X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=backend.c;h=2e6a377cfb4cb4317294eb854cd61712f402caf7;hp=2f46329376451f18a49c94696e84687dda4390d8;hb=9107a641e67f27003fa6cbe7b55b1ec6a0239197;hpb=2041bd343da1c1e955253f62374588718c64f0f3

diff --git a/backend.c b/backend.c
index 2f463293..2e6a377c 100644
--- a/backend.c
+++ b/backend.c
@@ -62,17 +62,22 @@ struct io_log *agg_io_log[DDIR_RWDIR_CNT];
 
 int groupid = 0;
 unsigned int thread_number = 0;
+unsigned int nr_segments = 0;
+unsigned int cur_segment = 0;
 unsigned int stat_number = 0;
-int shm_id = 0;
 int temp_stall_ts;
 unsigned long done_secs = 0;
+#ifdef PTHREAD_ERRORCHECK_MUTEX_INITIALIZER_NP
+pthread_mutex_t overlap_check = PTHREAD_ERRORCHECK_MUTEX_INITIALIZER_NP;
+#else
 pthread_mutex_t overlap_check = PTHREAD_MUTEX_INITIALIZER;
+#endif
 
 #define JOB_START_TIMEOUT	(5 * 1000)
 
 static void sig_int(int sig)
 {
-	if (threads) {
+	if (nr_segments) {
 		if (is_backend)
 			fio_server_got_signal(sig);
 		else {
@@ -81,7 +86,7 @@ static void sig_int(int sig)
 			exit_value = 128;
 		}
 
-		fio_terminate_threads(TERMINATE_ALL);
+		fio_terminate_threads(TERMINATE_ALL, TERMINATE_ALL);
 	}
 }
 
@@ -134,8 +139,8 @@ static bool __check_min_rate(struct thread_data *td, struct timespec *now,
 	unsigned long long bytes = 0;
 	unsigned long iops = 0;
 	unsigned long spent;
-	unsigned long rate;
-	unsigned int ratemin = 0;
+	unsigned long long rate;
+	unsigned long long ratemin = 0;
 	unsigned int rate_iops = 0;
 	unsigned int rate_iops_min = 0;
 
@@ -169,7 +174,7 @@ static bool __check_min_rate(struct thread_data *td, struct timespec *now,
 			 * check bandwidth specified rate
 			 */
 			if (bytes < td->rate_bytes[ddir]) {
-				log_err("%s: rate_min=%uB/s not met, only transferred %lluB\n",
+				log_err("%s: rate_min=%lluB/s not met, only transferred %lluB\n",
 					td->o.name, ratemin, bytes);
 				return true;
 			} else {
@@ -180,7 +185,7 @@ static bool __check_min_rate(struct thread_data *td, struct timespec *now,
 
 				if (rate < ratemin ||
 				    bytes < td->rate_bytes[ddir]) {
-					log_err("%s: rate_min=%uB/s not met, got %luB/s\n",
+					log_err("%s: rate_min=%lluB/s not met, got %lluB/s\n",
 						td->o.name, ratemin, rate);
 					return true;
 				}
@@ -201,7 +206,7 @@ static bool __check_min_rate(struct thread_data *td, struct timespec *now,
 
 				if (rate < rate_iops_min ||
 				    iops < td->rate_blocks[ddir]) {
-					log_err("%s: rate_iops_min=%u not met, got %lu IOPS\n",
+					log_err("%s: rate_iops_min=%u not met, got %llu IOPS\n",
 						td->o.name, rate_iops_min, rate);
 					return true;
 				}
@@ -219,12 +224,10 @@ static bool check_min_rate(struct thread_data *td, struct timespec *now)
 {
 	bool ret = false;
 
-	if (td->bytes_done[DDIR_READ])
-		ret |= __check_min_rate(td, now, DDIR_READ);
-	if (td->bytes_done[DDIR_WRITE])
-		ret |= __check_min_rate(td, now, DDIR_WRITE);
-	if (td->bytes_done[DDIR_TRIM])
-		ret |= __check_min_rate(td, now, DDIR_TRIM);
+	for_each_rw_ddir(ddir) {
+		if (td->bytes_done[ddir])
+			ret |= __check_min_rate(td, now, ddir);
+	}
 
 	return ret;
 }
@@ -237,15 +240,10 @@ static void cleanup_pending_aio(struct thread_data *td)
 {
 	int r;
 
-	if (td->error)
-		return;
-
 	/*
 	 * get immediately available events, if any
 	 */
 	r = io_u_queued_complete(td, 0);
-	if (r < 0)
-		return;
 
 	/*
 	 * now cancel remaining active events
@@ -281,6 +279,7 @@ static bool fio_io_sync(struct thread_data *td, struct fio_file *f)
 
 	io_u->ddir = DDIR_SYNC;
 	io_u->file = f;
+	io_u_set(td, io_u, IO_U_F_NO_FILE_PUT);
 
 	if (td_io_prep(td, io_u)) {
 		put_io_u(td, io_u);
@@ -314,7 +313,7 @@ requeue:
 
 static int fio_file_fsync(struct thread_data *td, struct fio_file *f)
 {
-	int ret;
+	int ret, ret2;
 
 	if (fio_file_open(f))
 		return fio_io_sync(td, f);
@@ -323,8 +322,10 @@ static int fio_file_fsync(struct thread_data *td, struct fio_file *f)
 		return 1;
 
 	ret = fio_io_sync(td, f);
-	td_io_close_file(td, f);
-	return ret;
+	ret2 = 0;
+	if (fio_file_open(f))
+		ret2 = td_io_close_file(td, f);
+	return (ret || ret2);
 }
 
 static inline void __update_ts_cache(struct thread_data *td)
@@ -1008,12 +1009,6 @@ static void do_io(struct thread_data *td, uint64_t *bytes_done)
 		if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_READ &&
 		    ((io_u->flags & IO_U_F_VER_LIST) || !td_rw(td))) {
 
-			if (!td->o.verify_pattern_bytes) {
-				io_u->rand_seed = __rand(&td->verify_state);
-				if (sizeof(int) != sizeof(long *))
-					io_u->rand_seed *= __rand(&td->verify_state);
-			}
-
 			if (verify_state_should_stop(td, io_u)) {
 				put_io_u(td, io_u);
 				break;
@@ -1088,7 +1083,7 @@ reap:
 		if (!in_ramp_time(td) && should_check_rate(td)) {
 			if (check_min_rate(td, &comp_time)) {
 				if (exitall_on_terminate || td->o.exitall_error)
-					fio_terminate_threads(td->groupid);
+					fio_terminate_threads(td->groupid, td->o.exit_what);
 				td_verror(td, EIO, "check_min_rate");
 				break;
 			}
@@ -1124,7 +1119,7 @@ reap:
 				td->error = 0;
 		}
 
-		if (should_fsync(td) && td->o.end_fsync) {
+		if (should_fsync(td) && (td->o.end_fsync || td->o.fsync_on_close)) {
 			td_set_runstate(td, TD_FSYNCING);
 
 			for_each_file(td, f, i) {
@@ -1233,7 +1228,7 @@ static int init_io_u(struct thread_data *td)
 		ptr = fio_memalign(cl_align, sizeof(*io_u), td_offload_overlap(td));
 		if (!ptr) {
 			log_err("fio: unable to allocate aligned memory\n");
-			break;
+			return 1;
 		}
 
 		io_u = ptr;
@@ -1464,16 +1459,17 @@ static bool keep_running(struct thread_data *td)
 	return false;
 }
 
-static int exec_string(struct thread_options *o, const char *string, const char *mode)
+static int exec_string(struct thread_options *o, const char *string,
+		       const char *mode)
 {
-	size_t newlen = strlen(string) + strlen(o->name) + strlen(mode) + 9 + 1;
 	int ret;
 	char *str;
 
-	str = malloc(newlen);
-	sprintf(str, "%s &> %s.%s.txt", string, o->name, mode);
+	if (asprintf(&str, "%s > %s.%s.txt 2>&1", string, o->name, mode) < 0)
+		return -1;
 
-	log_info("%s : Saving output of %s in %s.%s.txt\n",o->name, mode, o->name, mode);
+	log_info("%s : Saving output of %s in %s.%s.txt\n", o->name, mode,
+		 o->name, mode);
 	ret = system(str);
 	if (ret == -1)
 		log_err("fio: exec of cmd <%s> failed\n", str);
@@ -1543,7 +1539,7 @@ static void *thread_main(void *data)
 	uint64_t bytes_done[DDIR_RWDIR_CNT];
 	int deadlock_loop_cnt;
 	bool clear_state;
-	int ret;
+	int res, ret;
 
 	sk_out_assign(sk_out);
 	free(fd);
@@ -1737,7 +1733,7 @@ static void *thread_main(void *data)
 	if (!init_random_map(td))
 		goto err;
 
-	if (o->exec_prerun && exec_string(o, o->exec_prerun, (const char *)"prerun"))
+	if (o->exec_prerun && exec_string(o, o->exec_prerun, "prerun"))
 		goto err;
 
 	if (o->pre_read && !pre_read_files(td))
@@ -1868,17 +1864,21 @@ static void *thread_main(void *data)
 	 * offload mode so that we don't clean up this job while
 	 * another thread is checking its io_u's for overlap
 	 */
-	if (td_offload_overlap(td))
-		pthread_mutex_lock(&overlap_check);
+	if (td_offload_overlap(td)) {
+		int res = pthread_mutex_lock(&overlap_check);
+		assert(res == 0);
+	}
 	td_set_runstate(td, TD_FINISHING);
-	if (td_offload_overlap(td))
-		pthread_mutex_unlock(&overlap_check);
+	if (td_offload_overlap(td)) {
+		res = pthread_mutex_unlock(&overlap_check);
+		assert(res == 0);
+	}
 
 	update_rusage_stat(td);
 	td->ts.total_run_time = mtime_since_now(&td->epoch);
-	td->ts.io_bytes[DDIR_READ] = td->io_bytes[DDIR_READ];
-	td->ts.io_bytes[DDIR_WRITE] = td->io_bytes[DDIR_WRITE];
-	td->ts.io_bytes[DDIR_TRIM] = td->io_bytes[DDIR_TRIM];
+	for_each_rw_ddir(ddir) {
+		td->ts.io_bytes[ddir] = td->io_bytes[ddir];
+	}
 
 	if (td->o.verify_state_save && !(td->flags & TD_F_VSTATE_SAVED) &&
 	    (td->o.verify != VERIFY_NONE && td_write(td)))
@@ -1892,10 +1892,10 @@ static void *thread_main(void *data)
 	rate_submit_exit(td);
 
 	if (o->exec_postrun)
-		exec_string(o, o->exec_postrun, (const char *)"postrun");
+		exec_string(o, o->exec_postrun, "postrun");
 
 	if (exitall_on_terminate || (o->exitall_error && td->error))
-		fio_terminate_threads(td->groupid);
+		fio_terminate_threads(td->groupid, td->o.exit_what);
 
 err:
 	if (td->error)
@@ -2044,10 +2044,11 @@ reaped:
 
 		done_secs += mtime_since_now(&td->epoch) / 1000;
 		profile_td_exit(td);
+		flow_exit_job(td);
 	}
 
 	if (*nr_running == cputhreads && !pending && realthreads)
-		fio_terminate_threads(TERMINATE_ALL);
+		fio_terminate_threads(TERMINATE_ALL, TERMINATE_ALL);
 }
 
 static bool __check_trigger_file(void)
@@ -2097,7 +2098,7 @@ void check_trigger_file(void)
 			fio_clients_send_trigger(trigger_remote_cmd);
 		else {
 			verify_save_state(IO_LIST_ALL);
-			fio_terminate_threads(TERMINATE_ALL);
+			fio_terminate_threads(TERMINATE_ALL, TERMINATE_ALL);
 			exec_trigger(trigger_cmd);
 		}
 	}
@@ -2117,8 +2118,16 @@ static int fio_verify_load_state(struct thread_data *td)
 					td->thread_number - 1, &data);
 		if (!ret)
 			verify_assign_state(td, data);
-	} else
-		ret = verify_load_state(td, "local");
+	} else {
+		char prefix[PATH_MAX];
+
+		if (aux_path)
+			sprintf(prefix, "%s%clocal", aux_path,
+					FIO_OS_PATH_SEPARATOR);
+		else
+			strcpy(prefix, "local");
+		ret = verify_load_state(td, prefix);
+	}
 
 	return ret;
 }
@@ -2370,7 +2379,7 @@ reap:
 			dprint(FD_MUTEX, "wait on startup_sem\n");
 			if (fio_sem_down_timeout(startup_sem, 10000)) {
 				log_err("fio: job startup hung? exiting.\n");
-				fio_terminate_threads(TERMINATE_ALL);
+				fio_terminate_threads(TERMINATE_ALL, TERMINATE_ALL);
 				fio_abort = true;
 				nr_started--;
 				free(fd);
@@ -2492,7 +2501,8 @@ int fio_backend(struct sk_out *sk_out)
 
 	set_genesis_time();
 	stat_init();
-	helper_thread_create(startup_sem, sk_out);
+	if (helper_thread_create(startup_sem, sk_out))
+		log_err("fio: failed to create helper thread\n");
 
 	cgroup_list = smalloc(sizeof(*cgroup_list));
 	if (cgroup_list)