X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=fio.c;h=8cc8fcf3361be69e39bd4003a8d93e8ec81466a2;hp=f6ed89f2cc7fc78644ded12a3090e3083754d4a8;hb=dfc6e7513c82d83491e90b8583fb52e301053d1b;hpb=8a4f8a24a23a1f95ce28338d697c9b552e7e06b4

diff --git a/fio.c b/fio.c
index f6ed89f2..8cc8fcf3 100644
--- a/fio.c
+++ b/fio.c
@@ -28,6 +28,7 @@
 #include <time.h>
 #include <locale.h>
 #include <assert.h>
+#include <time.h>
 #include <sys/stat.h>
 #include <sys/wait.h>
 #include <sys/ipc.h>
@@ -38,7 +39,11 @@
 #include "hash.h"
 #include "smalloc.h"
 #include "verify.h"
+#include "trim.h"
 #include "diskutil.h"
+#include "cgroup.h"
+#include "profile.h"
+#include "lib/rand.h"
 
 unsigned long page_mask;
 unsigned long page_size;
@@ -58,8 +63,10 @@ static struct fio_mutex *startup_mutex;
 static struct fio_mutex *writeout_mutex;
 static volatile int fio_abort;
 static int exit_value;
-static struct itimerval itimer;
+static timer_t ival_timer;
 static pthread_t gtod_thread;
+static struct flist_head *cgroup_list;
+static char *cgroup_mnt;
 
 struct io_log *agg_io_log[2];
 
@@ -107,17 +114,21 @@ static void terminate_threads(int group_id)
 
 static void status_timer_arm(void)
 {
-	itimer.it_value.tv_sec = 0;
-	itimer.it_value.tv_usec = DISK_UTIL_MSEC * 1000;
-	setitimer(ITIMER_REAL, &itimer, NULL);
+	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 sig_alrm(int fio_unused sig)
+static void ival_fn(union sigval sig)
 {
 	if (threads) {
 		update_io_ticks();
 		print_thread_status();
-		status_timer_arm();
 	}
 }
 
@@ -131,42 +142,38 @@ static void sig_quit(int sig)
 static void sig_int(int sig)
 {
 	if (threads) {
-		printf("\nfio: terminating on signal %d\n", sig);
+		log_info("\nfio: terminating on signal %d\n", sig);
 		fflush(stdout);
 		terminate_threads(TERMINATE_ALL);
 	}
 }
 
-static void sig_ill(int fio_unused sig)
+static void posix_timer_teardown(void)
 {
-	if (!threads)
-		return;
+	timer_delete(ival_timer);
+}
+
+static void posix_timer_setup(void)
+{
+	struct sigevent evt;
+
+	memset(&evt, 0, sizeof(evt));
+	evt.sigev_notify = SIGEV_THREAD;
+	evt.sigev_notify_function = ival_fn;
 
-	log_err("fio: illegal instruction. your cpu does not support "
-		"the sse4.2 instruction for crc32c\n");
-	terminate_threads(TERMINATE_ALL);
-	exit(4);
+	if (timer_create(CLOCK_MONOTONIC, &evt, &ival_timer) < 0)
+		perror("timer_create");
 }
 
 static void set_sig_handlers(void)
 {
 	struct sigaction act;
 
-	memset(&act, 0, sizeof(act));
-	act.sa_handler = sig_alrm;
-	act.sa_flags = SA_RESTART;
-	sigaction(SIGALRM, &act, NULL);
-
 	memset(&act, 0, sizeof(act));
 	act.sa_handler = sig_int;
 	act.sa_flags = SA_RESTART;
 	sigaction(SIGINT, &act, NULL);
 
-	memset(&act, 0, sizeof(act));
-	act.sa_handler = sig_ill;
-	act.sa_flags = SA_RESTART;
-	sigaction(SIGILL, &act, NULL);
-
 	memset(&act, 0, sizeof(act));
 	act.sa_handler = sig_quit;
 	act.sa_flags = SA_RESTART;
@@ -187,6 +194,8 @@ static int __check_min_rate(struct thread_data *td, struct timeval *now,
 	unsigned int rate_iops = 0;
 	unsigned int rate_iops_min = 0;
 
+	assert(ddir_rw(ddir));
+
 	if (!td->o.ratemin[ddir] && !td->o.rate_iops_min[ddir])
 		return 0;
 
@@ -223,7 +232,7 @@ static int __check_min_rate(struct thread_data *td, struct timeval *now,
 				if (rate < ratemin ||
 				    bytes < td->rate_bytes[ddir]) {
 					log_err("%s: min rate %u not met, got"
-						" %luKiB/sec\n", td->o.name,
+						" %luKB/sec\n", td->o.name,
 							ratemin, rate);
 					return 1;
 				}
@@ -429,6 +438,8 @@ static void do_verify(struct thread_data *td)
 	int ret, min_events;
 	unsigned int i;
 
+	dprint(FD_VERIFY, "starting loop\n");
+
 	/*
 	 * sync io first and invalidate cache, to make sure we really
 	 * read from disk.
@@ -472,7 +483,10 @@ static void do_verify(struct thread_data *td)
 			break;
 		}
 
-		io_u->end_io = verify_io_u;
+		if (td->o.verify_async)
+			io_u->end_io = verify_io_u_async;
+		else
+			io_u->end_io = verify_io_u;
 
 		ret = td_io_queue(td, io_u);
 		switch (ret) {
@@ -497,7 +511,8 @@ static void do_verify(struct thread_data *td)
 				io_u->xfer_buf += bytes;
 				io_u->offset += bytes;
 
-				td->ts.short_io_u[io_u->ddir]++;
+				if (ddir_rw(io_u->ddir))
+					td->ts.short_io_u[io_u->ddir]++;
 
 				if (io_u->offset == f->real_file_size)
 					goto sync_done;
@@ -533,7 +548,8 @@ sync_done:
 		 */
 		full = queue_full(td) || ret == FIO_Q_BUSY;
 		if (full || !td->o.iodepth_batch_complete) {
-			min_events = td->o.iodepth_batch_complete;
+			min_events = min(td->o.iodepth_batch_complete,
+					 td->cur_depth);
 			if (full && !min_events)
 				min_events = 1;
 
@@ -562,6 +578,8 @@ sync_done:
 		cleanup_pending_aio(td);
 
 	td_set_runstate(td, TD_RUNNING);
+
+	dprint(FD_VERIFY, "exiting loop\n");
 }
 
 /*
@@ -578,7 +596,9 @@ static void do_io(struct thread_data *td)
 	else
 		td_set_runstate(td, TD_RUNNING);
 
-	while ((td->this_io_bytes[0] + td->this_io_bytes[1]) < td->o.size) {
+	while ( (td->o.read_iolog_file && !flist_empty(&td->io_log_list)) ||
+		(!flist_empty(&td->trim_list)) ||
+	        ((td->this_io_bytes[0] + td->this_io_bytes[1]) < td->o.size) ) {
 		struct timeval comp_time;
 		unsigned long bytes_done[2] = { 0, 0 };
 		int min_evts = 0;
@@ -603,8 +623,12 @@ static void do_io(struct thread_data *td)
 		 * Add verification end_io handler, if asked to verify
 		 * a previously written file.
 		 */
-		if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_READ) {
-			io_u->end_io = verify_io_u;
+		if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_READ &&
+		    !td_rw(td)) {
+			if (td->o.verify_async)
+				io_u->end_io = verify_io_u_async;
+			else
+				io_u->end_io = verify_io_u;
 			td_set_runstate(td, TD_VERIFYING);
 		} else if (in_ramp_time(td))
 			td_set_runstate(td, TD_RAMP);
@@ -634,7 +658,8 @@ static void do_io(struct thread_data *td)
 				io_u->xfer_buf += bytes;
 				io_u->offset += bytes;
 
-				td->ts.short_io_u[io_u->ddir]++;
+				if (ddir_rw(io_u->ddir))
+					td->ts.short_io_u[io_u->ddir]++;
 
 				if (io_u->offset == f->real_file_size)
 					goto sync_done;
@@ -680,7 +705,8 @@ sync_done:
 		 */
 		full = queue_full(td) || ret == FIO_Q_BUSY;
 		if (full || !td->o.iodepth_batch_complete) {
-			min_evts = td->o.iodepth_batch_complete;
+			min_evts = min(td->o.iodepth_batch_complete,
+					td->cur_depth);
 			if (full && !min_evts)
 				min_evts = 1;
 
@@ -727,6 +753,9 @@ sync_done:
 		}
 	}
 
+	if (td->trim_entries)
+		printf("trim entries %ld\n", td->trim_entries);
+
 	if (td->o.fill_device && td->error == ENOSPC) {
 		td->error = 0;
 		td->terminate = 1;
@@ -799,7 +828,8 @@ static int init_io_u(struct thread_data *td)
 	if (allocate_io_mem(td))
 		return 1;
 
-	if (td->o.mem_align)
+	if (td->o.odirect || td->o.mem_align ||
+	    (td->io_ops->flags & FIO_RAWIO))
 		p = PAGE_ALIGN(td->orig_buffer) + td->o.mem_align;
 	else
 		p = td->orig_buffer;
@@ -828,6 +858,13 @@ static int init_io_u(struct thread_data *td)
 
 			if (td_write(td) && !td->o.refill_buffers)
 				io_u_fill_buffer(td, io_u, max_bs);
+			else 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);
+			}
 		}
 
 		io_u->index = i;
@@ -952,6 +989,8 @@ void reset_all_stats(struct thread_data *td)
 	}
 	
 	fio_gettime(&tv, NULL);
+	td->ts.runtime[0] = 0;
+	td->ts.runtime[1] = 0;
 	memcpy(&td->epoch, &tv, sizeof(tv));
 	memcpy(&td->start, &tv, sizeof(tv));
 }
@@ -990,8 +1029,9 @@ static int exec_string(const char *string)
  */
 static void *thread_main(void *data)
 {
-	unsigned long long runtime[2], elapsed;
+	unsigned long long elapsed;
 	struct thread_data *td = data;
+	pthread_condattr_t attr;
 	int clear_state;
 
 	if (!td->o.use_thread)
@@ -1006,8 +1046,15 @@ static void *thread_main(void *data)
 	INIT_FLIST_HEAD(&td->io_u_requeues);
 	INIT_FLIST_HEAD(&td->io_log_list);
 	INIT_FLIST_HEAD(&td->io_hist_list);
+	INIT_FLIST_HEAD(&td->verify_list);
+	INIT_FLIST_HEAD(&td->trim_list);
+	pthread_mutex_init(&td->io_u_lock, NULL);
 	td->io_hist_tree = RB_ROOT;
 
+	pthread_condattr_init(&attr);
+	pthread_cond_init(&td->verify_cond, &attr);
+	pthread_cond_init(&td->free_cond, &attr);
+
 	td_set_runstate(td, TD_INITIALIZED);
 	dprint(FD_MUTEX, "up startup_mutex\n");
 	fio_mutex_up(startup_mutex);
@@ -1021,6 +1068,19 @@ 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;
+	}
+
 	/*
 	 * May alter parameters that init_io_u() will use, so we need to
 	 * do this first.
@@ -1031,7 +1091,10 @@ static void *thread_main(void *data)
 	if (init_io_u(td))
 		goto err;
 
-	if (td->o.cpumask_set && fio_setaffinity(td) == -1) {
+	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;
 	}
@@ -1042,7 +1105,7 @@ static void *thread_main(void *data)
 	 */
 	if (td->o.gtod_cpu) {
 		fio_cpu_clear(&td->o.cpumask, td->o.gtod_cpu);
-		if (fio_setaffinity(td) == -1) {
+		if (fio_setaffinity(td->pid, td->o.cpumask) == -1) {
 			td_verror(td, errno, "cpu_set_affinity");
 			goto err;
 		}
@@ -1055,6 +1118,9 @@ static void *thread_main(void *data)
 		}
 	}
 
+	if (td->o.cgroup_weight && cgroup_setup(td, cgroup_list, &cgroup_mnt))
+		goto err;
+
 	if (nice(td->o.nice) == -1) {
 		td_verror(td, errno, "nice");
 		goto err;
@@ -1085,11 +1151,13 @@ static void *thread_main(void *data)
 	fio_gettime(&td->epoch, NULL);
 	getrusage(RUSAGE_SELF, &td->ts.ru_start);
 
-	runtime[0] = runtime[1] = 0;
 	clear_state = 0;
 	while (keep_running(td)) {
 		fio_gettime(&td->start, NULL);
-		memcpy(&td->ts.stat_sample_time, &td->start, sizeof(td->start));
+		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->tv_cache, &td->start, sizeof(td->start));
 
 		if (td->o.ratemin[0] || td->o.ratemin[1])
@@ -1107,11 +1175,11 @@ static void *thread_main(void *data)
 
 		if (td_read(td) && td->io_bytes[DDIR_READ]) {
 			elapsed = utime_since_now(&td->start);
-			runtime[DDIR_READ] += elapsed;
+			td->ts.runtime[DDIR_READ] += elapsed;
 		}
 		if (td_write(td) && td->io_bytes[DDIR_WRITE]) {
 			elapsed = utime_since_now(&td->start);
-			runtime[DDIR_WRITE] += elapsed;
+			td->ts.runtime[DDIR_WRITE] += elapsed;
 		}
 
 		if (td->error || td->terminate)
@@ -1128,15 +1196,15 @@ static void *thread_main(void *data)
 
 		do_verify(td);
 
-		runtime[DDIR_READ] += utime_since_now(&td->start);
+		td->ts.runtime[DDIR_READ] += utime_since_now(&td->start);
 
 		if (td->error || td->terminate)
 			break;
 	}
 
 	update_rusage_stat(td);
-	td->ts.runtime[0] = (runtime[0] + 999) / 1000;
-	td->ts.runtime[1] = (runtime[1] + 999) / 1000;
+	td->ts.runtime[0] = (td->ts.runtime[0] + 999) / 1000;
+	td->ts.runtime[1] = (td->ts.runtime[1] + 999) / 1000;
 	td->ts.total_run_time = mtime_since_now(&td->epoch);
 	td->ts.io_bytes[0] = td->io_bytes[0];
 	td->ts.io_bytes[1] = td->io_bytes[1];
@@ -1149,6 +1217,13 @@ static void *thread_main(void *data)
 		} else
 			finish_log(td, td->ts.bw_log, "bw");
 	}
+	if (td->ts.lat_log) {
+		if (td->o.lat_log_file) {
+			finish_log_named(td, td->ts.lat_log,
+						td->o.lat_log_file, "lat");
+		} else
+			finish_log(td, td->ts.lat_log, "lat");
+	}
 	if (td->ts.slat_log) {
 		if (td->o.lat_log_file) {
 			finish_log_named(td, td->ts.slat_log,
@@ -1172,11 +1247,16 @@ static void *thread_main(void *data)
 
 err:
 	if (td->error)
-		printf("fio: pid=%d, err=%d/%s\n", (int) td->pid, td->error,
+		log_info("fio: pid=%d, err=%d/%s\n", (int) td->pid, td->error,
 							td->verror);
+
+	if (td->o.verify_async)
+		verify_async_exit(td);
+
 	close_and_free_files(td);
 	close_ioengine(td);
 	cleanup_io_u(td);
+	cgroup_shutdown(td, &cgroup_mnt);
 
 	if (td->o.cpumask_set) {
 		int ret = fio_cpuset_exit(&td->o.cpumask);
@@ -1333,9 +1413,13 @@ static void *gtod_thread_main(void *data)
 
 static int fio_start_gtod_thread(void)
 {
+	pthread_attr_t attr;
 	int ret;
 
-	ret = pthread_create(&gtod_thread, NULL, gtod_thread_main, NULL);
+	pthread_attr_init(&attr);
+	pthread_attr_setstacksize(&attr, PTHREAD_STACK_MIN);
+	ret = pthread_create(&gtod_thread, &attr, gtod_thread_main, NULL);
+	pthread_attr_destroy(&attr);
 	if (ret) {
 		log_err("Can't create gtod thread: %s\n", strerror(ret));
 		return 1;
@@ -1369,17 +1453,17 @@ static void run_threads(void)
 		return;
 
 	if (!terse_output) {
-		printf("Starting ");
+		log_info("Starting ");
 		if (nr_thread)
-			printf("%d thread%s", nr_thread,
+			log_info("%d thread%s", nr_thread,
 						nr_thread > 1 ? "s" : "");
 		if (nr_process) {
 			if (nr_thread)
 				printf(" and ");
-			printf("%d process%s", nr_process,
+			log_info("%d process%s", nr_process,
 						nr_process > 1 ? "es" : "");
 		}
-		printf("\n");
+		log_info("\n");
 		fflush(stdout);
 	}
 
@@ -1500,7 +1584,13 @@ static void run_threads(void)
 					*fio_debug_jobp = pid;
 			}
 			dprint(FD_MUTEX, "wait on startup_mutex\n");
-			fio_mutex_down(startup_mutex);
+			if (fio_mutex_down_timeout(startup_mutex, 10)) {
+				log_err("fio: job startup hung? exiting.\n");
+				terminate_threads(TERMINATE_ALL);
+				fio_abort = 1;
+				nr_started--;
+				break;
+			}
 			dprint(FD_MUTEX, "done waiting on startup_mutex\n");
 		}
 
@@ -1582,6 +1672,7 @@ int main(int argc, char *argv[])
 	long ps;
 
 	sinit();
+	init_rand(&__fio_rand_state);
 
 	/*
 	 * We need locale for number printing, if it isn't set then just
@@ -1590,12 +1681,6 @@ int main(int argc, char *argv[])
 	if (!getenv("LC_NUMERIC"))
 		setlocale(LC_NUMERIC, "en_US");
 
-	if (parse_options(argc, argv))
-		return 1;
-
-	if (!thread_number)
-		return 0;
-
 	ps = sysconf(_SC_PAGESIZE);
 	if (ps < 0) {
 		log_err("Failed to get page size\n");
@@ -1605,6 +1690,17 @@ int main(int argc, char *argv[])
 	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;
+
+	if (!thread_number)
+		return 0;
+
 	if (write_bw_log) {
 		setup_log(&agg_io_log[DDIR_READ]);
 		setup_log(&agg_io_log[DDIR_WRITE]);
@@ -1615,8 +1711,12 @@ int main(int argc, char *argv[])
 
 	set_genesis_time();
 
+	posix_timer_setup();
 	status_timer_arm();
 
+	cgroup_list = smalloc(sizeof(*cgroup_list));
+	INIT_FLIST_HEAD(cgroup_list);
+
 	run_threads();
 
 	if (!fio_abort) {
@@ -1628,6 +1728,11 @@ int main(int argc, char *argv[])
 		}
 	}
 
+	cgroup_kill(cgroup_list);
+	sfree(cgroup_list);
+	sfree(cgroup_mnt);
+
+	posix_timer_teardown();
 	fio_mutex_remove(startup_mutex);
 	fio_mutex_remove(writeout_mutex);
 	return exit_value;