1 The Linux Watchdog driver API.
3 Copyright 2002 Christer Weingel <wingel@nano-system.com>
5 Some parts of this document are copied verbatim from the sbc60xxwdt
6 driver which is (c) Copyright 2000 Jakob Oestergaard <jakob@ostenfeld.dk>
8 This document describes the state of the Linux 2.4.18 kernel.
12 A Watchdog Timer (WDT) is a hardware circuit that can reset the
13 computer system in case of a software fault. You probably knew that
16 Usually a userspace daemon will notify the kernel watchdog driver via the
17 /dev/watchdog special device file that userspace is still alive, at
18 regular intervals. When such a notification occurs, the driver will
19 usually tell the hardware watchdog that everything is in order, and
20 that the watchdog should wait for yet another little while to reset
21 the system. If userspace fails (RAM error, kernel bug, whatever), the
22 notifications cease to occur, and the hardware watchdog will reset the
23 system (causing a reboot) after the timeout occurs.
25 The Linux watchdog API is a rather AD hoc construction and different
26 drivers implement different, and sometimes incompatible, parts of it.
27 This file is an attempt to document the existing usage and allow
28 future driver writers to use it as a reference.
32 All drivers support the basic mode of operation, where the watchdog
33 activates as soon as /dev/watchdog is opened and will reboot unless
34 the watchdog is pinged within a certain time, this time is called the
35 timeout or margin. The simplest way to ping the watchdog is to write
36 some data to the device. So a very simple watchdog daemon would look
42 int main(int argc, const char *argv[]) {
43 int fd=open("/dev/watchdog",O_WRONLY);
54 A more advanced driver could for example check that a HTTP server is
55 still responding before doing the write call to ping the watchdog.
57 When the device is closed, the watchdog is disabled. This is not
58 always such a good idea, since if there is a bug in the watchdog
59 daemon and it crashes the system will not reboot. Because of this,
60 some of the drivers support the configuration option "Disable watchdog
61 shutdown on close", CONFIG_WATCHDOG_NOWAYOUT. If it is set to Y when
62 compiling the kernel, there is no way of disabling the watchdog once
63 it has been started. So, if the watchdog dameon crashes, the system
64 will reboot after the timeout has passed.
66 Some other drivers will not disable the watchdog, unless a specific
67 magic character 'V' has been sent /dev/watchdog just before closing
68 the file. If the userspace daemon closes the file without sending
69 this special character, the driver will assume that the daemon (and
70 userspace in general) died, and will stop pinging the watchdog without
71 disabling it first. This will then cause a reboot.
75 All conforming drivers also support an ioctl API.
77 Pinging the watchdog using an ioctl:
79 All drivers that have an ioctl interface support at least one ioctl,
80 KEEPALIVE. This ioctl does exactly the same thing as a write to the
81 watchdog device, so the main loop in the above program could be
85 ioctl(fd, WDIOC_KEEPALIVE, 0);
89 the argument to the ioctl is ignored.
91 Setting and getting the timeout:
93 For some drivers it is possible to modify the watchdog timeout on the
94 fly with the SETTIMEOUT ioctl, those drivers have the WDIOF_SETTIMEOUT
95 flag set in their option field. The argument is an integer
96 representing the timeout in seconds. The driver returns the real
97 timeout used in the same variable, and this timeout might differ from
98 the requested one due to limitation of the hardware.
101 ioctl(fd, WDIOC_SETTIMEOUT, &timeout);
102 printf("The timeout was set to %d seconds\n", timeout);
104 This example might actually print "The timeout was set to 60 seconds"
105 if the device has a granularity of minutes for its timeout.
107 Starting with the Linux 2.4.18 kernel, it is possible to query the
108 current timeout using the GETTIMEOUT ioctl.
110 ioctl(fd, WDIOC_GETTIMEOUT, &timeout);
111 printf("The timeout was is %d seconds\n", timeout);
115 Some watchdog timers can be set to have a trigger go off before the
116 actual time they will reset the system. This can be done with an NMI,
117 interrupt, or other mechanism. This allows Linux to record useful
118 information (like panic information and kernel coredumps) before it
122 ioctl(fd, WDIOC_SETPRETIMEOUT, &pretimeout);
124 Note that the pretimeout is the number of seconds before the time
125 when the timeout will go off. It is not the number of seconds until
126 the pretimeout. So, for instance, if you set the timeout to 60 seconds
127 and the pretimeout to 10 seconds, the pretimout will go of in 50
128 seconds. Setting a pretimeout to zero disables it.
130 There is also a get function for getting the pretimeout:
132 ioctl(fd, WDIOC_GETPRETIMEOUT, &timeout);
133 printf("The pretimeout was is %d seconds\n", timeout);
135 Not all watchdog drivers will support a pretimeout.
137 Get the number of seconds before reboot:
139 Some watchdog drivers have the ability to report the remaining time
140 before the system will reboot. The WDIOC_GETTIMELEFT is the ioctl
141 that returns the number of seconds before reboot.
143 ioctl(fd, WDIOC_GETTIMELEFT, &timeleft);
144 printf("The timeout was is %d seconds\n", timeleft);
146 Environmental monitoring:
148 All watchdog drivers are required return more information about the system,
149 some do temperature, fan and power level monitoring, some can tell you
150 the reason for the last reboot of the system. The GETSUPPORT ioctl is
151 available to ask what the device can do:
153 struct watchdog_info ident;
154 ioctl(fd, WDIOC_GETSUPPORT, &ident);
156 the fields returned in the ident struct are:
158 identity a string identifying the watchdog driver
159 firmware_version the firmware version of the card if available
160 options a flags describing what the device supports
162 the options field can have the following bits set, and describes what
163 kind of information that the GET_STATUS and GET_BOOT_STATUS ioctls can
164 return. [FIXME -- Is this correct?]
166 WDIOF_OVERHEAT Reset due to CPU overheat
168 The machine was last rebooted by the watchdog because the thermal limit was
171 WDIOF_FANFAULT Fan failed
173 A system fan monitored by the watchdog card has failed
175 WDIOF_EXTERN1 External relay 1
177 External monitoring relay/source 1 was triggered. Controllers intended for
178 real world applications include external monitoring pins that will trigger
181 WDIOF_EXTERN2 External relay 2
183 External monitoring relay/source 2 was triggered
185 WDIOF_POWERUNDER Power bad/power fault
187 The machine is showing an undervoltage status
189 WDIOF_CARDRESET Card previously reset the CPU
191 The last reboot was caused by the watchdog card
193 WDIOF_POWEROVER Power over voltage
195 The machine is showing an overvoltage status. Note that if one level is
196 under and one over both bits will be set - this may seem odd but makes
199 WDIOF_KEEPALIVEPING Keep alive ping reply
201 The watchdog saw a keepalive ping since it was last queried.
203 WDIOF_SETTIMEOUT Can set/get the timeout
205 The watchdog can do pretimeouts.
207 WDIOF_PRETIMEOUT Pretimeout (in seconds), get/set
210 For those drivers that return any bits set in the option field, the
211 GETSTATUS and GETBOOTSTATUS ioctls can be used to ask for the current
212 status, and the status at the last reboot, respectively.
215 ioctl(fd, WDIOC_GETSTATUS, &flags);
219 ioctl(fd, WDIOC_GETBOOTSTATUS, &flags);
221 Note that not all devices support these two calls, and some only
222 support the GETBOOTSTATUS call.
224 Some drivers can measure the temperature using the GETTEMP ioctl. The
225 returned value is the temperature in degrees farenheit.
228 ioctl(fd, WDIOC_GETTEMP, &temperature);
230 Finally the SETOPTIONS ioctl can be used to control some aspects of
231 the cards operation; right now the pcwd driver is the only one
232 supporting thiss ioctl.
235 ioctl(fd, WDIOC_SETOPTIONS, options);
237 The following options are available:
239 WDIOS_DISABLECARD Turn off the watchdog timer
240 WDIOS_ENABLECARD Turn on the watchdog timer
241 WDIOS_TEMPPANIC Kernel panic on temperature trip
243 [FIXME -- better explanations]
245 Implementations in the current drivers in the kernel tree:
247 Here I have tried to summarize what the different drivers support and
248 where they do strange things compared to the other drivers.
250 acquirewdt.c -- Acquire Single Board Computer
252 This driver has a hardcoded timeout of 1 minute
254 Supports CONFIG_WATCHDOG_NOWAYOUT
256 GETSUPPORT returns KEEPALIVEPING. GETSTATUS will return 1 if
257 the device is open, 0 if not. [FIXME -- isn't this rather
258 silly? To be able to use the ioctl, the device must be open
259 and so GETSTATUS will always return 1].
261 advantechwdt.c -- Advantech Single Board Computer
263 Timeout that defaults to 60 seconds, supports SETTIMEOUT.
265 Supports CONFIG_WATCHDOG_NOWAYOUT
267 GETSUPPORT returns WDIOF_KEEPALIVEPING and WDIOF_SETTIMEOUT.
268 The GETSTATUS call returns if the device is open or not.
269 [FIXME -- silliness again?]
271 booke_wdt.c -- PowerPC BookE Watchdog Timer
273 Timeout default varies according to frequency, supports
276 Watchdog can not be turned off, CONFIG_WATCHDOG_NOWAYOUT
279 GETSUPPORT returns the watchdog_info struct, and
280 GETSTATUS returns the supported options. GETBOOTSTATUS
281 returns a 1 if the last reset was caused by the
282 watchdog and a 0 otherwise. This watchdog can not be
283 disabled once it has been started. The wdt_period kernel
284 parameter selects which bit of the time base changing
285 from 0->1 will trigger the watchdog exception. Changing
286 the timeout from the ioctl calls will change the
287 wdt_period as defined above. Finally if you would like to
288 replace the default Watchdog Handler you can implement the
289 WatchdogHandler() function in your own code.
291 eurotechwdt.c -- Eurotech CPU-1220/1410
293 The timeout can be set using the SETTIMEOUT ioctl and defaults
296 Also has a module parameter "ev", event type which controls
297 what should happen on a timeout, the string "int" or anything
298 else that causes a reboot. [FIXME -- better description]
300 Supports CONFIG_WATCHDOG_NOWAYOUT
302 GETSUPPORT returns CARDRESET and WDIOF_SETTIMEOUT but
303 GETSTATUS is not supported and GETBOOTSTATUS just returns 0.
305 i810-tco.c -- Intel 810 chipset
307 Also has support for a lot of other i8x0 stuff, but the
308 watchdog is one of the things.
310 The timeout is set using the module parameter "i810_margin",
311 which is in steps of 0.6 seconds where 2<i810_margin<64. The
312 driver supports the SETTIMEOUT ioctl.
314 Supports CONFIG_WATCHDOG_NOWAYOUT.
316 GETSUPPORT returns WDIOF_SETTIMEOUT. The GETSTATUS call
317 returns some kind of timer value which ist not compatible with
318 the other drivers. GETBOOT status returns some kind of
319 hardware specific boot status. [FIXME -- describe this]
321 ib700wdt.c -- IB700 Single Board Computer
323 Default timeout of 30 seconds and the timeout is settable
324 using the SETTIMEOUT ioctl. Note that only a few timeout
325 values are supported.
327 Supports CONFIG_WATCHDOG_NOWAYOUT
329 GETSUPPORT returns WDIOF_KEEPALIVEPING and WDIOF_SETTIMEOUT.
330 The GETSTATUS call returns if the device is open or not.
331 [FIXME -- silliness again?]
333 machzwd.c -- MachZ ZF-Logic
335 Hardcoded timeout of 10 seconds
337 Has a module parameter "action" that controls what happens
338 when the timeout runs out which can be 0 = RESET (default),
339 1 = SMI, 2 = NMI, 3 = SCI.
341 Supports CONFIG_WATCHDOG_NOWAYOUT and the magic character
344 GETSUPPORT returns WDIOF_KEEPALIVEPING, and the GETSTATUS call
345 returns if the device is open or not. [FIXME -- silliness
348 mixcomwd.c -- MixCom Watchdog
350 [FIXME -- I'm unable to tell what the timeout is]
352 Supports CONFIG_WATCHDOG_NOWAYOUT
354 GETSUPPORT returns WDIOF_KEEPALIVEPING, GETSTATUS returns if
355 the device is opened or not [FIXME -- I'm not really sure how
356 this works, there seems to be some magic connected to
357 CONFIG_WATCHDOG_NOWAYOUT]
359 pcwd.c -- Berkshire PC Watchdog
361 Hardcoded timeout of 1.5 seconds
363 Supports CONFIG_WATCHDOG_NOWAYOUT
365 GETSUPPORT returns WDIOF_OVERHEAT|WDIOF_CARDRESET and both
366 GETSTATUS and GETBOOTSTATUS return something useful.
368 The SETOPTIONS call can be used to enable and disable the card
369 and to ask the driver to call panic if the system overheats.
371 sbc60xxwdt.c -- 60xx Single Board Computer
373 Hardcoded timeout of 10 seconds
375 Does not support CONFIG_WATCHDOG_NOWAYOUT, but has the magic
376 character 'V' close handling.
378 No bits set in GETSUPPORT
380 scx200.c -- National SCx200 CPUs
382 Not in the kernel yet.
384 The timeout is set using a module parameter "margin" which
385 defaults to 60 seconds. The timeout can also be set using
386 SETTIMEOUT and read using GETTIMEOUT.
388 Supports a module parameter "nowayout" that is initialized
389 with the value of CONFIG_WATCHDOG_NOWAYOUT. Also supports the
390 magic character 'V' handling.
392 shwdt.c -- SuperH 3/4 processors
394 [FIXME -- I'm unable to tell what the timeout is]
396 Supports CONFIG_WATCHDOG_NOWAYOUT
398 GETSUPPORT returns WDIOF_KEEPALIVEPING, and the GETSTATUS call
399 returns if the device is open or not. [FIXME -- silliness
402 softdog.c -- Software watchdog
404 The timeout is set with the module parameter "soft_margin"
405 which defaults to 60 seconds, the timeout is also settable
406 using the SETTIMEOUT ioctl.
408 Supports CONFIG_WATCHDOG_NOWAYOUT
410 WDIOF_SETTIMEOUT bit set in GETSUPPORT
412 w83877f_wdt.c -- W83877F Computer
414 Hardcoded timeout of 30 seconds
416 Does not support CONFIG_WATCHDOG_NOWAYOUT, but has the magic
417 character 'V' close handling.
419 No bits set in GETSUPPORT
421 w83627hf_wdt.c -- w83627hf watchdog
423 Timeout that defaults to 60 seconds, supports SETTIMEOUT.
425 Supports CONFIG_WATCHDOG_NOWAYOUT
427 GETSUPPORT returns WDIOF_KEEPALIVEPING and WDIOF_SETTIMEOUT.
428 The GETSTATUS call returns if the device is open or not.
430 wdt.c -- ICS WDT500/501 ISA and
431 wdt_pci.c -- ICS WDT500/501 PCI
433 Default timeout of 60 seconds. The timeout is also settable
434 using the SETTIMEOUT ioctl.
436 Supports CONFIG_WATCHDOG_NOWAYOUT
438 GETSUPPORT returns with bits set depending on the actual
439 card. The WDT501 supports a lot of external monitoring, the
442 wdt285.c -- Footbridge watchdog
444 The timeout is set with the module parameter "soft_margin"
445 which defaults to 60 seconds. The timeout is also settable
446 using the SETTIMEOUT ioctl.
448 Does not support CONFIG_WATCHDOG_NOWAYOUT
450 WDIOF_SETTIMEOUT bit set in GETSUPPORT
452 wdt977.c -- Netwinder W83977AF chip
454 Hardcoded timeout of 3 minutes
456 Supports CONFIG_WATCHDOG_NOWAYOUT
458 Does not support any ioctls at all.