Difference between revisions of "AVR watch-dog daemon for Linkstation"

Latest revision as of 15:42, 18 January 2013

WARNING! There is a possibility that you could brick your NAS with these instructions. Please make sure that you read the entire page carefully. The daemon discussed in this article is essential to the LinkStation's operation. If its keep-alive messages are not sent, the LS' watchdog timer will trigger a shutdown.

Description

avr_evtd is an operating system daemon and should be started from /etc/init.d or /usr/local/sbin. It will return immediately, so you don't need to start it with '&'. It can replace the propritary ppc_uartd (LS1) and mc_ctld (LS2) daemons.

avr_evtd is a simple and small user space interface to the Linkstation AVR micro-controller (note: technically, the LS2 doesn't contain an AVR, but avr_evtd, works with the LS2, too). It doesn't have a lot of special features, but it's main task is to provide 'keep-alive' messages to the Linkstation's on-board AVR device. This device controls/monitors the fan, various LEDs, timed power up and two buttons. This daemon provides the necessary initialisation to the device and also stimulates the LEDs depending on various fault conditions. It also monitors a power button (located at the front) and a reset button (located at the rear).

avr_evtd searches for a configuration file located within the /etc/melco at start-up time. If this file does not exist, then it reverts to reading the file located at /etc/default/avr_evtd. Additionally, it periodically checks the root /dev/hda1 partition and the user data partition /dev/hda3 to ensure they are mounted. Also, if requested, it will determine if sufficient space remains and if not then the AVR is requested to illuminate the DISK LED. This periodic checking also takes place on the configuration files. If they are deemed to have been updated, then the daemon will respond accordingly.

Any failures are normally routed through the log files. With timed shutdown/power up, a warning will be broadcast to all users (console message) when within 5 or less minutes of power off. Also, fan failure alerts will be broadcast in this fashion. Failure to determine mounted partition /dev/hda3 will result in the DIAG LED flashing three times, repeatedly.

A new feature of this daemon is the ability to code events for single or groups of days. This allows the user to add any number of power-on/off events as required. This also has the added benifit of being able to shut down the device for longer periods. The internal AVR timer has a 12-bit resolution timer which can power up the linkstation from a maximum of a sixty eight hour sleep: from time of invocation. The AVR is updated again at time of power down/shutdown to re-validate the timer, in case of time updates (either by user or ntp). This will also preserve the 68 hour sleep resolution.

Install

Get the latest version from https://sourceforge.net/projects/ppc-evtd

 cd / 
 wget <link to latest version from sourcecforge site>
 tar -zxvf <filename from wget download>

• build the source and install process daemon and configuration files

 cd /usr/src/avr_evtd
 make install avr_evtd

• alternatively, if you do not have gcc installed

 cd /usr/src/avr_evtd
 make -n install avr_evtd

• modify /etc/default/avr_evtd.sample to suit your needs (refer to below for more information) and save as avr_evtd.
• remove current daemon ppc_uartd (or mc_ctld for MIPS) if applicable

FreeLink

For those with ppc_uartd (LS1, HG, HS, KURO)

 update-rc.d -f ppc_uartd remove

or for the MIPSel LS2

 update-rc.d -f mc_ctld remove

To manually remove

 rm /etc/rc2.d/S95ppc_uartd
 rm /etc/rc0.d/K95ppc_uartd

or (MIPSel)

 rm /etc/rc2.d/mc_ctld
 rm /etc/rc0.d/mc_ctld

OpenLink PPC

 rm /etc/rc.d/rc2.d/S95ppc_uartd
 rm /etc/rc.d/rc0.d/K95ppc_uartd

• add the new daemon links

FreeLink

 update-rc.d avr_evtd start 12 2 . stop 95 0 6 .

Manually

 cd /etc/rc2.d
 ln -s ../init.d/avr_evtd S12avr_evtd
 cd /etc/init.d/rc0.d
 ln -s ../init.d/avr_evtd K95avr_evtd
 cd /etc/rc6.d
 ln -s ../init.d/avr_evtd K95avr_evtd

OpenLink

 cd /etc/rc.d/rc2.d
 ln -s ../init.d/avr_evtd S12avr_evtd
 cd /etc/init.d/rc0.d
 ln -s ../init.d/avr_evtd K95avr_evtd
 cd /etc/rc.d/rc6.d
 ln -s ../init.d/avr_evtd K95avr_evtd

• check the daemon is running

 ps aux | grep avr_evtd

• observe /usr/local/sbin/avr_evtd in the output line

Setup

/etc/default/avr_evtd configuration file is the fallback file in the event that the stock melco files do not exist. The file is read at initial start-up in order to determine if timed shutdown is required and if the disk usage is to be monitored. The file should be always located within the /etc/default directory and a sample file is provided. The file format is similar to other Unix configuration files - comments begin with a # character and extend to the end of the line; blank lines are ignored. Configuration commands consist of an initial keyword followed by an argument. Arguments may be strings or times written in HH:MM (UTC) format. Optional arguments are delimited by [ ] in the following descriptions, while alternatives are separated by |.

This is an example configuration file:

# Sample avr_evtd daemon configuration file
# PLEASE EDIT THIS FILE TO SUIT
LOG=/var/log
DEBUG=ON
# Set to YES to enable the EM-Mode feature, default no
EMMODE=YES
#
# Timed shutdown flag (ON enables timed power down/up only
# or use days of the week MONWEDSAT etc
# if SHUTDOWN and POWERON are also specified)
TIMER=ON
# To override the device scan, specify the desired
# serial port connection to the AVR below and remove the
# comment
#DEVICE=/dev/ttyS1
# MACRO day/group switching ON/OFF times in 24hr HH:MM
MON=ON=08:00,OFF=21:00,ON=23:30
TUE=OFF=02:30,ON=09:00,OFF=21:45
WED-FRI=ON=08:00,OFF=22:00
SAT=ON=07:30,OFF=08:30
# Shutdown (default) time specified in 24hr format HH:MM
SHUTDOWN=
# Power on (default) time specified in 24hr format HH:MM
POWERON=
# Disk check, set to OFF or a value (0-100) specifying a
# percentage used at which the disk LED will illuminate
DISKCHECK=90
# Specifies the root partition to check, no default
ROOT=hda1
# Specifies the working partition to check, no default
WORK=hda3
# Disk/AVR refresh rate (seconds), default 40
REFRESH=40
# Hold time (seconds) for button power-off, default 3
HOLD=3
# Enable/disable continous disk full messages, default off
DISKNAG=OFF
# Fan stationary fault timer (seconds), default 30
FANSTOP=30

Meaning:

LOG=/var/log

Specifies the location of the debug log files. Note, this size of this file is not restricited so if enabled, some logrotate mechanism maybe desirable.

DEBUG=ON

Debug data will be generated. This can be disabled by setting this entry to NO.

EMMODE=YES

Enables the selection of EM-Mode from the reset button (hold for > 20 seconds). To disable, set this entry to NO. If this is changed whilst the daemon is up, then it will be ignored. This is only observed at startup (to protect against accidental enable) and the daemon will need to be restarted (either by reboot or from /etc/init.d/avr_evtd restart) if the user requires instant on. When enabled, a ps aux will show a -e in the avr_evtd command line.

TIMER=YES

Timed shutdown/power on is enabled. To disable, set this to NO. This will turn-off the sleep mode option of the box.

The following are macro'd power switching events. These can be removed if this is not required. Shown here as an example only.

TUE=OFF=02:30,ON=09:00,OFF=21:45
WED-FRI=ON=08:00,OFF=22:00
SAT=ON=07:30,OFF=08:30

• For monday, the box will power on at 08:00, off at 21:00 and on again at 23:30
• Tuesday, the box will power off at 02:30, on at 09:00 and off at 21:45
• Wednesday throught friday, the box will power on at 08:00 and off at 22:00
• Saturday, the box will power on at 07:30 and off at 08:30
• Sunday, the box will not power up

SHUTDOWN=
POWERON=

No default shutdown or power on time is defined. If macro'd timing is not required and then power switching can be specified through these options only. Alternatively, if power-on of 07:30 and off at 23:30 is all that is required then it can be specified through the macro event handler via:

• SUN-MON=ON=07:30,OFF=23:30

NOTE:The macro list of days must be in the correct order for the scheduler to work correctly:

• This order is SUN, MON, TUE, WED, THR, FRI and SAT.

DISKCHECK=90

The disk led will illuminate if either /dev/hda1 or /dev/hda3 is above 90% used. This can be disabled by setting to 0 (zero) or another value from 1-100%. When the disk usage exceeds this value, then an event is triggered through the EventScript which can be used to additional clean-up if required.

ROOT=hda1

This allows the user to specify what the root partition is for the DISKCHECK process. There is no default partition for this so setting to 'blank' means no checking of the root partition is performed. Also note, no syntax or validation is performed on this entry.

WORK=hda3

This allows the user to specify what the working partition is for the DISKCHECK process. There is no default partition for this so setting to 'blank' means no checking of the work partition is performed. Also note, no syntax or validation is performed on this entry.

REFRESH=40

The avr refresh rate is set to 40 seconds. This is the rate at which the daemon pings the micro-controller and checks on disk usage. This can be pushed out to 300 seconds if less frequent pinging is required.

HOLD=3

The button hold time for power off is set to 3 seconds. Can be changed from 3 to 10 seconds. This is the time the power button has to be depressed for before the daemon issues the shutdown event.

DISKNAG=OFF

Continous disk event messages on disk full is disabled. If repeated messages are required all the time the disk is full, then set this to ON.

FANSTOP=30

A fan event will be generated if stationary for more than 30 seconds. This can be changed from 1 to 60 seconds or to OFF. For HG uses a lower time maybe more desirable as these units run hotter.

When set to OFF, ensure that there is enough ventilation to allow the heat build-up to escape. This simply sets the daemon to ignore fan failure messages.

Files

The following files are used by the avr_evtd daemon:

• /etc/default/avr_evtd
• /etc/avr_evtd/EventScript
• /etc/avr_evtd/emergency_eth0
• $LOG/avr_evtd.log (default is /var/log/avr_evtd/log)

The /etc/default/avr_evtd is the configuration file, as descibed in the setup.

The next file, EventScript, is what is invoked by the daemon on an event. This being a button and/or action event request. This is described below and is scripted in this fashion to allow users to perform any other desired action following the event.

The next file, emergency_eth0, is used to configure the ethernet device if the reset button is double pressed. This establishes an EM IP of 192.168.11.150 and also requests a DHCP server.

The log file, whose directory is specified by the LOG entry in the configuration file and is enabled by setting DEBUG=ON records all the events.

Event Handling

Any event, such as button or fault, is handled by the daemon and some are made available externally through /etc/avr_evtd/EventScript. A default example of this is shown below:

DEVICE=$2

#
# Populate the configured settings
#
[ -f /etc/default/avr_evtd ] && . /etc/default/avr_evtd /dev/null 2>&1

if [ "$DEBUG" = "ON" ]; then
  echo "`date` command is $1 for $DEVICE" >> /etc/avr_evtd/events.log
fi

case "$1" in
  0)  if [ -e /sbin/utelnetd ]; then
        /sbin/utelnetd -p1234 -l /bin/bash &
      fi ;;
  1)  echo -n "[avr_evtd]: AVR failure dectected!"
      # Perform halt
      /sbin/shutdown -h now ;;
  2)  echo -n "[avr_evtd]: Shutdown request"
      # Perform shutdown request
      /sbin/shutdown -h now ;;
  3)  echo -n "[avr_evtd]: Power Button Up"
      ;;
  4)  echo -n "[avr_evtd]: Power Button Down"
      ;;
  5)  echo -n "[avr_evtd]: Reset Button Up"
      ;;
  6)  echo -n "[avr_evtd]: Reset Button Down"
      ;;
  7)  echo -n "[avr_evtd]: Timed shutdown"
      echo -n "EEEE" > $DEVICE
      # Perform shutdown request
      /sbin/shutdown -h now ;;
  8)  echo -n "[avr_evtd]: AVR demanded reset"
      echo -n "CCCC" > $DEVICE
      # Perform reboot
      /sbin/reboot ;;
  9)  echo -n "[avr_evtd]: Disk used > $DISKCHECK%"
      ;;
  E)  echo -n "[avr_evtd]: EM mode selected"
      # Determine flash block to use
      if [ -f /proc/mtd ]; then
        MTD=`cat /proc/mtd | grep mtd_status`
        BLOCK_ID=${MTD:3:1}
        FLASH=/dev/mtdblock$BLOCK_ID
      else
        FLASH=/dev/fl3
      fi

      # Flash device available?
      if [ -n "$FLASH" ]; then
        DUMP=`cat $FLASH`
        STATE=${DUMP:2:4}
        # Check state is currently NORMAL
        if [ "$STATE" = "OKOK" ]; then
          echo -n "NGNGNG" > $FLASH
        fi
      fi
      echo -n "CCCC" > $DEVICE
      /sbin/reboot ;;
  F)  echo -n "[avr_evtd]: Fan failure detected"
      # Illuminate relevant LED and wait for AVR halt message
      echo -n "iiii" > $DEVICE ;;
  *)  exit 1
      ;;
esac

exit 0

Troubleshooting

EventScript doesn't run?

If EventScript fails to run, then it could be because /bin/sh isn't symlinked to /bin/bash. The script uses ${param:start:offset} notation for a 'substring' operation which /bin/sh doesn't understand (on mflint's Ubuntu Kurobox, anyway).

To fix this, either make a symlink from /bin/sh to /bin/bash, or (preferred) change #!/bin/sh to #!/bin/bash at the top of EventScript.