X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=fio.c;h=10746062927f32b0c6956c0770ec88ba416f8747;hp=93482f5294ed55c245e25a0e98f8151fe7354ced;hb=cf451d1ede3bbbd2fed3619eb43def054d5b5a5a;hpb=ecc314ba7c5f02b7e90ac1dfbce1a74cd4e6d6fe

diff --git a/fio.c b/fio.c
index 93482f52..10746062 100644
--- a/fio.c
+++ b/fio.c
@@ -24,6 +24,7 @@
 #include <unistd.h>
 #include <fcntl.h>
 #include <string.h>
+#include <limits.h>
 #include <signal.h>
 #include <time.h>
 #include <locale.h>
@@ -44,6 +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;
@@ -63,11 +66,13 @@ static struct fio_mutex *startup_mutex;
 static struct fio_mutex *writeout_mutex;
 static volatile int fio_abort;
 static int exit_value;
-static timer_t ival_timer;
 static pthread_t gtod_thread;
+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)
@@ -100,7 +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;
@@ -112,58 +119,56 @@ static void terminate_threads(int group_id)
 	}
 }
 
-static void status_timer_arm(void)
-{
-	struct itimerspec value;
-
-	value.it_value.tv_sec = 0;
-	value.it_value.tv_nsec = DISK_UTIL_MSEC * 1000000;
-	value.it_interval.tv_sec = 0;
-	value.it_interval.tv_nsec = DISK_UTIL_MSEC * 1000000;
-
-	timer_settime(ival_timer, 0, &value, NULL);
-}
-
-static void ival_fn(union sigval sig)
-{
-	if (threads) {
-		update_io_ticks();
-		print_thread_status();
-	}
-}
-
-/*
- * 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);
 	}
 }
 
-static void posix_timer_teardown(void)
+static void *disk_thread_main(void *data)
 {
-	timer_delete(ival_timer);
+	fio_mutex_up(startup_mutex);
+
+	while (threads) {
+		usleep(DISK_UTIL_MSEC * 1000);
+		if (!threads)
+			break;
+		update_io_ticks();
+
+		if (is_backend)
+			fio_server_send_status();
+		else
+			print_thread_status();
+	}
+
+	return NULL;
 }
 
-static void posix_timer_setup(void)
+static int create_disk_util_thread(void)
 {
-	struct sigevent evt;
+	int ret;
 
-	memset(&evt, 0, sizeof(evt));
-	evt.sigev_notify = SIGEV_THREAD;
-	evt.sigev_notify_function = ival_fn;
+	ret = pthread_create(&disk_util_thread, NULL, disk_thread_main, NULL);
+	if (ret) {
+		log_err("Can't create disk util thread: %s\n", strerror(ret));
+		return 1;
+	}
 
-	if (timer_create(FIO_TIMER_CLOCK, &evt, &ival_timer) < 0)
-		perror("timer_create");
+	ret = pthread_detach(disk_util_thread);
+	if (ret) {
+		log_err("Can't detatch disk util thread: %s\n", strerror(ret));
+		return 1;
+	}
 
+	dprint(FD_MUTEX, "wait on startup_mutex\n");
+	fio_mutex_down(startup_mutex);
+	dprint(FD_MUTEX, "done waiting on startup_mutex\n");
+	return 0;
 }
 
 static void set_sig_handlers(void)
@@ -176,7 +181,7 @@ 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 +190,7 @@ 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 +382,15 @@ 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 +476,11 @@ 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);
@@ -497,7 +510,6 @@ static void do_verify(struct thread_data *td)
 				clear_io_u(td, io_u);
 			} else if (io_u->resid) {
 				int bytes = io_u->xfer_buflen - io_u->resid;
-				struct fio_file *f = io_u->file;
 
 				/*
 				 * zero read, fail
@@ -515,6 +527,7 @@ static void do_verify(struct thread_data *td)
 				if (ddir_rw(io_u->ddir))
 					td->ts.short_io_u[io_u->ddir]++;
 
+				f = io_u->file;
 				if (io_u->offset == f->real_file_size)
 					goto sync_done;
 
@@ -545,13 +558,14 @@ 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 +626,11 @@ 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 +719,15 @@ 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 +784,11 @@ 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 +818,7 @@ 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 +865,9 @@ 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;
 		}
 
@@ -857,14 +880,14 @@ 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.
 				 */
-				fill_pattern(td, io_u->buf, max_bs, io_u);
+				fill_pattern(td, io_u->buf, max_bs, io_u, 0, 0);
 			}
 		}
 
@@ -955,7 +978,7 @@ static int keep_running(struct thread_data *td)
 
 static void reset_io_counters(struct thread_data *td)
 {
-	td->ts.stat_io_bytes[0] = td->ts.stat_io_bytes[1] = 0;
+	td->stat_io_bytes[0] = td->stat_io_bytes[1] = 0;
 	td->this_io_bytes[0] = td->this_io_bytes[1] = 0;
 	td->zone_bytes = 0;
 	td->rate_bytes[0] = td->rate_bytes[1] = 0;
@@ -968,11 +991,6 @@ 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)
@@ -988,7 +1006,7 @@ void reset_all_stats(struct thread_data *td)
 		td->io_issues[i] = 0;
 		td->ts.total_io_u[i] = 0;
 	}
-	
+
 	fio_gettime(&tv, NULL);
 	td->ts.runtime[0] = 0;
 	td->ts.runtime[1] = 0;
@@ -1006,6 +1024,11 @@ 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 +1058,11 @@ 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);
 
@@ -1069,12 +1093,32 @@ static void *thread_main(void *data)
 	 */
 	fio_mutex_remove(td->mutex);
 
+	/*
+	 * A new gid requires privilege, so we need to do this before setting
+	 * the uid.
+	 */
+	if (td->o.gid != -1U && setgid(td->o.gid)) {
+		td_verror(td, errno, "setgid");
+		goto err;
+	}
 	if (td->o.uid != -1U && setuid(td->o.uid)) {
 		td_verror(td, errno, "setuid");
 		goto err;
 	}
-	if (td->o.gid != -1U && setgid(td->o.gid)) {
-		td_verror(td, errno, "setgid");
+
+	/*
+	 * 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;
 	}
 
@@ -1091,23 +1135,6 @@ 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");
@@ -1146,19 +1173,17 @@ static void *thread_main(void *data)
 	}
 
 	fio_gettime(&td->epoch, NULL);
-	getrusage(RUSAGE_SELF, &td->ts.ru_start);
+	getrusage(RUSAGE_SELF, &td->ru_start);
 
 	clear_state = 0;
 	while (keep_running(td)) {
 		fio_gettime(&td->start, NULL);
-		memcpy(&td->ts.stat_sample_time[0], &td->start,
-				sizeof(td->start));
-		memcpy(&td->ts.stat_sample_time[1], &td->start,
-				sizeof(td->start));
+		memcpy(&td->stat_sample_time[0], &td->start, sizeof(td->start));
+		memcpy(&td->stat_sample_time[1], &td->start, sizeof(td->start));
 		memcpy(&td->tv_cache, &td->start, sizeof(td->start));
 
 		if (td->o.ratemin[0] || td->o.ratemin[1])
-			memcpy(&td->lastrate, &td->ts.stat_sample_time,
+			memcpy(&td->lastrate, &td->stat_sample_time,
 							sizeof(td->lastrate));
 
 		if (clear_state)
@@ -1207,33 +1232,33 @@ static void *thread_main(void *data)
 	td->ts.io_bytes[1] = td->io_bytes[1];
 
 	fio_mutex_down(writeout_mutex);
-	if (td->ts.bw_log) {
+	if (td->bw_log) {
 		if (td->o.bw_log_file) {
-			finish_log_named(td, td->ts.bw_log,
+			finish_log_named(td, td->bw_log,
 						td->o.bw_log_file, "bw");
 		} else
-			finish_log(td, td->ts.bw_log, "bw");
+			finish_log(td, td->bw_log, "bw");
 	}
-	if (td->ts.lat_log) {
+	if (td->lat_log) {
 		if (td->o.lat_log_file) {
-			finish_log_named(td, td->ts.lat_log,
+			finish_log_named(td, td->lat_log,
 						td->o.lat_log_file, "lat");
 		} else
-			finish_log(td, td->ts.lat_log, "lat");
+			finish_log(td, td->lat_log, "lat");
 	}
-	if (td->ts.slat_log) {
+	if (td->slat_log) {
 		if (td->o.lat_log_file) {
-			finish_log_named(td, td->ts.slat_log,
+			finish_log_named(td, td->slat_log,
 						td->o.lat_log_file, "slat");
 		} else
-			finish_log(td, td->ts.slat_log, "slat");
+			finish_log(td, td->slat_log, "slat");
 	}
-	if (td->ts.clat_log) {
+	if (td->clat_log) {
 		if (td->o.lat_log_file) {
-			finish_log_named(td, td->ts.clat_log,
+			finish_log_named(td, td->clat_log,
 						td->o.lat_log_file, "clat");
 		} else
-			finish_log(td, td->ts.clat_log, "clat");
+			finish_log(td, td->clat_log, "clat");
 	}
 	fio_mutex_up(writeout_mutex);
 	if (td->o.exec_postrun)
@@ -1281,6 +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;
@@ -1288,6 +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);
@@ -1474,10 +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,
@@ -1493,26 +1523,24 @@ static void run_threads(void)
 			todo--;
 		} else {
 			struct fio_file *f;
-			unsigned int i;
+			unsigned int j;
 
 			/*
 			 * for sharing to work, each job must always open
 			 * its own files. so close them, if we opened them
 			 * for creation
 			 */
-			for_each_file(td, f, i) {
+			for_each_file(td, f, j) {
 				if (fio_file_open(f))
 					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;
 
@@ -1545,6 +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.
@@ -1664,34 +1694,10 @@ 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 (nr_clients)
+		return fio_handle_clients();
 	if (exec_profile && load_profile(exec_profile))
 		return 1;
 
@@ -1711,9 +1717,7 @@ int main(int argc, char *argv[])
 		return 1;
 
 	set_genesis_time();
-
-	posix_timer_setup();
-	status_timer_arm();
+	create_disk_util_thread();
 
 	cgroup_list = smalloc(sizeof(*cgroup_list));
 	INIT_FLIST_HEAD(cgroup_list);
@@ -1733,8 +1737,48 @@ int main(int argc, char *argv[])
 	sfree(cgroup_list);
 	sfree(cgroup_mnt);
 
-	posix_timer_teardown();
 	fio_mutex_remove(startup_mutex);
 	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();
+}