sysmobts: Add "normal" actions to execute

Instead of keeping state to remember what was done and needs
to be undone this patch introduces actions that will be executed
when the system is back to normal.

By design the system is considered to be in the normal state
and these actions will be only executed after the system is
coming back to the normal state.

One advantage of this scheme is that an operator can decide
that an overheated systems hould be off duty and requires manual
interaction to be allowed back in service.

The change has only been smoke tested

Fixes: SYS#833

3 files changed, 113 insertions, 10 deletions

diff --git a/src/osmo-bts-sysmo/misc/sysmobts_mgr.h b/src/osmo-bts-sysmo/misc/sysmobts_mgr.h
index f4058674..ac760835 100644
--- a/src/osmo-bts-sysmo/misc/sysmobts_mgr.h
+++ b/src/osmo-bts-sysmo/misc/sysmobts_mgr.h
@@ -19,6 +19,15 @@ enum {
 	TEMP_ACT_BTS_SRV_OFF	=	0x10,
 };
 
+/* actions only for normal state */
+enum {
+	TEMP_ACT_NORM_PW_CONTRL	=	0x1,
+	TEMP_ACT_NORM_MASTER_ON	=	0x2,
+	TEMP_ACT_NORM_SLAVE_ON	=	0x4,
+	TEMP_ACT_NORM_PA_ON	=	0x8,
+	TEMP_ACT_NORM_BTS_SRV_ON=	0x10,
+};
+
 enum sysmobts_temp_state {
 	STATE_NORMAL,		/* Everything is fine */
 	STATE_WARNING_HYST,	/* Go back to normal next? */
@@ -39,6 +48,7 @@ struct sysmobts_temp_limit {
 enum mgr_vty_node {
 	MGR_NODE = _LAST_OSMOVTY_NODE + 1,
 
+	ACT_NORM_NODE,
 	ACT_WARN_NODE,
 	ACT_CRIT_NODE,
 	LIMIT_RF_NODE,
@@ -57,6 +67,7 @@ struct sysmobts_mgr_instance {
 	struct sysmobts_temp_limit board_limit;
 	struct sysmobts_temp_limit pa_limit;
 
+	int action_norm;
 	int action_warn;
 	int action_crit;
 
diff --git a/src/osmo-bts-sysmo/misc/sysmobts_mgr_temp.c b/src/osmo-bts-sysmo/misc/sysmobts_mgr_temp.c
index c757f63a..39b20dca 100644
--- a/src/osmo-bts-sysmo/misc/sysmobts_mgr_temp.c
+++ b/src/osmo-bts-sysmo/misc/sysmobts_mgr_temp.c
@@ -77,6 +77,34 @@ static int next_state(enum sysmobts_temp_state current_state, int critical, int
 	return next_state;
 }
 
+static void handle_normal_actions(int actions)
+{
+	/* switch off the PA */
+	if (actions & TEMP_ACT_NORM_PA_ON) {
+		if (!is_sbts2050()) {
+			LOGP(DTEMP, LOGL_NOTICE,
+				"PA can only be switched-on on the master\n");
+		} else if (sbts2050_uc_set_pa_power(1) != 0) {
+			LOGP(DTEMP, LOGL_ERROR,
+				"Failed to switch on the PA\n");
+		} else {
+			LOGP(DTEMP, LOGL_NOTICE,
+				"Switched on the PA as normal action.\n");
+		}
+	}
+
+	if (actions & TEMP_ACT_NORM_BTS_SRV_ON) {
+		LOGP(DTEMP, LOGL_NOTICE,
+			"Going to switch on the BTS service\n");
+		/*
+		 * TODO: use/create something like nspawn that serializes
+		 * and used SIGCHLD/waitpid to pick up the dead processes
+		 * without invoking shell.
+		 */
+		system("/bin/systemctl start sysmobts.service");
+	}
+}
+
 static void handle_actions(int actions)
 {
 	/* switch off the PA */
@@ -91,12 +119,6 @@ static void handle_actions(int actions)
 			LOGP(DTEMP, LOGL_NOTICE,
 				"Switched off the PA due temperature.\n");
 		}
-		/*
-		 * TODO: remember we switched off things so we could switch
-		 * it back on. But we would need to make sure that the BTS
-		 * will not transmit with full power at that time. This
-		 * requires the control protocol.
-		 */
 	}
 
 	if (actions & TEMP_ACT_BTS_SRV_OFF) {
@@ -112,14 +134,16 @@ static void handle_actions(int actions)
 }
 
 /**
- * Go back to normal! Undo everything we did in the other states. For
- * reducint the transmit power, the question is if we should slowly set
- * it back to normal, let the BTS slowly increase it.. or handle it here
- * as well?
+ * Go back to normal! Depending on the configuration execute the normal
+ * actions that could (start to) undo everything we did in the other
+ * states. What is still missing is the power increase/decrease depending
+ * on the state. E.g. starting from WARNING_HYST we might want to slowly
+ * ramp up the output power again.
  */
 static void execute_normal_act(struct sysmobts_mgr_instance *manager)
 {
 	LOGP(DTEMP, LOGL_NOTICE, "System is back to normal temperature.\n");
+	handle_normal_actions(manager->action_norm);
 }
 
 static void execute_warning_act(struct sysmobts_mgr_instance *manager)
diff --git a/src/osmo-bts-sysmo/misc/sysmobts_mgr_vty.c b/src/osmo-bts-sysmo/misc/sysmobts_mgr_vty.c
index 5c75840e..9c7e7d30 100644
--- a/src/osmo-bts-sysmo/misc/sysmobts_mgr_vty.c
+++ b/src/osmo-bts-sysmo/misc/sysmobts_mgr_vty.c
@@ -54,6 +54,7 @@ static enum node_type go_to_parent(struct vty *vty)
 	case MGR_NODE:
 		vty->node = CONFIG_NODE;
 		break;
+	case ACT_NORM_NODE:
 	case ACT_WARN_NODE:
 	case ACT_CRIT_NODE:
 	case LIMIT_RF_NODE:
@@ -72,6 +73,7 @@ static int is_config_node(struct vty *vty, int node)
 {
 	switch (node) {
 	case MGR_NODE:
+	case ACT_NORM_NODE:
 	case ACT_WARN_NODE:
 	case ACT_CRIT_NODE:
 	case LIMIT_RF_NODE:
@@ -101,6 +103,12 @@ static struct cmd_node mgr_node = {
 	1,
 };
 
+static struct cmd_node act_norm_node = {
+	ACT_NORM_NODE,
+	"%s(action-normal)# ",
+	1,
+};
+
 static struct cmd_node act_warn_node = {
 	ACT_WARN_NODE,
 	"%s(action-warn)# ",
@@ -155,6 +163,15 @@ static void write_temp_limit(struct vty *vty, const char *name,
 		limit->thresh_crit, VTY_NEWLINE);
 }
 
+static void write_norm_action(struct vty *vty, const char *name, int actions)
+{
+	vty_out(vty, " %s%s", name, VTY_NEWLINE);
+	vty_out(vty, "  %spa-on%s",
+		(actions & TEMP_ACT_NORM_PA_ON) ? "" : "no ", VTY_NEWLINE);
+	vty_out(vty, "  %sbts-service-on%s",
+		(actions & TEMP_ACT_NORM_BTS_SRV_ON) ? "" : "no ", VTY_NEWLINE);
+}
+
 static void write_action(struct vty *vty, const char *name, int actions)
 {
 	vty_out(vty, " %s%s", name, VTY_NEWLINE);
@@ -183,6 +200,7 @@ static int config_write_mgr(struct vty *vty)
 	write_temp_limit(vty, "limits board", &s_mgr->board_limit);
 	write_temp_limit(vty, "limits pa", &s_mgr->pa_limit);
 
+	write_norm_action(vty, "actions normal", s_mgr->action_norm);
 	write_action(vty, "actions warn", s_mgr->action_warn);
 	write_action(vty, "actions critical", s_mgr->action_crit);
 
@@ -237,10 +255,47 @@ DEFUN(cfg_action_##name, cfg_action_##name##_cmd,			\
 	vty->index = &s_mgr->variable;					\
 	return CMD_SUCCESS;						\
 }
+CFG_ACTION(normal, "Normal Actions\n", ACT_NORM_NODE, action_norm)
 CFG_ACTION(warn, "Warning Actions\n", ACT_WARN_NODE, action_warn)
 CFG_ACTION(critical, "Critical Actions\n", ACT_CRIT_NODE, action_crit)
 #undef CFG_ACTION
 
+DEFUN(cfg_action_pa_on, cfg_action_pa_on_cmd,
+	"pa-on",
+	"Switch the Power Amplifier on\n")
+{
+	int *action = vty->index;
+	*action |= TEMP_ACT_NORM_PA_ON;
+	return CMD_SUCCESS;
+}
+
+DEFUN(cfg_no_action_pa_on, cfg_no_action_pa_on_cmd,
+	"no pa-on",
+	NO_STR "Switch the Power Amplifier on\n")
+{
+	int *action = vty->index;
+	*action &= ~TEMP_ACT_NORM_PA_ON;
+	return CMD_SUCCESS;
+}
+
+DEFUN(cfg_action_bts_srv_on, cfg_action_bts_srv_on_cmd,
+	"bts-service-on",
+	"Start the systemd sysmobts.service\n")
+{
+	int *action = vty->index;
+	*action |= TEMP_ACT_NORM_BTS_SRV_ON;
+	return CMD_SUCCESS;
+}
+
+DEFUN(cfg_no_action_bts_srv_on, cfg_no_action_bts_srv_on_cmd,
+	"no bts-service-on",
+	NO_STR "Start the systemd sysmobts.service\n")
+{
+	int *action = vty->index;
+	*action &= ~TEMP_ACT_NORM_BTS_SRV_ON;
+	return CMD_SUCCESS;
+}
+
 DEFUN(cfg_action_pa_off, cfg_action_pa_off_cmd,
 	"pa-off",
 	"Switch the Power Amplifier off\n")
@@ -333,6 +388,14 @@ static void register_limit(int limit)
 	install_element(limit, &cfg_thresh_crit_cmd);
 }
 
+static void register_normal_action(int act)
+{
+	install_element(act, &cfg_action_pa_on_cmd);
+	install_element(act, &cfg_no_action_pa_on_cmd);
+	install_element(act, &cfg_action_bts_srv_on_cmd);
+	install_element(act, &cfg_no_action_bts_srv_on_cmd);
+}
+
 static void register_action(int act)
 {
 #if 0
@@ -382,6 +445,11 @@ int sysmobts_mgr_vty_init(void)
 	register_limit(LIMIT_PA_NODE);
 	vty_install_default(LIMIT_PA_NODE);
 
+	/* install the normal node */
+	install_node(&act_norm_node, config_write_dummy);
+	install_element(MGR_NODE, &cfg_action_normal_cmd);
+	register_normal_action(ACT_NORM_NODE);
+
 	/* install the warning and critical node */
 	install_node(&act_warn_node, config_write_dummy);
 	install_element(MGR_NODE, &cfg_action_warn_cmd);