Fix #26: add support for multiple watchdog devices

Signed-off-by: Joachim Wiberg <[email protected]>

man/watchdogd.conf.5

@@ -23,10 +23,11 @@
 The default
 .Xr watchdogd 8
 use-case does not require a configuration file.  However, enabling a
-health monitor plugin or the process supervisor is done using
+health monitor plugin, the process supervisor, or multiple watchdog
+device nodes, is done using
 .Pa /etc/watchdogd.conf .
 .Pp
-Available monitor plugins are:
+Available health monitor plugins:
 .Bl -tag -width supervisor
 .It Cm supervisor
 Process supervisor, monitor the heartbeat of processes
@@ -78,6 +79,24 @@ script.sh {filenr, loadavg, meminfo} {crit, warn} VALUE
 .Ed
 .Pp
 Health monitor plugins also have their own local script setting.
+.It Cm device Ar /path/to/device Ar {}
+.Nm watchdogd
+supports kicking multiple watchdog devices.  By default, and with no
+command line arguemts,
+.Pa /dev/watchdog
+is used.  If that is your system, this section is not necessary.  This
+section is only useful if you want everything in the configuration file
+or have multiple watchdog devices.  See
+.Sx EXAMPLE
+section below.
+.Bl -tag -width TERM
+.It Cm timeout = Ar SEC
+Same as global option.
+.It Cm interval = Ar SEC
+Same as global option.
+.It Cm safe-exit = Ar true | false
+Same as global option.
+.El
 .It Cm reset-reason Ar {}
 This section controls the reset reason, including the reset counter.  By
 default this is disabled, since not all systems allow writing to disk,
@@ -429,17 +448,36 @@ active.  The format is not guaranteed to be stable between releases, but
 will most likely be anyway.  See
 .Ql /run/watchdogd/tempmon.json .
 .Sh EXAMPLE
-.Bd -unfilled -offset indent
+.Bd -unfilled
 ### /etc/watchdogd.conf
+
+### Watchdogs ##########################################################
+# Global settings that can be overridden per watchdog
 timeout   = 20
 interval  = 10
 safe-exit = false
 
+# Multiple watchdogs can be kicked, the default, even if no .conf file
+# is found or device node given on the command line, is /dev/watchdog
+device /dev/watchdog {
+    timeout    = 20
+    interval   = 10
+    safe-exit  = false
+}
+
+#device /dev/watchdog2 {
+#    timeout    = 20
+#    interval   = 10
+#    safe-exit  = false
+#}
+
+### Process Supervisor #################################################
 supervisor {
     enabled  = true
     priority = 98
 }
 
+### Reset Reason #######################################################
 reset-reason {
     enabled = true
     file    = "/var/lib/misc/watchdogd.state"

src/conf.c

@@ -95,6 +95,26 @@ static int generic_checker(uev_ctx_t *ctx, cfg_t *cfg)
 }
 #endif
 
+static int wdt_checker(uev_ctx_t *ctx, cfg_t *cfg, const char *sect)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < cfg_size(cfg, sect); i++) {
+		cfg_t *sec = cfg_getnsec(cfg, sect, i);
+		const char *name = cfg_title(sec);
+		int interval, timeout, safe;
+
+		interval = cfg_getint(sec, "interval");
+		timeout  = cfg_getint(sec, "timeout");
+		safe     = cfg_getbool(sec, "safe-exit");
+
+		rc += wdt_add(name, interval, timeout, safe, 0);
+	}
+
+	return rc;
+}
+
 static int validate_reset_reason(uev_ctx_t *ctx, cfg_t *cfg)
 {
 	if (!cfg)
@@ -189,6 +209,12 @@ static void conf_errfunc(cfg_t *cfg, const char *format, va_list args)
 
 int conf_parse_file(uev_ctx_t *ctx, char *file)
 {
+	cfg_opt_t device_opts[] =  {
+		CFG_INT ("interval",    WDT_KICK_DEFAULT, CFGF_NONE),
+		CFG_INT ("timeout",     WDT_TIMEOUT_DEFAULT, CFGF_NONE),
+		CFG_BOOL("safe-exit",   cfg_false, CFGF_NONE),
+		CFG_END()
+	};
 	cfg_opt_t supervisor_opts[] =  {
 		CFG_BOOL("enabled",  cfg_false, CFGF_NONE),
 		CFG_INT ("priority", 0, CFGF_NONE),
@@ -224,6 +250,7 @@ int conf_parse_file(uev_ctx_t *ctx, char *file)
 		CFG_INT ("interval",    WDT_KICK_DEFAULT, CFGF_NONE),
 		CFG_INT ("timeout",     WDT_TIMEOUT_DEFAULT, CFGF_NONE),
 		CFG_BOOL("safe-exit",   cfg_false, CFGF_NONE),
+		CFG_SEC ("device",      device_opts, CFGF_MULTI | CFGF_TITLE),
 		CFG_SEC ("supervisor",  supervisor_opts, CFGF_NONE),
 		CFG_SEC ("reset-cause", reset_reason_opts, CFGF_NONE), /* Compat only */
 		CFG_SEC ("reset-reason", reset_reason_opts, CFGF_NONE),
@@ -294,6 +321,10 @@ int conf_parse_file(uev_ctx_t *ctx, char *file)
 		opt = cfg_getnsec(cfg, "reset-cause", 0); /* Compat only */
 	validate_reset_reason(ctx, opt);
 
+	wdt_mark();
+	wdt_checker(ctx, cfg, "device");
+	wdt_sweep();
+
 #ifdef FILENR_PLUGIN
 	checker(ctx, cfg, "filenr", filenr_init);
 #endif

src/watchdogd.c

@@ -62,11 +62,11 @@ static void exit_cb(uev_t *w, void *arg, int events)
 {
 	DEBUG("Got signal %d, rebooting:%d ...", w->signo, rebooting);
 	if (rebooting) {
-		wdt_exit(w->ctx);
+		wdt_reboot(w->ctx);
 		return;
 	}
 
-	wdt_close(w->ctx);
+	wdt_exit(w->ctx);
 }
 
 static void reboot_cb(uev_t *w, void *arg, int events)
@@ -75,7 +75,7 @@ static void reboot_cb(uev_t *w, void *arg, int events)
 
 	DEBUG("Got signal %d, rebooting:%d ...", w->signo, rebooting);
 	if (rebooting) {
-		wdt_exit(w->ctx);
+		wdt_reboot(w->ctx);
 		return;
 	}