#include <inttypes.h>
#include <stdio.h>
#include <stdlib.h>
+#include <time.h>
+#include <common/defaults.h>
#include <common/error.h>
#include "health.h"
+static const struct timespec time_delta = {
+ .tv_sec = DEFAULT_HEALTH_CHECK_DELTA_S,
+ .tv_nsec = DEFAULT_HEALTH_CHECK_DELTA_NS,
+};
+
+/* Define TLS health state. */
+DEFINE_URCU_TLS(struct health_state, health_state);
+
+/*
+ * It ensures that TLS memory used for the node and its container structure
+ * don't get reclaimed after the TLS owner thread exits until we have finished
+ * using it.
+ */
+static pthread_mutex_t health_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+static struct health_tls_state_list health_state_list = {
+ .head = CDS_LIST_HEAD_INIT(health_state_list.head),
+};
+
+/*
+ * This keeps track of the error state for unregistered thread. A thread
+ * reporting a health error, normally unregisters and quits. This makes the TLS
+ * health state not available to the health_check_state() call so on unregister
+ * we update this global error array so we can keep track of which thread was
+ * on error if the TLS health state has been removed.
+ */
+static enum health_flags global_error_state[HEALTH_NUM_TYPE];
+
+/*
+ * Lock health state global list mutex.
+ */
+static void state_lock(void)
+{
+ pthread_mutex_lock(&health_mutex);
+}
+
+/*
+ * Unlock health state global list mutex.
+ */
+static void state_unlock(void)
+{
+ pthread_mutex_unlock(&health_mutex);
+}
+
/*
- * Check health of a specific health state counter.
+ * Set time difference in res from time_a and time_b.
+ */
+static void time_diff(const struct timespec *time_a,
+ const struct timespec *time_b, struct timespec *res)
+{
+ if (time_a->tv_nsec - time_b->tv_nsec < 0) {
+ res->tv_sec = time_a->tv_sec - time_b->tv_sec - 1;
+ res->tv_nsec = 1000000000L + time_a->tv_sec - time_b->tv_sec;
+ } else {
+ res->tv_sec = time_a->tv_sec - time_b->tv_sec;
+ res->tv_nsec = time_a->tv_nsec - time_b->tv_nsec;
+ }
+}
+
+/*
+ * Return true if time_a - time_b > diff, else false.
+ */
+static int time_diff_gt(const struct timespec *time_a,
+ const struct timespec *time_b, const struct timespec *diff)
+{
+ struct timespec res;
+
+ time_diff(time_a, time_b, &res);
+ time_diff(&res, diff, &res);
+
+ if (res.tv_sec > 0) {
+ return 1;
+ } else if (res.tv_sec == 0 && res.tv_nsec > 0) {
+ return 1;
+ }
+
+ return 0;
+}
+
+/*
+ * Validate health state. Checks for the error flag or health conditions.
*
* Return 0 if health is bad or else 1.
*/
-int health_check_state(struct health_state *state)
+static int validate_state(struct health_state *state)
{
+ int retval = 1, ret;
unsigned long current, last;
- int ret = 1;
+ struct timespec current_time;
assert(state);
last = state->last;
current = uatomic_read(&state->current);
+ ret = clock_gettime(CLOCK_MONOTONIC, ¤t_time);
+ if (ret < 0) {
+ PERROR("Error reading time\n");
+ /* error */
+ retval = 0;
+ goto end;
+ }
+
/*
- * Here are the conditions for a bad health. Either flag HEALTH_ERROR is
- * set, or the progress counter is the same as the last one and we are NOT
- * waiting for a poll() call.
+ * Thread is in bad health if flag HEALTH_ERROR is set. It is also in bad
+ * health if, after the delta delay has passed, its the progress counter
+ * has not moved and it has NOT been waiting for a poll() call.
*/
- if ((uatomic_read(&state->flags) & HEALTH_ERROR) ||
- (current == last && !HEALTH_IS_IN_POLL(current))) {
- /* error */
- ret = 0;
+ if (uatomic_read(&state->flags) & HEALTH_ERROR) {
+ retval = 0;
+ goto end;
}
- DBG("Health state current %" PRIu64 ", last %" PRIu64 ", ret %d",
+ /*
+ * Initial condition need to update the last counter and sample time, but
+ * should not check health in this initial case, because we don't know how
+ * much time has passed.
+ */
+ if (state->last_time.tv_sec == 0 && state->last_time.tv_nsec == 0) {
+ /* update last counter and last sample time */
+ state->last = current;
+ memcpy(&state->last_time, ¤t_time, sizeof(current_time));
+ } else {
+ if (time_diff_gt(¤t_time, &state->last_time, &time_delta)) {
+ if (current == last && !HEALTH_IS_IN_POLL(current)) {
+ /* error */
+ retval = 0;
+ }
+ /* update last counter and last sample time */
+ state->last = current;
+ memcpy(&state->last_time, ¤t_time, sizeof(current_time));
+
+ /* On error, stop right now and notify caller. */
+ if (retval == 0) {
+ goto end;
+ }
+ }
+ }
+
+end:
+ DBG("Health state current %lu, last %lu, ret %d",
current, last, ret);
+ return retval;
+}
+
+/*
+ * Check health of a specific health type. Note that if a thread has not yet
+ * initialize its health subsystem or has quit, it's considered in a good
+ * state.
+ *
+ * Return 0 if health is bad or else 1.
+ */
+int health_check_state(enum health_type type)
+{
+ int retval = 1;
+ struct health_state *state;
+
+ assert(type < HEALTH_NUM_TYPE);
+
+ state_lock();
+
+ cds_list_for_each_entry(state, &health_state_list.head, node) {
+ int ret;
+
+ if (state->type != type) {
+ continue;
+ }
+
+ ret = validate_state(state);
+ if (!ret) {
+ retval = 0;
+ goto end;
+ }
+ }
+
+ /* Check the global state since some state might not be visible anymore. */
+ if (global_error_state[type] & HEALTH_ERROR) {
+ retval = 0;
+ }
+
+end:
+ state_unlock();
+
+ DBG("Health check for type %d is %s", (int) type,
+ (retval == 0) ? "BAD" : "GOOD");
+ return retval;
+}
+
+/*
+ * Init health state.
+ */
+void health_register(enum health_type type)
+{
+ assert(type < HEALTH_NUM_TYPE);
+
+ /* Init TLS state. */
+ uatomic_set(&URCU_TLS(health_state).last, 0);
+ uatomic_set(&URCU_TLS(health_state).last_time.tv_sec, 0);
+ uatomic_set(&URCU_TLS(health_state).last_time.tv_nsec, 0);
+ uatomic_set(&URCU_TLS(health_state).current, 0);
+ uatomic_set(&URCU_TLS(health_state).flags, 0);
+ uatomic_set(&URCU_TLS(health_state).type, type);
+ /* Add it to the global TLS state list. */
+ state_lock();
+ cds_list_add(&URCU_TLS(health_state).node, &health_state_list.head);
+ state_unlock();
+}
+
+/*
+ * Remove node from global list.
+ */
+void health_unregister(void)
+{
+ state_lock();
/*
- * Update last counter. This value is and MUST be access only in this
- * function.
+ * On error, set the global_error_state since we are about to remove
+ * the node from the global list.
*/
- state->last = current;
-
- return ret;
+ if (uatomic_read(&URCU_TLS(health_state).flags) & HEALTH_ERROR) {
+ uatomic_set(&global_error_state[URCU_TLS(health_state).type],
+ HEALTH_ERROR);
+ }
+ cds_list_del(&URCU_TLS(health_state).node);
+ state_unlock();
}