[systemd-devel] [PATCH] core: send sigabrt on watchdog timeout
Umut Tezduyar Lindskog
umut.tezduyar at axis.com
Sat Oct 25 14:58:34 PDT 2014
if sigabrt doesn't do the job, follow regular shutdown
routine, sigterm > sigkill.
Umut:
- I have done basic testing with suppressing sigabrt,
suppressing both sigabrt & sigterm on client application.
- Documentation needs to be updated. I wasn't sure if we
should mention it in KillMode, KillSignal or WatchdogSec
---
TODO | 2 --
src/core/busname.c | 1 +
src/core/mount.c | 1 +
src/core/scope.c | 1 +
src/core/service.c | 38 +++++++++++++++++++++++++++-----------
src/core/service.h | 1 +
src/core/socket.c | 1 +
src/core/swap.c | 1 +
src/core/unit.c | 3 ++-
src/core/unit.h | 2 +-
10 files changed, 36 insertions(+), 15 deletions(-)
diff --git a/TODO b/TODO
index ed00661..08f8d20 100644
--- a/TODO
+++ b/TODO
@@ -54,8 +54,6 @@ Features:
* consider showing the unit names during boot up in the status output, not just the unit descriptions
-* send SIGABRT when a service watchdog is triggered, by default, so that we acquire a backtrace of the hang.
-
* shouldn't RouteMetric= in networkd's [DHCP] section move to [Network]?
* dhcp: do we allow configuring dhcp routes on interfaces that are not the one we got the dhcp info from?
diff --git a/src/core/busname.c b/src/core/busname.c
index 22d2a6d..c97b0a8 100644
--- a/src/core/busname.c
+++ b/src/core/busname.c
@@ -447,6 +447,7 @@ static void busname_enter_signal(BusName *n, BusNameState state, BusNameResult f
r = unit_kill_context(UNIT(n),
&kill_context,
state != BUSNAME_SIGTERM,
+ false,
-1,
n->control_pid,
false);
diff --git a/src/core/mount.c b/src/core/mount.c
index e284357..4b3a73b 100644
--- a/src/core/mount.c
+++ b/src/core/mount.c
@@ -776,6 +776,7 @@ static void mount_enter_signal(Mount *m, MountState state, MountResult f) {
UNIT(m),
&m->kill_context,
state != MOUNT_MOUNTING_SIGTERM && state != MOUNT_UNMOUNTING_SIGTERM && state != MOUNT_REMOUNTING_SIGTERM,
+ false,
-1,
m->control_pid,
false);
diff --git a/src/core/scope.c b/src/core/scope.c
index e8f9e8d..4745d15 100644
--- a/src/core/scope.c
+++ b/src/core/scope.c
@@ -244,6 +244,7 @@ static void scope_enter_signal(Scope *s, ScopeState state, ScopeResult f) {
UNIT(s),
&s->kill_context,
state != SCOPE_STOP_SIGTERM,
+ false,
-1, -1, false);
if (r < 0)
goto fail;
diff --git a/src/core/service.c b/src/core/service.c
index d160c4e..76618ff 100644
--- a/src/core/service.c
+++ b/src/core/service.c
@@ -56,6 +56,7 @@ static const UnitActiveState state_translation_table[_SERVICE_STATE_MAX] = {
[SERVICE_EXITED] = UNIT_ACTIVE,
[SERVICE_RELOAD] = UNIT_RELOADING,
[SERVICE_STOP] = UNIT_DEACTIVATING,
+ [SERVICE_STOP_SIGABRT] = UNIT_DEACTIVATING,
[SERVICE_STOP_SIGTERM] = UNIT_DEACTIVATING,
[SERVICE_STOP_SIGKILL] = UNIT_DEACTIVATING,
[SERVICE_STOP_POST] = UNIT_DEACTIVATING,
@@ -76,6 +77,7 @@ static const UnitActiveState state_translation_table_idle[_SERVICE_STATE_MAX] =
[SERVICE_EXITED] = UNIT_ACTIVE,
[SERVICE_RELOAD] = UNIT_RELOADING,
[SERVICE_STOP] = UNIT_DEACTIVATING,
+ [SERVICE_STOP_SIGABRT] = UNIT_DEACTIVATING,
[SERVICE_STOP_SIGTERM] = UNIT_DEACTIVATING,
[SERVICE_STOP_SIGKILL] = UNIT_DEACTIVATING,
[SERVICE_STOP_POST] = UNIT_DEACTIVATING,
@@ -663,7 +665,7 @@ static void service_set_state(Service *s, ServiceState state) {
SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST,
SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL,
- SERVICE_STOP_POST,
+ SERVICE_STOP_SIGABRT, SERVICE_STOP_POST,
SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL,
SERVICE_AUTO_RESTART))
s->timer_event_source = sd_event_source_unref(s->timer_event_source);
@@ -672,7 +674,7 @@ static void service_set_state(Service *s, ServiceState state) {
SERVICE_START, SERVICE_START_POST,
SERVICE_RUNNING, SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL,
- SERVICE_STOP_POST,
+ SERVICE_STOP_SIGABRT, SERVICE_STOP_POST,
SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) {
service_unwatch_main_pid(s);
s->main_command = NULL;
@@ -682,7 +684,7 @@ static void service_set_state(Service *s, ServiceState state) {
SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST,
SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL,
- SERVICE_STOP_POST,
+ SERVICE_STOP_SIGABRT, SERVICE_STOP_POST,
SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) {
service_unwatch_control_pid(s);
s->control_command = NULL;
@@ -696,7 +698,7 @@ static void service_set_state(Service *s, ServiceState state) {
SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST,
SERVICE_RUNNING, SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST,
- SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL) &&
+ SERVICE_STOP_SIGABRT, SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL) &&
!(state == SERVICE_DEAD && UNIT(s)->job)) {
service_close_socket_fd(s);
service_connection_unref(s);
@@ -750,7 +752,7 @@ static int service_coldplug(Unit *u) {
SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST,
SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL,
- SERVICE_STOP_POST,
+ SERVICE_STOP_SIGABRT, SERVICE_STOP_POST,
SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) {
usec_t k;
@@ -779,7 +781,7 @@ static int service_coldplug(Unit *u) {
SERVICE_START, SERVICE_START_POST,
SERVICE_RUNNING, SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL,
- SERVICE_STOP_POST,
+ SERVICE_STOP_SIGABRT, SERVICE_STOP_POST,
SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL))) {
r = unit_watch_pid(UNIT(s), s->main_pid);
if (r < 0)
@@ -791,7 +793,7 @@ static int service_coldplug(Unit *u) {
SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST,
SERVICE_RELOAD,
SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL,
- SERVICE_STOP_POST,
+ SERVICE_STOP_SIGABRT, SERVICE_STOP_POST,
SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) {
r = unit_watch_pid(UNIT(s), s->control_pid);
if (r < 0)
@@ -1181,7 +1183,9 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f
r = unit_kill_context(
UNIT(s),
&s->kill_context,
- state != SERVICE_STOP_SIGTERM && state != SERVICE_FINAL_SIGTERM,
+ state != SERVICE_STOP_SIGTERM && state != SERVICE_FINAL_SIGTERM &&
+ state != SERVICE_STOP_SIGABRT,
+ state == SERVICE_STOP_SIGABRT,
s->main_pid,
s->control_pid,
s->main_pid_alien);
@@ -1197,7 +1201,7 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f
}
service_set_state(s, state);
- } else if (state == SERVICE_STOP_SIGTERM)
+ } else if (state == SERVICE_STOP_SIGTERM || state == SERVICE_STOP_SIGABRT)
service_enter_signal(s, SERVICE_STOP_SIGKILL, SERVICE_SUCCESS);
else if (state == SERVICE_STOP_SIGKILL)
service_enter_stop_post(s, SERVICE_SUCCESS);
@@ -1211,7 +1215,8 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f
fail:
log_warning_unit(UNIT(s)->id, "%s failed to kill processes: %s", UNIT(s)->id, strerror(-r));
- if (state == SERVICE_STOP_SIGTERM || state == SERVICE_STOP_SIGKILL)
+ if (state == SERVICE_STOP_SIGTERM || state == SERVICE_STOP_SIGKILL ||
+ state == SERVICE_STOP_SIGABRT)
service_enter_stop_post(s, SERVICE_FAILURE_RESOURCES);
else
service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true);
@@ -1637,6 +1642,7 @@ static int service_start(Unit *u) {
/* We cannot fulfill this request right now, try again later
* please! */
if (s->state == SERVICE_STOP ||
+ s->state == SERVICE_STOP_SIGABRT ||
s->state == SERVICE_STOP_SIGTERM ||
s->state == SERVICE_STOP_SIGKILL ||
s->state == SERVICE_STOP_POST ||
@@ -1695,6 +1701,7 @@ static int service_stop(Unit *u) {
/* Already on it */
if (s->state == SERVICE_STOP ||
+ s->state == SERVICE_STOP_SIGABRT ||
s->state == SERVICE_STOP_SIGTERM ||
s->state == SERVICE_STOP_SIGKILL ||
s->state == SERVICE_STOP_POST ||
@@ -2126,6 +2133,7 @@ static void service_notify_cgroup_empty_event(Unit *u) {
service_enter_running(s, SERVICE_SUCCESS);
break;
+ case SERVICE_STOP_SIGABRT:
case SERVICE_STOP_SIGTERM:
case SERVICE_STOP_SIGKILL:
@@ -2252,6 +2260,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
service_enter_running(s, f);
break;
+ case SERVICE_STOP_SIGABRT:
case SERVICE_STOP_SIGTERM:
case SERVICE_STOP_SIGKILL:
@@ -2392,6 +2401,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
service_enter_signal(s, SERVICE_STOP_SIGTERM, f);
break;
+ case SERVICE_STOP_SIGABRT:
case SERVICE_STOP_SIGTERM:
case SERVICE_STOP_SIGKILL:
if (main_pid_good(s) <= 0)
@@ -2461,6 +2471,12 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us
service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_TIMEOUT);
break;
+ case SERVICE_STOP_SIGABRT:
+ log_warning_unit(UNIT(s)->id,
+ "%s stop-sigabrt timed out. Terminating.", UNIT(s)->id);
+ service_enter_signal(s, SERVICE_STOP_SIGTERM, s->result);
+ break;
+
case SERVICE_STOP_SIGTERM:
if (s->kill_context.send_sigkill) {
log_warning_unit(UNIT(s)->id, "%s stop-sigterm timed out. Killing.", UNIT(s)->id);
@@ -2528,7 +2544,7 @@ static int service_dispatch_watchdog(sd_event_source *source, usec_t usec, void
log_error_unit(UNIT(s)->id, "%s watchdog timeout (limit %s)!", UNIT(s)->id,
format_timespan(t, sizeof(t), s->watchdog_usec, 1));
- service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_WATCHDOG);
+ service_enter_signal(s, SERVICE_STOP_SIGABRT, SERVICE_FAILURE_WATCHDOG);
return 0;
}
diff --git a/src/core/service.h b/src/core/service.h
index 0db0c4d..54fbe46 100644
--- a/src/core/service.h
+++ b/src/core/service.h
@@ -39,6 +39,7 @@ typedef enum ServiceState {
SERVICE_EXITED, /* Nothing is running anymore, but RemainAfterExit is true hence this is OK */
SERVICE_RELOAD,
SERVICE_STOP, /* No STOP_PRE state, instead just register multiple STOP executables */
+ SERVICE_STOP_SIGABRT, /* Watchdog timeout */
SERVICE_STOP_SIGTERM,
SERVICE_STOP_SIGKILL,
SERVICE_STOP_POST,
diff --git a/src/core/socket.c b/src/core/socket.c
index 9004cb4..70ff481 100644
--- a/src/core/socket.c
+++ b/src/core/socket.c
@@ -1579,6 +1579,7 @@ static void socket_enter_signal(Socket *s, SocketState state, SocketResult f) {
UNIT(s),
&s->kill_context,
state != SOCKET_STOP_PRE_SIGTERM && state != SOCKET_FINAL_SIGTERM,
+ false,
-1,
s->control_pid,
false);
diff --git a/src/core/swap.c b/src/core/swap.c
index b2ca048..3448c1e 100644
--- a/src/core/swap.c
+++ b/src/core/swap.c
@@ -711,6 +711,7 @@ static void swap_enter_signal(Swap *s, SwapState state, SwapResult f) {
UNIT(s),
&s->kill_context,
state != SWAP_ACTIVATING_SIGTERM && state != SWAP_DEACTIVATING_SIGTERM,
+ false,
-1,
s->control_pid,
false);
diff --git a/src/core/unit.c b/src/core/unit.c
index e40e6f2..66804c9 100644
--- a/src/core/unit.c
+++ b/src/core/unit.c
@@ -3305,6 +3305,7 @@ int unit_kill_context(
Unit *u,
KillContext *c,
bool sigkill,
+ bool sigabrt,
pid_t main_pid,
pid_t control_pid,
bool main_pid_alien) {
@@ -3317,7 +3318,7 @@ int unit_kill_context(
if (c->kill_mode == KILL_NONE)
return 0;
- sig = sigkill ? SIGKILL : c->kill_signal;
+ sig = sigkill ? SIGKILL : (sigabrt ? SIGABRT : c->kill_signal);
if (main_pid > 0) {
r = kill_and_sigcont(main_pid, sig);
diff --git a/src/core/unit.h b/src/core/unit.h
index 43ab4d1..b4b8fa1 100644
--- a/src/core/unit.h
+++ b/src/core/unit.h
@@ -572,7 +572,7 @@ int unit_write_drop_in_private_format(Unit *u, UnitSetPropertiesMode mode, const
int unit_remove_drop_in(Unit *u, UnitSetPropertiesMode mode, const char *name);
-int unit_kill_context(Unit *u, KillContext *c, bool sigkill, pid_t main_pid, pid_t control_pid, bool main_pid_alien);
+int unit_kill_context(Unit *u, KillContext *c, bool sigkill, bool sigabrt, pid_t main_pid, pid_t control_pid, bool main_pid_alien);
int unit_make_transient(Unit *u);
--
2.1.1
More information about the systemd-devel
mailing list