed + verbose-message "pl_netinit: unable to find a usable device, check to make sure" + verbose-message "pl_netinit: the NET_DEVICE field in the configuration file" + verbose-message "pl_netinit: corresponds with a network adapter on this system" + net-init-failed fi # within a systemd-driven startup, we often see this stage # triggered before the network interface is actually exposed # by udev/kernel -# although of course we have network-online.target -# as a requirement; go figure what systemd actually does.. - -# in any case, let us try to work around that by allowing some delay -# here - -# tmp: Thierry June 2015 -# on fedora 21 nodes we see this running in a context where eth0 is not known to the system -# could be related to a dependency that we poorly describe to systemd -# I am increasing this timeout to 2 minutes in order to check that conjecture -ALLOW=60 +# so allow for some small delay here +ALLOW=10 COUNTER=0 while true; do - if /sbin/ifconfig $ETH_DEVICE >& /dev/null; then - echo $(date "+%H:%M:%S") " pl_netinit: device present $ETH_DEVICE, proceeding (${COUNTER}s/${ALLOW}s)" - break + ### check if the interface is known + # instead of using ifconfig, check the kernel area + if [ -e /sys/class/net/$ETH_DEVICE ]; then + verbose-message "pl_netinit: device present $ETH_DEVICE" + break fi - echo $(date "+%H:%M:%S") " pl_netinit: waiting for device $ETH_DEVICE - ${COUNTER}s/${ALLOW}s" - set -x - /sbin/ifconfig - journalctl -b | egrep 'eth|bnx|udev' - systemctl list-unit-files | grep -i network - set +x + verbose-message "pl_netinit: waiting for device $ETH_DEVICE - ${COUNTER}s/${ALLOW}s" + sleep 1 COUNTER=$(($COUNTER+1)) - [ $COUNTER -ge $ALLOW ] && net_init_failed + [ $COUNTER -ge $ALLOW ] && net-init-failed sleep 1 done # actually check to make sure ifconfig succeeds -/sbin/ifconfig $ETH_DEVICE up 2>&1 > /dev/null || { - echo $(date "+%H:%M:%S") " pl_netinit: device $ETH_DEVICE does not exist, most likely" - echo $(date "+%H:%M:%S") " pl_netinit: this CD does not have hardware support for your" - echo $(date "+%H:%M:%S") " pl_netinit: network adapter. please send the following lines" - echo $(date "+%H:%M:%S") " pl_netinit: to your PlanetLab support for further assistance" - net_init_failed +/usr/libexec/nm-ifup $ETH_DEVICE >& /dev/null || { + verbose-message "pl_netinit: device $ETH_DEVICE does not exist, most likely" + verbose-message "pl_netinit: this CD does not have hardware support for your" + verbose-message "pl_netinit: network adapter. please send the following lines" + verbose-message "pl_netinit: to your PlanetLab support for further assistance" + net-init-failed } -echo $(date "+%H:%M:%S") " pl_netinit: attempting to start networking" +verbose-message "pl_netinit: attempting to start networking" /sbin/service network start # for backwards compatibility /sbin/ifconfig $ETH_DEVICE > $IFCONFIG_OUTPUT -echo $(date "+%H:%M:%S") " pl_netinit: network online" +verbose-message "pl_netinit: network online" -echo "BEG post pl_netinit forensics" +# patch for f22 and above - if /etc/resolv.conf is empty or missing +function pl-netinit-patch-resolv-conf () { + file="$1"; shift + needed="" + # missing file : patch needed + if ! [ -f $file ]; then + needed=true + # empty file : patch needed + else + contents=$(grep -v '^#' $file) + contents=$(echo $contents) + [ -z "$contents" ] && needed=true + fi + if [ -z "$needed" ]; then + echo pl_netinit has no need to patch $file + return + fi -file=/root/.ssh/authorized_keys -echo "$file" -if ! [ -f "$file" ]; then - echo "!!!!!!!!!! missing $file" -else - echo "---------- $file" - cat $file -fi + verbose-message "pl_netinit: patching $file" + source /etc/sysconfig/network-scripts/ifcfg-${ETH_DEVICE} + # delete because it may be a symlink to /run/systemd/resolve/resolv.conf + # which looks really weird (ls -lL /etc/resolv.conf does not show anything) + rm -f $file + # mention the DNS servers defined in the config - if any + [ -n "$DNS1" ] && echo nameserver $DNS1 >> $file + [ -n "$DNS2" ] && echo nameserver $DNS2 >> $file + # also add landmark for good measure + echo "nameserver 8.8.8.8" >> $file +} -file="/etc/ssh/sshd_config" -if ! [ -f "$file" ]; then - echo "!!!!!!!!!! missing $file" -else - echo '---------- egrep Pass|Auth|PAM' $file - egrep 'Pass|Auth|PAM' $file -fi +pl-netinit-patch-resolv-conf /etc/resolv.conf -echo "END post pl_netinit forensics" +verbose-forensics "pl_netinit epilogue"