wdogctl
—
Watchdog timer control utility
wdogctl |
-e [-A ]
[-p seconds]
timer |
wdogctl |
-k [-A ]
[-p seconds]
timer |
wdogctl |
-u [-A ]
[-p seconds]
timer |
wdogctl |
-x [-A ]
[-p seconds]
timer |
wdogctl
is used to manipulate watchdog timers. Watchdog
timers provide a means of ensuring that a system continues to make progress.
This is accomplished by use of a timer, provided by either hardware or
software; when the timer expires, the watchdog resets the system. In this case
of a hardware watchdog timer, this is accomplished by asserting the system's
hardware reset signal. In the case of a software watchdog timer, this is
accomplished by calling the kernel's normal reboot path. In order to prevent
the system from rebooting, something must refresh the timer to prevent it from
expiring.
The NetBSD kernel provides three basic
modes in which watchdog timers may operate: kernel tickle mode, user tickle
mode, and external tickle mode. In kernel tickle mode, a timer in the kernel
refreshes the watchdog timer. In user tickle mode,
wdogctl
runs in the background and refreshes the
watchdog timer. In kernel tickle mode, progress of the kernel is ensured. In
user tickle mode, the ability for user programs to run within a known period
of time is ensured. Note that user tickle mode must be used with caution; on
a heavily loaded system, the timer may expire accidentally, even though user
programs may be making (very slow) progress. A user-mode timer is disarmed
(if possible) when the device is closed, unless the timer is activated with
the -x
option.
External-mode watchdogs are similar to user-mode watchdogs, except
that the tickle must be done explicitly by a separate invocation of the
program with the -t
option.
In the first two modes, an attempt is made to refresh the watchdog
timer in one half the timer's configured period. That is, if the watchdog
timer has a period of 30 seconds, a refresh attempt is made every 15
seconds.
If called without arguments, wdogctl
will
list the timers available on the system. When arming a watchdog timer, the
timer argument is the name of the timer to arm.
Only one timer may be armed at a time; if an attempt is made to
arm a timer when one is already armed, an error message will be displayed
and no action will be taken.
The options are as follows:
-A
- When arming a timer, this flag indicates that an audible alarm is to sound
when the watchdog timer expires and resets the system. If the selected
timer does not support an audible alarm, this option will be silently
ignored.
-d
- This flag disarms the currently active timer. Note that not all watchdog
timers can be disabled once armed. If the selected timer can not be
disabled, an error message will be displayed and the timer will remain
armed.
-e
- Arm timer in external tickle mode.
-k
- Arm timer in kernel tickle mode.
-p
period
- When arming a timer, this flag configures the timer period to
period seconds. If the specified period is outside
the timer's range, an error message will be displayed and no action will
be taken.
-t
- This flag tickles an external mode timer.
-u
- Arm timer in user tickle mode.
-x
- Arm timer in a modified user tickle mode: closing
the device will not disarm the timer.
/dev/watchdog -- the system monitor watchdog timer
device
acpiwdrt(4),
evbarm/iopwdog(4),
i386/elansc(4),
i386/gcscpcib(4),
i386/geodewdog(4),
ipmi(4),
itesio(4),
pcweasel(4),
pwdog(4),
swwdog(4),
x86/ichlpcib(4)
The wdogctl
command first appeared in
NetBSD 1.6.
The wdogctl
command and the
NetBSD watchdog timer framework were written by
Jason R. Thorpe ⟨thorpej@zembu.com⟩, and
contributed by Zembu Labs, Inc.