X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=backend.c;h=981625b61095dd42995d32291ec3fdc4b58cb597;hp=e32d84165273cb5c71fc520d4926b2afd1708a18;hb=86d59660ca54dd2e3e2457f3683198afc2b2701c;hpb=aee2ab6775d96609a4632703827c409a7f9abcca

diff --git a/backend.c b/backend.c
index e32d8416..981625b6 100644
--- a/backend.c
+++ b/backend.c
@@ -53,7 +53,7 @@
 #include "lib/getrusage.h"
 #include "idletime.h"
 #include "err.h"
-#include "tp.h"
+#include "lib/tp.h"
 
 static pthread_t disk_util_thread;
 static struct fio_mutex *disk_thread_mutex;
@@ -391,7 +391,7 @@ static int break_on_this_error(struct thread_data *td, enum fio_ddir ddir,
 			 * fill_device option is set.
 			 */
 			td_clear_error(td);
-			td->terminate = 1;
+			fio_mark_td_terminate(td);
 			return 1;
 		} else {
 			/*
@@ -460,7 +460,7 @@ static void do_verify(struct thread_data *td, uint64_t verify_bytes)
 		if (runtime_exceeded(td, &td->tv_cache)) {
 			__update_tv_cache(td);
 			if (runtime_exceeded(td, &td->tv_cache)) {
-				td->terminate = 1;
+				fio_mark_td_terminate(td);
 				break;
 			}
 		}
@@ -714,7 +714,7 @@ static uint64_t do_io(struct thread_data *td)
 		if (runtime_exceeded(td, &td->tv_cache)) {
 			__update_tv_cache(td);
 			if (runtime_exceeded(td, &td->tv_cache)) {
-				td->terminate = 1;
+				fio_mark_td_terminate(td);
 				break;
 			}
 		}
@@ -922,7 +922,7 @@ reap:
 
 	if (td->o.fill_device && td->error == ENOSPC) {
 		td->error = 0;
-		td->terminate = 1;
+		fio_mark_td_terminate(td);
 	}
 	if (!td->error) {
 		struct fio_file *f;
@@ -1603,6 +1603,13 @@ static int fork_main(int shmid, int offset)
 	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.
  */
@@ -1680,6 +1687,17 @@ static void reap_threads(unsigned int *nr_running, unsigned int *t_rate,
 			}
 		}
 
+		/*
+		 * 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
 		 */
@@ -2050,7 +2068,7 @@ int fio_backend(void)
 	run_threads();
 
 	if (!fio_abort) {
-		show_run_stats();
+		__show_run_stats();
 		if (write_bw_log) {
 			int i;