return (int) (uintptr_t) ret;
}
+static void dump_td_info(struct thread_data *td)
+{
+ log_err("fio: job '%s' hasn't exited in %lu seconds, it appears to "
+ "be stuck. Doing forceful exit of this job.\n", td->o.name,
+ (unsigned long) time_since_now(&td->terminate_time));
+}
+
/*
* Run over the job map and reap the threads that have exited, if any.
*/
}
}
+ /*
+ * If the job is stuck, do a forceful timeout of it and
+ * move on.
+ */
+ if (td->terminate &&
+ time_since_now(&td->terminate_time) >= FIO_REAP_TIMEOUT) {
+ dump_td_info(td);
+ td_set_runstate(td, TD_REAPED);
+ goto reaped;
+ }
+
/*
* thread is not dead, continue
*/
struct timeval epoch; /* time job was started */
struct timeval last_issue;
struct timeval tv_cache;
+ struct timeval terminate_time;
unsigned int tv_cache_nr;
unsigned int tv_cache_mask;
unsigned int ramp_time_over;
extern int td_bump_runstate(struct thread_data *, int);
extern void td_restore_runstate(struct thread_data *, int);
+/*
+ * Allow 60 seconds for a job to quit on its own, otherwise reap with
+ * a vengeance.
+ */
+#define FIO_REAP_TIMEOUT 60
+
#define TERMINATE_ALL (-1)
extern void fio_terminate_threads(int);
if (group_id == TERMINATE_ALL || groupid == td->groupid) {
dprint(FD_PROCESS, "setting terminate on %s/%d\n",
td->o.name, (int) td->pid);
+
+ if (td->terminate)
+ continue;
+
td->terminate = 1;
td->o.start_delay = 0;
+ fio_gettime(&td->terminate_time, NULL);
/*
* if the thread is running, just let it exit