The Linux Watchdog driver API. Copyright 2002 Christer Weingel Some parts of this document are copied verbatim from the sbc60xxwdt driver which is (c) Copyright 2000 Jakob Oestergaard This document describes the state of the Linux 2.4.18 kernel. Introduction: A Watchdog Timer (WDT) is a hardware circuit that can reset the computer system in case of a software fault. You probably knew that already. Usually a userspace daemon will notify the kernel watchdog driver via the /dev/watchdog special device file that userspace is still alive, at regular intervals. When such a notification occurs, the driver will usually tell the hardware watchdog that everything is in order, and that the watchdog should wait for yet another little while to reset the system. If userspace fails (RAM error, kernel bug, whatever), the notifications cease to occur, and the hardware watchdog will reset the system (causing a reboot) after the timeout occurs. The Linux watchdog API is a rather AD hoc construction and different drivers implement different, and sometimes incompatible, parts of it. This file is an attempt to document the existing usage and allow future driver writers to use it as a reference. The simplest API: All drivers support the basic mode of operation, where the watchdog activates as soon as /dev/watchdog is opened and will reboot unless the watchdog is pinged within a certain time, this time is called the timeout or margin. The simplest way to ping the watchdog is to write some data to the device. So a very simple watchdog daemon would look like this: int main(int argc, const char *argv[]) { int fd=open("/dev/watchdog",O_WRONLY); if (fd==-1) { perror("watchdog"); exit(1); } while(1) { write(fd, "\0", 1); sleep(10); } } A more advanced driver could for example check that a HTTP server is still responding before doing the write call to ping the watchdog. When the device is closed, the watchdog is disabled. This is not always such a good idea, since if there is a bug in the watchdog daemon and it crashes the system will not reboot. Because of this, some of the drivers support the configuration option "Disable watchdog shutdown on close", CONFIG_WATCHDOG_NOWAYOUT. If it is set to Y when compiling the kernel, there is no way of disabling the watchdog once it has been started. So, if the watchdog dameon crashes, the system will reboot after the timeout has passed. Some other drivers will not disable the watchdog, unless a specific magic character 'V' has been sent /dev/watchdog just before closing the file. If the userspace daemon closes the file without sending this special character, the driver will assume that the daemon (and userspace in general) died, and will stop pinging the watchdog without disabling it first. This will then cause a reboot. The ioctl API: All conforming drivers also support an ioctl API. Pinging the watchdog using an ioctl: All drivers that have an ioctl interface support at least one ioctl, KEEPALIVE. This ioctl does exactly the same thing as a write to the watchdog device, so the main loop in the above program could be replaced with: while (1) { ioctl(fd, WDIOC_KEEPALIVE, 0); sleep(10); } the argument to the ioctl is ignored. Setting and getting the timeout: For some drivers it is possible to modify the watchdog timeout on the fly with the SETTIMEOUT ioctl, those drivers have the WDIOF_SETTIMEOUT flag set in their option field. The argument is an integer representing the timeout in seconds. The driver returns the real timeout used in the same variable, and this timeout might differ from the requested one due to limitation of the hardware. int timeout = 45; ioctl(fd, WDIOC_SETTIMEOUT, &timeout); printf("The timeout was set to %d seconds\n", timeout); This example might actually print "The timeout was set to 60 seconds" if the device has a granularity of minutes for its timeout. Starting with the Linux 2.4.18 kernel, it is possible to query the current timeout using the GETTIMEOUT ioctl. ioctl(fd, WDIOC_GETTIMEOUT, &timeout); printf("The timeout was is %d seconds\n", timeout); Envinronmental monitoring: All watchdog drivers are required return more information about the system, some do temperature, fan and power level monitoring, some can tell you the reason for the last reboot of the system. The GETSUPPORT ioctl is available to ask what the device can do: struct watchdog_info ident; ioctl(fd, WDIOC_GETSUPPORT, &ident); the fields returned in the ident struct are: identity a string identifying the watchdog driver firmware_version the firmware version of the card if available options a flags describing what the device supports the options field can have the following bits set, and describes what kind of information that the GET_STATUS and GET_BOOT_STATUS ioctls can return. [FIXME -- Is this correct?] WDIOF_OVERHEAT Reset due to CPU overheat The machine was last rebooted by the watchdog because the thermal limit was exceeded WDIOF_FANFAULT Fan failed A system fan monitored by the watchdog card has failed WDIOF_EXTERN1 External relay 1 External monitoring relay/source 1 was triggered. Controllers intended for real world applications include external monitoring pins that will trigger a reset. WDIOF_EXTERN2 External relay 2 External monitoring relay/source 2 was triggered WDIOF_POWERUNDER Power bad/power fault The machine is showing an undervoltage status WDIOF_CARDRESET Card previously reset the CPU The last reboot was caused by the watchdog card WDIOF_POWEROVER Power over voltage The machine is showing an overvoltage status. Note that if one level is under and one over both bits will be set - this may seem odd but makes sense. WDIOF_KEEPALIVEPING Keep alive ping reply The watchdog saw a keepalive ping since it was last queried. WDIOF_SETTIMEOUT Can set/get the timeout For those drivers that return any bits set in the option field, the GETSTATUS and GETBOOTSTATUS ioctls can be used to ask for the current status, and the status at the last reboot, respectively. int flags; ioctl(fd, WDIOC_GETSTATUS, &flags); or ioctl(fd, WDIOC_GETBOOTSTATUS, &flags); Note that not all devices support these two calls, and some only support the GETBOOTSTATUS call. Some drivers can measure the temperature using the GETTEMP ioctl. The returned value is the temperature in degrees farenheit. int temperature; ioctl(fd, WDIOC_GETTEMP, &temperature); Finally the SETOPTIONS ioctl can be used to control some aspects of the cards operation; right now the pcwd driver is the only one supporting thiss ioctl. int options = 0; ioctl(fd, WDIOC_SETOPTIONS, options); The following options are available: WDIOS_DISABLECARD Turn off the watchdog timer WDIOS_ENABLECARD Turn on the watchdog timer WDIOS_TEMPPANIC Kernel panic on temperature trip [FIXME -- better explanations] Implementations in the current drivers in the kernel tree: Here I have tried to summarize what the different drivers support and where they do strange things compared to the other drivers. acquirewdt.c -- Acquire Single Board Computer This driver has a hardcoded timeout of 1 minute Supports CONFIG_WATCHDOG_NOWAYOUT GETSUPPORT returns KEEPALIVEPING. GETSTATUS will return 1 if the device is open, 0 if not. [FIXME -- isn't this rather silly? To be able to use the ioctl, the device must be open and so GETSTATUS will always return 1]. advantechwdt.c -- Advantech Single Board Computer Timeout that defaults to 60 seconds, supports SETTIMEOUT. Supports CONFIG_WATCHDOG_NOWAYOUT GETSUPPORT returns WDIOF_KEEPALIVEPING and WDIOF_SETTIMEOUT. The GETSTATUS call returns if the device is open or not. [FIXME -- silliness again?] eurotechwdt.c -- Eurotech CPU-1220/1410 The timeout can be set using the SETTIMEOUT ioctl and defaults to 60 seconds. Also has a module parameter "ev", event type which controls what should happen on a timeout, the string "int" or anything else that causes a reboot. [FIXME -- better description] Supports CONFIG_WATCHDOG_NOWAYOUT GETSUPPORT returns CARDRESET and WDIOF_SETTIMEOUT but GETSTATUS is not supported and GETBOOTSTATUS just returns 0. i810-tco.c -- Intel 810 chipset Also has support for a lot of other i8x0 stuff, but the watchdog is one of the things. The timeout is set using the module parameter "i810_margin", which is in steps of 0.6 seconds where 2