X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=fio.c;h=a15080b4a26e89bebea6edf5def458e701e0cddd;hp=e205b3aa52ac9f9a693fd249c52de1b84f45afb8;hb=009b1be41b32bf7e32b441c6a22e3ae628ec9b89;hpb=2b13e716c0921356c0930522718e00b8df34293a

diff --git a/fio.c b/fio.c
index e205b3aa..a15080b4 100644
--- a/fio.c
+++ b/fio.c
@@ -45,6 +45,8 @@
 #include "cgroup.h"
 #include "profile.h"
 #include "lib/rand.h"
+#include "memalign.h"
+#include "server.h"
 
 unsigned long page_mask;
 unsigned long page_size;
@@ -69,6 +71,8 @@ static pthread_t disk_util_thread;
 static struct flist_head *cgroup_list;
 static char *cgroup_mnt;
 
+unsigned long arch_flags = 0;
+
 struct io_log *agg_io_log[2];
 
 #define TERMINATE_ALL		(-1)
@@ -101,7 +105,9 @@ static void terminate_threads(int group_id)
 			/*
 			 * if the thread is running, just let it exit
 			 */
-			if (td->runstate < TD_RUNNING)
+			if (!td->pid)
+				continue;
+			else if (td->runstate < TD_RAMP)
 				kill(td->pid, SIGTERM);
 			else {
 				struct ioengine_ops *ops = td->io_ops;
@@ -113,18 +119,13 @@ static void terminate_threads(int group_id)
 	}
 }
 
-/*
- * Happens on thread runs with ctrl-c, ignore our own SIGQUIT
- */
-static void sig_quit(int sig)
-{
-}
-
 static void sig_int(int sig)
 {
 	if (threads) {
 		log_info("\nfio: terminating on signal %d\n", sig);
+		exit_backend = 1;
 		fflush(stdout);
+		exit_value = 128;
 		terminate_threads(TERMINATE_ALL);
 	}
 }
@@ -138,7 +139,9 @@ static void *disk_thread_main(void *data)
 		if (!threads)
 			break;
 		update_io_ticks();
-		print_thread_status();
+
+		if (!is_backend)
+			print_thread_status();
 	}
 
 	return NULL;
@@ -176,7 +179,7 @@ static void set_sig_handlers(void)
 	sigaction(SIGINT, &act, NULL);
 
 	memset(&act, 0, sizeof(act));
-	act.sa_handler = sig_quit;
+	act.sa_handler = sig_int;
 	act.sa_flags = SA_RESTART;
 	sigaction(SIGTERM, &act, NULL);
 }
@@ -185,7 +188,7 @@ static void set_sig_handlers(void)
  * Check if we are above the minimum rate given.
  */
 static int __check_min_rate(struct thread_data *td, struct timeval *now,
-			    enum td_ddir ddir)
+			    enum fio_ddir ddir)
 {
 	unsigned long long bytes = 0;
 	unsigned long iops = 0;
@@ -377,10 +380,15 @@ requeue:
 	return 0;
 }
 
+static inline void __update_tv_cache(struct thread_data *td)
+{
+	fio_gettime(&td->tv_cache, NULL);
+}
+
 static inline void update_tv_cache(struct thread_data *td)
 {
 	if ((++td->tv_cache_nr & td->tv_cache_mask) == td->tv_cache_mask)
-		fio_gettime(&td->tv_cache, NULL);
+		__update_tv_cache(td);
 }
 
 static int break_on_this_error(struct thread_data *td, int *retptr)
@@ -466,8 +474,11 @@ static void do_verify(struct thread_data *td)
 		update_tv_cache(td);
 
 		if (runtime_exceeded(td, &td->tv_cache)) {
-			td->terminate = 1;
-			break;
+			__update_tv_cache(td);
+			if (runtime_exceeded(td, &td->tv_cache)) {
+				td->terminate = 1;
+				break;
+			}
 		}
 
 		io_u = __get_io_u(td);
@@ -545,13 +556,14 @@ sync_done:
 
 		/*
 		 * if we can queue more, do so. but check if there are
-		 * completed io_u's first.
+		 * completed io_u's first. Note that we can get BUSY even
+		 * without IO queued, if the system is resource starved.
 		 */
-		full = queue_full(td) || ret == FIO_Q_BUSY;
+		full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth);
 		if (full || !td->o.iodepth_batch_complete) {
 			min_events = min(td->o.iodepth_batch_complete,
 					 td->cur_depth);
-			if (full && !min_events)
+			if (full && !min_events && td->o.iodepth_batch_complete != 0)
 				min_events = 1;
 
 			do {
@@ -612,8 +624,11 @@ static void do_io(struct thread_data *td)
 		update_tv_cache(td);
 
 		if (runtime_exceeded(td, &td->tv_cache)) {
-			td->terminate = 1;
-			break;
+			__update_tv_cache(td);
+			if (runtime_exceeded(td, &td->tv_cache)) {
+				td->terminate = 1;
+				break;
+			}
 		}
 
 		io_u = get_io_u(td);
@@ -702,13 +717,15 @@ sync_done:
 			break;
 
 		/*
-		 * See if we need to complete some commands
+		 * See if we need to complete some commands. Note that we
+		 * can get BUSY even without IO queued, if the system is
+		 * resource starved.
 		 */
-		full = queue_full(td) || ret == FIO_Q_BUSY;
+		full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth);
 		if (full || !td->o.iodepth_batch_complete) {
 			min_evts = min(td->o.iodepth_batch_complete,
 					td->cur_depth);
-			if (full && !min_evts)
+			if (full && !min_evts && td->o.iodepth_batch_complete != 0)
 				min_evts = 1;
 
 			if (__should_check_rate(td, 0) ||
@@ -765,8 +782,11 @@ sync_done:
 		struct fio_file *f;
 
 		i = td->cur_depth;
-		if (i)
+		if (i) {
 			ret = io_u_queued_complete(td, i, NULL);
+			if (td->o.fill_device && td->error == ENOSPC)
+				td->error = 0;
+		}
 
 		if (should_fsync(td) && td->o.end_fsync) {
 			td_set_runstate(td, TD_FSYNCING);
@@ -796,7 +816,7 @@ static void cleanup_io_u(struct thread_data *td)
 		io_u = flist_entry(entry, struct io_u, list);
 
 		flist_del(&io_u->list);
-		free(io_u);
+		fio_memfree(io_u, sizeof(*io_u));
 	}
 
 	free_io_mem(td);
@@ -843,8 +863,9 @@ static int init_io_u(struct thread_data *td)
 		if (td->terminate)
 			return 1;
 
-		if (posix_memalign(&ptr, cl_align, sizeof(*io_u))) {
-			log_err("fio: posix_memalign=%s\n", strerror(errno));
+		ptr = fio_memalign(cl_align, sizeof(*io_u));
+		if (!ptr) {
+			log_err("fio: unable to allocate aligned memory\n");
 			break;
 		}
 
@@ -857,9 +878,9 @@ static int init_io_u(struct thread_data *td)
 			io_u->buf = p + max_bs * i;
 			dprint(FD_MEM, "io_u %p, mem %p\n", io_u, io_u->buf);
 
-			if (td_write(td) && !td->o.refill_buffers)
+			if (td_write(td))
 				io_u_fill_buffer(td, io_u, max_bs);
-			else if (td_write(td) && td->o.verify_pattern_bytes) {
+			if (td_write(td) && td->o.verify_pattern_bytes) {
 				/*
 				 * Fill the buffer with the pattern if we are
 				 * going to be doing writes.
@@ -968,11 +989,6 @@ static void reset_io_counters(struct thread_data *td)
 	 */
 	if (td->o.time_based || td->o.loops)
 		td->nr_done_files = 0;
-
-	/*
-	 * Set the same seed to get repeatable runs
-	 */
-	td_fill_rand_seeds(td);
 }
 
 void reset_all_stats(struct thread_data *td)
@@ -1006,6 +1022,11 @@ static void clear_io_state(struct thread_data *td)
 	close_files(td);
 	for_each_file(td, f, i)
 		fio_file_clear_done(f);
+
+	/*
+	 * Set the same seed to get repeatable runs
+	 */
+	td_fill_rand_seeds(td);
 }
 
 static int exec_string(const char *string)
@@ -1035,10 +1056,11 @@ static void *thread_main(void *data)
 	pthread_condattr_t attr;
 	int clear_state;
 
-	if (!td->o.use_thread)
+	if (!td->o.use_thread) {
 		setsid();
-
-	td->pid = getpid();
+		td->pid = getpid();
+	} else
+		td->pid = gettid();
 
 	dprint(FD_PROCESS, "jobs pid=%d started\n", (int) td->pid);
 
@@ -1082,6 +1104,22 @@ static void *thread_main(void *data)
 		goto err;
 	}
 
+	/*
+	 * If we have a gettimeofday() thread, make sure we exclude that
+	 * thread from this job
+	 */
+	if (td->o.gtod_cpu)
+		fio_cpu_clear(&td->o.cpumask, td->o.gtod_cpu);
+
+	/*
+	 * Set affinity first, in case it has an impact on the memory
+	 * allocations.
+	 */
+	if (td->o.cpumask_set && fio_setaffinity(td->pid, td->o.cpumask) == -1) {
+		td_verror(td, errno, "cpu_set_affinity");
+		goto err;
+	}
+
 	/*
 	 * May alter parameters that init_io_u() will use, so we need to
 	 * do this first.
@@ -1095,23 +1133,6 @@ static void *thread_main(void *data)
 	if (td->o.verify_async && verify_async_init(td))
 		goto err;
 
-	if (td->o.cpumask_set && fio_setaffinity(td->pid, td->o.cpumask) == -1) {
-		td_verror(td, errno, "cpu_set_affinity");
-		goto err;
-	}
-
-	/*
-	 * If we have a gettimeofday() thread, make sure we exclude that
-	 * thread from this job
-	 */
-	if (td->o.gtod_cpu) {
-		fio_cpu_clear(&td->o.cpumask, td->o.gtod_cpu);
-		if (fio_setaffinity(td->pid, td->o.cpumask) == -1) {
-			td_verror(td, errno, "cpu_set_affinity");
-			goto err;
-		}
-	}
-
 	if (td->ioprio_set) {
 		if (ioprio_set(IOPRIO_WHO_PROCESS, 0, td->ioprio) == -1) {
 			td_verror(td, errno, "ioprio_set");
@@ -1285,6 +1306,7 @@ static int fork_main(int shmid, int offset)
 	struct thread_data *td;
 	void *data, *ret;
 
+#ifndef __hpux
 	data = shmat(shmid, NULL, 0);
 	if (data == (void *) -1) {
 		int __err = errno;
@@ -1292,6 +1314,12 @@ static int fork_main(int shmid, int offset)
 		perror("shmat");
 		return __err;
 	}
+#else
+	/*
+	 * HP-UX inherits shm mappings?
+	 */
+	data = threads;
+#endif
 
 	td = data + offset * sizeof(struct thread_data);
 	ret = thread_main(td);
@@ -1478,10 +1506,8 @@ static void run_threads(void)
 	for_each_td(td, i) {
 		print_status_init(td->thread_number - 1);
 
-		if (!td->o.create_serialize) {
-			init_disk_util(td);
+		if (!td->o.create_serialize)
 			continue;
-		}
 
 		/*
 		 * do file setup here so it happens sequentially,
@@ -1509,14 +1535,12 @@ static void run_threads(void)
 					td_io_close_file(td, f);
 			}
 		}
-
-		init_disk_util(td);
 	}
 
 	set_genesis_time();
 
 	while (todo) {
-		struct thread_data *map[MAX_JOBS];
+		struct thread_data *map[REAL_MAX_JOBS];
 		struct timeval this_start;
 		int this_jobs = 0, left;
 
@@ -1549,6 +1573,8 @@ static void run_threads(void)
 				break;
 			}
 
+			init_disk_util(td);
+
 			/*
 			 * Set state to created. Thread will transition
 			 * to TD_INITIALIZED when it's done setting up.
@@ -1668,34 +1694,8 @@ static void run_threads(void)
 	fio_unpin_memory();
 }
 
-int main(int argc, char *argv[])
+int exec_run(void)
 {
-	long ps;
-
-	sinit();
-	init_rand(&__fio_rand_state);
-
-	/*
-	 * We need locale for number printing, if it isn't set then just
-	 * go with the US format.
-	 */
-	if (!getenv("LC_NUMERIC"))
-		setlocale(LC_NUMERIC, "en_US");
-
-	ps = sysconf(_SC_PAGESIZE);
-	if (ps < 0) {
-		log_err("Failed to get page size\n");
-		return 1;
-	}
-
-	page_size = ps;
-	page_mask = ps - 1;
-
-	fio_keywords_init();
-
-	if (parse_options(argc, argv))
-		return 1;
-
 	if (exec_profile && load_profile(exec_profile))
 		return 1;
 
@@ -1739,3 +1739,44 @@ int main(int argc, char *argv[])
 	fio_mutex_remove(writeout_mutex);
 	return exit_value;
 }
+
+void reset_fio_state(void)
+{
+	groupid = 0;
+	thread_number = 0;
+	nr_process = 0;
+	nr_thread = 0;
+	done_secs = 0;
+}
+
+int main(int argc, char *argv[], char *envp[])
+{
+	long ps;
+
+	arch_init(envp);
+
+	sinit();
+
+	/*
+	 * We need locale for number printing, if it isn't set then just
+	 * go with the US format.
+	 */
+	if (!getenv("LC_NUMERIC"))
+		setlocale(LC_NUMERIC, "en_US");
+
+	ps = sysconf(_SC_PAGESIZE);
+	if (ps < 0) {
+		log_err("Failed to get page size\n");
+		return 1;
+	}
+
+	page_size = ps;
+	page_mask = ps - 1;
+
+	fio_keywords_init();
+
+	if (parse_options(argc, argv))
+		return 1;
+
+	return exec_run();
+}