X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=client.c;h=5dcc5f5410bf712e1947b1e66391eaab7335a545;hp=af1dba44ff77d8f8d0d11b6d90fda29521a2b8cf;hb=bc0fec0e12f19dd424f4bf83cfca89d434184c8d;hpb=2f99deb0895b2d0b618e6e72c74add7192b4d73f

diff --git a/client.c b/client.c
index af1dba44..637cd3fb 100644
--- a/client.c
+++ b/client.c
@@ -14,12 +14,16 @@
 #include <arpa/inet.h>
 #include <netdb.h>
 #include <signal.h>
+#ifdef CONFIG_ZLIB
+#include <zlib.h>
+#endif
 
 #include "fio.h"
 #include "client.h"
 #include "server.h"
 #include "flist.h"
 #include "hash.h"
+#include "verify.h"
 
 static void handle_du(struct fio_client *client, struct fio_net_cmd *cmd);
 static void handle_ts(struct fio_client *client, struct fio_net_cmd *cmd);
@@ -27,15 +31,19 @@ static void handle_gs(struct fio_client *client, struct fio_net_cmd *cmd);
 static void handle_probe(struct fio_client *client, struct fio_net_cmd *cmd);
 static void handle_text(struct fio_client *client, struct fio_net_cmd *cmd);
 static void handle_stop(struct fio_client *client, struct fio_net_cmd *cmd);
+static void handle_start(struct fio_client *client, struct fio_net_cmd *cmd);
 
 struct client_ops fio_client_ops = {
-	.text_op	= handle_text,
+	.text		= handle_text,
 	.disk_util	= handle_du,
 	.thread_status	= handle_ts,
 	.group_stats	= handle_gs,
 	.stop		= handle_stop,
+	.start		= handle_start,
 	.eta		= display_thread_status,
 	.probe		= handle_probe,
+	.eta_msec	= FIO_CLIENT_DEF_ETA_MSEC,
+	.client_type	= FIO_CLIENT_TYPE_CLI,
 };
 
 static struct timeval eta_tv;
@@ -50,6 +58,12 @@ struct group_run_stats client_gs;
 int sum_stat_clients;
 
 static int sum_stat_nr;
+static struct json_object *root = NULL;
+static struct json_object *job_opt_object = NULL;
+static struct json_array *clients_array = NULL;
+static struct json_array *du_array = NULL;
+
+static int error_clients;
 
 #define FIO_CLIENT_HASH_BITS	7
 #define FIO_CLIENT_HASH_SZ	(1 << FIO_CLIENT_HASH_BITS)
@@ -78,6 +92,69 @@ static void fio_init fio_client_hash_init(void)
 		INIT_FLIST_HEAD(&client_hash[i]);
 }
 
+static int read_data(int fd, void *data, size_t size)
+{
+	ssize_t ret;
+
+	while (size) {
+		ret = read(fd, data, size);
+		if (ret < 0) {
+			if (errno == EAGAIN || errno == EINTR)
+				continue;
+			break;
+		} else if (!ret)
+			break;
+		else {
+			data += ret;
+			size -= ret;
+		}
+	}
+
+	if (size)
+		return EAGAIN;
+
+	return 0;
+}
+
+static void fio_client_json_init(void)
+{
+	char time_buf[32];
+	time_t time_p;
+
+	if (!(output_format & FIO_OUTPUT_JSON))
+		return;
+
+	time(&time_p);
+	os_ctime_r((const time_t *) &time_p, time_buf, sizeof(time_buf));
+	time_buf[strlen(time_buf) - 1] = '\0';
+
+	root = json_create_object();
+	json_object_add_value_string(root, "fio version", fio_version_string);
+	json_object_add_value_int(root, "timestamp", time_p);
+	json_object_add_value_string(root, "time", time_buf);
+
+	job_opt_object = json_create_object();
+	json_object_add_value_object(root, "global options", job_opt_object);
+	clients_array = json_create_array();
+	json_object_add_value_array(root, "client_stats", clients_array);
+	du_array = json_create_array();
+	json_object_add_value_array(root, "disk_util", du_array);
+}
+
+static void fio_client_json_fini(void)
+{
+	if (!(output_format & FIO_OUTPUT_JSON))
+		return;
+
+	log_info("\n");
+	json_print_object(root, NULL);
+	log_info("\n");
+	json_free_object(root);
+	root = NULL;
+	clients_array = NULL;
+	du_array = NULL;
+}
+
 static struct fio_client *find_client_by_fd(int fd)
 {
 	int bucket = hash_long(fd, FIO_CLIENT_HASH_BITS) & FIO_CLIENT_HASH_MASK;
@@ -96,15 +173,54 @@ static struct fio_client *find_client_by_fd(int fd)
 	return NULL;
 }
 
-static void remove_client(struct fio_client *client)
+void fio_put_client(struct fio_client *client)
 {
-	assert(client->refs);
-
 	if (--client->refs)
 		return;
 
+	free(client->hostname);
+	if (client->argv)
+		free(client->argv);
+	if (client->name)
+		free(client->name);
+	while (client->nr_files) {
+		struct client_file *cf = &client->files[--client->nr_files];
+
+		free(cf->file);
+	}
+	if (client->files)
+		free(client->files);
+	if (client->opt_lists)
+		free(client->opt_lists);
+
+	if (!client->did_stat)
+		sum_stat_clients--;
+
+	if (client->error)
+		error_clients++;
+
+	free(client);
+}
+
+static int fio_client_dec_jobs_eta(struct client_eta *eta, client_eta_op eta_fn)
+{
+	if (!--eta->pending) {
+		eta_fn(&eta->eta);
+		free(eta);
+		return 0;
+	}
+
+	return 1;
+}
+
+static void remove_client(struct fio_client *client)
+{
+	assert(client->refs);
+
 	dprint(FD_NET, "client: removed <%s>\n", client->hostname);
-	flist_del(&client->list);
+
+	if (!flist_empty(&client->list))
+		flist_del_init(&client->list);
 
 	fio_client_remove_hash(client);
 
@@ -113,20 +229,20 @@ static void remove_client(struct fio_client *client)
 		fio_client_dec_jobs_eta(client->eta_in_flight, client->ops->eta);
 	}
 
-	free(client->hostname);
-	if (client->argv)
-		free(client->argv);
-	if (client->name)
-		free(client->name);
+	close(client->fd);
+	client->fd = -1;
+
+	if (client->ops->removed)
+		client->ops->removed(client);
 
-	free(client);
 	nr_clients--;
-	sum_stat_clients--;
+	fio_put_client(client);
 }
 
-static void put_client(struct fio_client *client)
+struct fio_client *fio_get_client(struct fio_client *client)
 {
-	remove_client(client);
+	client->refs++;
+	return client;
 }
 
 static void __fio_client_add_cmd_option(struct fio_client *client,
@@ -183,7 +299,7 @@ struct fio_client *fio_client_add_explicit(struct client_ops *ops,
 		int ipv6;
 
 		ipv6 = type == Fio_client_ipv6;
-		if (fio_server_parse_host(hostname, &ipv6,
+		if (fio_server_parse_host(hostname, ipv6,
 						&client->addr.sin_addr,
 						&client->addr6.sin6_addr))
 			goto err;
@@ -194,6 +310,7 @@ struct fio_client *fio_client_add_explicit(struct client_ops *ops,
 	client->fd = -1;
 	client->ops = ops;
 	client->refs = 1;
+	client->type = ops->client_type;
 
 	__fio_client_add_cmd_option(client, "fio");
 
@@ -206,6 +323,31 @@ err:
 	return NULL;
 }
 
+int fio_client_add_ini_file(void *cookie, const char *ini_file, int remote)
+{
+	struct fio_client *client = cookie;
+	struct client_file *cf;
+	size_t new_size;
+	void *new_files;
+
+	if (!client)
+		return 1;
+
+	dprint(FD_NET, "client <%s>: add ini %s\n", client->hostname, ini_file);
+
+	new_size = (client->nr_files + 1) * sizeof(struct client_file);
+	new_files = realloc(client->files, new_size);
+	if (!new_files)
+		return 1;
+
+	client->files = new_files;
+	cf = &client->files[client->nr_files];
+	cf->file = strdup(ini_file);
+	cf->remote = remote;
+	client->nr_files++;
+	return 0;
+}
+
 int fio_client_add(struct client_ops *ops, const char *hostname, void **cookie)
 {
 	struct fio_client *existing = *cookie;
@@ -243,6 +385,7 @@ int fio_client_add(struct client_ops *ops, const char *hostname, void **cookie)
 	client->fd = -1;
 	client->ops = ops;
 	client->refs = 1;
+	client->type = ops->client_type;
 
 	__fio_client_add_cmd_option(client, "fio");
 
@@ -253,10 +396,47 @@ int fio_client_add(struct client_ops *ops, const char *hostname, void **cookie)
 	return 0;
 }
 
+static const char *server_name(struct fio_client *client, char *buf,
+			       size_t bufsize)
+{
+	const char *from;
+
+	if (client->ipv6)
+		from = inet_ntop(AF_INET6, (struct sockaddr *) &client->addr6.sin6_addr, buf, bufsize);
+	else if (client->is_sock)
+		from = "sock";
+	else
+		from = inet_ntop(AF_INET, (struct sockaddr *) &client->addr.sin_addr, buf, bufsize);
+
+	return from;
+}
+
+static void probe_client(struct fio_client *client)
+{
+	struct cmd_client_probe_pdu pdu;
+	const char *sname;
+	uint64_t tag;
+	char buf[64];
+
+	dprint(FD_NET, "client: send probe\n");
+
+#ifdef CONFIG_ZLIB
+	pdu.flags = __le64_to_cpu(FIO_PROBE_FLAG_ZLIB);
+#else
+	pdu.flags = 0;
+#endif
+
+	sname = server_name(client, buf, sizeof(buf));
+	memset(pdu.server, 0, sizeof(pdu.server));
+	strncpy((char *) pdu.server, sname, sizeof(pdu.server) - 1);
+
+	fio_net_send_cmd(client->fd, FIO_NET_CMD_PROBE, &pdu, sizeof(pdu), &tag, &client->cmd_list);
+}
+
 static int fio_client_connect_ip(struct fio_client *client)
 {
 	struct sockaddr *addr;
-	fio_socklen_t socklen;
+	socklen_t socklen;
 	int fd, domain;
 
 	if (client->ipv6) {
@@ -275,16 +455,20 @@ static int fio_client_connect_ip(struct fio_client *client)
 
 	fd = socket(domain, SOCK_STREAM, 0);
 	if (fd < 0) {
+		int ret = -errno;
+
 		log_err("fio: socket: %s\n", strerror(errno));
-		return -1;
+		return ret;
 	}
 
 	if (connect(fd, addr, socklen) < 0) {
+		int ret = -errno;
+
 		log_err("fio: connect: %s\n", strerror(errno));
 		log_err("fio: failed to connect to %s:%u\n", client->hostname,
 								client->port);
 		close(fd);
-		return -1;
+		return ret;
 	}
 
 	return fd;
@@ -293,24 +477,28 @@ static int fio_client_connect_ip(struct fio_client *client)
 static int fio_client_connect_sock(struct fio_client *client)
 {
 	struct sockaddr_un *addr = &client->addr_un;
-	fio_socklen_t len;
+	socklen_t len;
 	int fd;
 
 	memset(addr, 0, sizeof(*addr));
 	addr->sun_family = AF_UNIX;
-	strcpy(addr->sun_path, client->hostname);
+	strncpy(addr->sun_path, client->hostname, sizeof(addr->sun_path) - 1);
 
 	fd = socket(AF_UNIX, SOCK_STREAM, 0);
 	if (fd < 0) {
+		int ret = -errno;
+
 		log_err("fio: socket: %s\n", strerror(errno));
-		return -1;
+		return ret;
 	}
 
 	len = sizeof(addr->sun_family) + strlen(addr->sun_path) + 1;
 	if (connect(fd, (struct sockaddr *) addr, len) < 0) {
+		int ret = -errno;
+
 		log_err("fio: connect; %s\n", strerror(errno));
 		close(fd);
-		return -1;
+		return ret;
 	}
 
 	return fd;
@@ -330,17 +518,19 @@ int fio_client_connect(struct fio_client *client)
 	dprint(FD_NET, "client: %s connected %d\n", client->hostname, fd);
 
 	if (fd < 0)
-		return 1;
+		return fd;
 
 	client->fd = fd;
 	fio_client_add_hash(client);
 	client->state = Client_connected;
+
+	probe_client(client);
 	return 0;
 }
 
-void fio_client_terminate(struct fio_client *client)
+int fio_client_terminate(struct fio_client *client)
 {
-	fio_net_send_simple_cmd(client->fd, FIO_NET_CMD_QUIT, 0, NULL);
+	return fio_net_send_quit(client->fd);
 }
 
 void fio_clients_terminate(void)
@@ -375,13 +565,19 @@ static void client_signal_handler(void)
 	act.sa_handler = sig_int;
 	act.sa_flags = SA_RESTART;
 	sigaction(SIGTERM, &act, NULL);
-}
 
-static void probe_client(struct fio_client *client)
-{
-	dprint(FD_NET, "client: send probe\n");
+/* Windows uses SIGBREAK as a quit signal from other applications */
+#ifdef WIN32
+	memset(&act, 0, sizeof(act));
+	act.sa_handler = sig_int;
+	act.sa_flags = SA_RESTART;
+	sigaction(SIGBREAK, &act, NULL);
+#endif
 
-	fio_net_send_simple_cmd(client->fd, FIO_NET_CMD_PROBE, 0, &client->cmd_list);
+	memset(&act, 0, sizeof(act));
+	act.sa_handler = sig_show_status;
+	act.sa_flags = SA_RESTART;
+	sigaction(SIGUSR1, &act, NULL);
 }
 
 static int send_client_cmd_line(struct fio_client *client)
@@ -426,7 +622,8 @@ static int send_client_cmd_line(struct fio_client *client)
 
 	free(lens);
 	clp->lines = cpu_to_le16(client->argc);
-	ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_JOBLINE, pdu, mem, 0);
+	clp->client_type = __cpu_to_le16(client->type);
+	ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_JOBLINE, pdu, mem, NULL, NULL);
 	free(pdu);
 	return ret;
 }
@@ -439,7 +636,7 @@ int fio_clients_connect(void)
 
 #ifdef WIN32
 	WSADATA wsd;
-	WSAStartup(MAKEWORD(2,2), &wsd);
+	WSAStartup(MAKEWORD(2, 2), &wsd);
 #endif
 
 	dprint(FD_NET, "client: connect all\n");
@@ -455,8 +652,6 @@ int fio_clients_connect(void)
 			continue;
 		}
 
-		probe_client(client);
-
 		if (client->argc > 1)
 			send_client_cmd_line(client);
 	}
@@ -478,6 +673,8 @@ int fio_start_all_clients(void)
 
 	dprint(FD_NET, "client: start all\n");
 
+	fio_client_json_init();
+
 	flist_for_each_safe(entry, tmp, &client_list) {
 		client = flist_entry(entry, struct fio_client, list);
 
@@ -491,14 +688,40 @@ int fio_start_all_clients(void)
 	return flist_empty(&client_list);
 }
 
+static int __fio_client_send_remote_ini(struct fio_client *client,
+					const char *filename)
+{
+	struct cmd_load_file_pdu *pdu;
+	size_t p_size;
+	int ret;
+
+	dprint(FD_NET, "send remote ini %s to %s\n", filename, client->hostname);
+
+	p_size = sizeof(*pdu) + strlen(filename) + 1;
+	pdu = malloc(p_size);
+	memset(pdu, 0, p_size);
+	pdu->name_len = strlen(filename);
+	strcpy((char *) pdu->file, filename);
+	pdu->client_type = cpu_to_le16((uint16_t) client->type);
+
+	client->sent_job = 1;
+	ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_LOAD_FILE, pdu, p_size,NULL, NULL);
+	free(pdu);
+	return ret;
+}
+
 /*
  * Send file contents to server backend. We could use sendfile(), but to remain
  * more portable lets just read/write the darn thing.
  */
-static int fio_client_send_ini(struct fio_client *client, const char *filename)
+static int __fio_client_send_local_ini(struct fio_client *client,
+				       const char *filename)
 {
+	struct cmd_job_pdu *pdu;
+	size_t p_size;
 	struct stat sb;
-	char *p, *buf;
+	char *p;
+	void *buf;
 	off_t len;
 	int fd, ret;
 
@@ -506,48 +729,63 @@ static int fio_client_send_ini(struct fio_client *client, const char *filename)
 
 	fd = open(filename, O_RDONLY);
 	if (fd < 0) {
+		ret = -errno;
 		log_err("fio: job file <%s> open: %s\n", filename, strerror(errno));
-		return 1;
+		return ret;
 	}
 
 	if (fstat(fd, &sb) < 0) {
+		ret = -errno;
 		log_err("fio: job file stat: %s\n", strerror(errno));
 		close(fd);
-		return 1;
+		return ret;
 	}
 
-	buf = malloc(sb.st_size);
+	p_size = sb.st_size + sizeof(*pdu);
+	pdu = malloc(p_size);
+	buf = pdu->buf;
 
 	len = sb.st_size;
 	p = buf;
-	do {
-		ret = read(fd, p, len);
-		if (ret > 0) {
-			len -= ret;
-			if (!len)
-				break;
-			p += ret;
-			continue;
-		} else if (!ret)
-			break;
-		else if (errno == EAGAIN || errno == EINTR)
-			continue;
-	} while (1);
-
-	if (len) {
+	if (read_data(fd, p, len)) {
 		log_err("fio: failed reading job file %s\n", filename);
 		close(fd);
-		free(buf);
+		free(pdu);
 		return 1;
 	}
 
+	pdu->buf_len = __cpu_to_le32(sb.st_size);
+	pdu->client_type = cpu_to_le32(client->type);
+
 	client->sent_job = 1;
-	ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_JOB, buf, sb.st_size, 0);
-	free(buf);
+	ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_JOB, pdu, p_size, NULL, NULL);
+	free(pdu);
 	close(fd);
 	return ret;
 }
 
+int fio_client_send_ini(struct fio_client *client, const char *filename,
+			int remote)
+{
+	int ret;
+
+	if (!remote)
+		ret = __fio_client_send_local_ini(client, filename);
+	else
+		ret = __fio_client_send_remote_ini(client, filename);
+
+	if (!ret)
+		client->sent_job = 1;
+
+	return ret;
+}
+
+static int fio_client_send_cf(struct fio_client *client,
+			      struct client_file *cf)
+{
+	return fio_client_send_ini(client, cf->file, cf->remote);
+}
+
 int fio_clients_send_ini(const char *filename)
 {
 	struct fio_client *client;
@@ -556,15 +794,41 @@ int fio_clients_send_ini(const char *filename)
 	flist_for_each_safe(entry, tmp, &client_list) {
 		client = flist_entry(entry, struct fio_client, list);
 
-		if (fio_client_send_ini(client, filename))
-			remove_client(client);
+		if (client->nr_files) {
+			int i;
 
-		client->sent_job = 1;
+			for (i = 0; i < client->nr_files; i++) {
+				struct client_file *cf;
+
+				cf = &client->files[i];
+
+				if (fio_client_send_cf(client, cf)) {
+					remove_client(client);
+					break;
+				}
+			}
+		}
+		if (client->sent_job)
+			continue;
+		if (!filename || fio_client_send_ini(client, filename, 0))
+			remove_client(client);
 	}
 
 	return !nr_clients;
 }
 
+int fio_client_update_options(struct fio_client *client,
+			      struct thread_options *o, uint64_t *tag)
+{
+	struct cmd_add_job_pdu pdu;
+
+	pdu.thread_number = cpu_to_le32(client->thread_number);
+	pdu.groupid = cpu_to_le32(client->groupid);
+	convert_thread_options_to_net(&pdu.top, o);
+
+	return fio_net_send_cmd(client->fd, FIO_NET_CMD_UPDATE_JOB, &pdu, sizeof(pdu), tag, &client->cmd_list);
+}
+
 static void convert_io_stat(struct io_stat *dst, struct io_stat *src)
 {
 	dst->max_val	= le64_to_cpu(src->max_val);
@@ -582,12 +846,14 @@ static void convert_ts(struct thread_stat *dst, struct thread_stat *src)
 {
 	int i, j;
 
-	dst->error	= le32_to_cpu(src->error);
-	dst->groupid	= le32_to_cpu(src->groupid);
-	dst->pid	= le32_to_cpu(src->pid);
-	dst->members	= le32_to_cpu(src->members);
+	dst->error		= le32_to_cpu(src->error);
+	dst->thread_number	= le32_to_cpu(src->thread_number);
+	dst->groupid		= le32_to_cpu(src->groupid);
+	dst->pid		= le32_to_cpu(src->pid);
+	dst->members		= le32_to_cpu(src->members);
+	dst->unified_rw_rep	= le32_to_cpu(src->unified_rw_rep);
 
-	for (i = 0; i < 2; i++) {
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 		convert_io_stat(&dst->clat_stat[i], &src->clat_stat[i]);
 		convert_io_stat(&dst->slat_stat[i], &src->slat_stat[i]);
 		convert_io_stat(&dst->lat_stat[i], &src->lat_stat[i]);
@@ -600,6 +866,7 @@ static void convert_ts(struct thread_stat *dst, struct thread_stat *src)
 	dst->minf		= le64_to_cpu(src->minf);
 	dst->majf		= le64_to_cpu(src->majf);
 	dst->clat_percentiles	= le64_to_cpu(src->clat_percentiles);
+	dst->percentile_precision = le64_to_cpu(src->percentile_precision);
 
 	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++) {
 		fio_fp64_t *fps = &src->percentile_list[i];
@@ -619,19 +886,20 @@ static void convert_ts(struct thread_stat *dst, struct thread_stat *src)
 		dst->io_u_lat_m[i]	= le32_to_cpu(src->io_u_lat_m[i]);
 	}
 
-	for (i = 0; i < 2; i++)
+	for (i = 0; i < DDIR_RWDIR_CNT; i++)
 		for (j = 0; j < FIO_IO_U_PLAT_NR; j++)
 			dst->io_u_plat[i][j] = le32_to_cpu(src->io_u_plat[i][j]);
 
-	for (i = 0; i < 3; i++) {
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 		dst->total_io_u[i]	= le64_to_cpu(src->total_io_u[i]);
 		dst->short_io_u[i]	= le64_to_cpu(src->short_io_u[i]);
+		dst->drop_io_u[i]	= le64_to_cpu(src->drop_io_u[i]);
 	}
 
 	dst->total_submit	= le64_to_cpu(src->total_submit);
 	dst->total_complete	= le64_to_cpu(src->total_complete);
 
-	for (i = 0; i < 2; i++) {
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 		dst->io_bytes[i]	= le64_to_cpu(src->io_bytes[i]);
 		dst->runtime[i]		= le64_to_cpu(src->runtime[i]);
 	}
@@ -641,13 +909,23 @@ static void convert_ts(struct thread_stat *dst, struct thread_stat *src)
 	dst->total_err_count	= le64_to_cpu(src->total_err_count);
 	dst->first_error	= le32_to_cpu(src->first_error);
 	dst->kb_base		= le32_to_cpu(src->kb_base);
+	dst->unit_base		= le32_to_cpu(src->unit_base);
+
+	dst->latency_depth	= le32_to_cpu(src->latency_depth);
+	dst->latency_target	= le64_to_cpu(src->latency_target);
+	dst->latency_window	= le64_to_cpu(src->latency_window);
+	dst->latency_percentile.u.f = fio_uint64_to_double(le64_to_cpu(src->latency_percentile.u.i));
+
+	dst->nr_block_infos	= le64_to_cpu(src->nr_block_infos);
+	for (i = 0; i < dst->nr_block_infos; i++)
+		dst->block_infos[i] = le32_to_cpu(src->block_infos[i]);
 }
 
 static void convert_gs(struct group_run_stats *dst, struct group_run_stats *src)
 {
 	int i;
 
-	for (i = 0; i < 2; i++) {
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 		dst->max_run[i]		= le64_to_cpu(src->max_run[i]);
 		dst->min_run[i]		= le64_to_cpu(src->min_run[i]);
 		dst->max_bw[i]		= le64_to_cpu(src->max_bw[i]);
@@ -657,27 +935,54 @@ static void convert_gs(struct group_run_stats *dst, struct group_run_stats *src)
 	}
 
 	dst->kb_base	= le32_to_cpu(src->kb_base);
+	dst->unit_base	= le32_to_cpu(src->unit_base);
 	dst->groupid	= le32_to_cpu(src->groupid);
+	dst->unified_rw_rep	= le32_to_cpu(src->unified_rw_rep);
+}
+
+static void json_object_add_client_info(struct json_object *obj,
+					struct fio_client *client)
+{
+	const char *hostname = client->hostname ? client->hostname : "";
+
+	json_object_add_value_string(obj, "hostname", hostname);
+	json_object_add_value_int(obj, "port", client->port);
 }
 
 static void handle_ts(struct fio_client *client, struct fio_net_cmd *cmd)
 {
 	struct cmd_ts_pdu *p = (struct cmd_ts_pdu *) cmd->payload;
+	struct flist_head *opt_list = NULL;
+	struct json_object *tsobj;
 
-	show_thread_status(&p->ts, &p->rs);
+	if (client->opt_lists && p->ts.thread_number <= client->jobs)
+		opt_list = &client->opt_lists[p->ts.thread_number - 1];
 
-	if (sum_stat_clients == 1)
+	tsobj = show_thread_status(&p->ts, &p->rs, opt_list, NULL);
+	client->did_stat = 1;
+	if (tsobj) {
+		json_object_add_client_info(tsobj, client);
+		json_array_add_value_object(clients_array, tsobj);
+	}
+
+	if (sum_stat_clients <= 1)
 		return;
 
-	sum_thread_stats(&client_ts, &p->ts, sum_stat_nr);
+	sum_thread_stats(&client_ts, &p->ts, sum_stat_nr == 1);
 	sum_group_stats(&client_gs, &p->rs);
 
 	client_ts.members++;
+	client_ts.thread_number = p->ts.thread_number;
 	client_ts.groupid = p->ts.groupid;
+	client_ts.unified_rw_rep = p->ts.unified_rw_rep;
 
 	if (++sum_stat_nr == sum_stat_clients) {
 		strcpy(client_ts.name, "All clients");
-		show_thread_status(&client_ts, &client_gs);
+		tsobj = show_thread_status(&client_ts, &client_gs, NULL, NULL);
+		if (tsobj) {
+			json_object_add_client_info(tsobj, client);
+			json_array_add_value_object(clients_array, tsobj);
+		}
 	}
 }
 
@@ -685,7 +990,41 @@ static void handle_gs(struct fio_client *client, struct fio_net_cmd *cmd)
 {
 	struct group_run_stats *gs = (struct group_run_stats *) cmd->payload;
 
-	show_group_stats(gs);
+	if (output_format & FIO_OUTPUT_NORMAL)
+		show_group_stats(gs, NULL);
+}
+
+static void handle_job_opt(struct fio_client *client, struct fio_net_cmd *cmd)
+{
+	struct cmd_job_option *pdu = (struct cmd_job_option *) cmd->payload;
+	struct print_option *p;
+
+	if (!job_opt_object)
+		return;
+
+	pdu->global = le16_to_cpu(pdu->global);
+	pdu->truncated = le16_to_cpu(pdu->truncated);
+	pdu->groupid = le32_to_cpu(pdu->groupid);
+
+	p = malloc(sizeof(*p));
+	p->name = strdup((char *) pdu->name);
+	if (pdu->value[0] != '\0')
+		p->value = strdup((char *) pdu->value);
+	else
+		p->value = NULL;
+
+	if (pdu->global) {
+		const char *pos = "";
+
+		if (p->value)
+			pos = p->value;
+
+		json_object_add_value_string(job_opt_object, p->name, pos);
+	} else if (client->opt_lists) {
+		struct flist_head *opt_list = &client->opt_lists[pdu->groupid];
+
+		flist_add_tail(&p->list, opt_list);
+	}
 }
 
 static void handle_text(struct fio_client *client, struct fio_net_cmd *cmd)
@@ -709,16 +1048,16 @@ static void convert_agg(struct disk_util_agg *agg)
 	int i;
 
 	for (i = 0; i < 2; i++) {
-		agg->ios[i]	= le32_to_cpu(agg->ios[i]);
-		agg->merges[i]	= le32_to_cpu(agg->merges[i]);
+		agg->ios[i]	= le64_to_cpu(agg->ios[i]);
+		agg->merges[i]	= le64_to_cpu(agg->merges[i]);
 		agg->sectors[i]	= le64_to_cpu(agg->sectors[i]);
-		agg->ticks[i]	= le32_to_cpu(agg->ticks[i]);
+		agg->ticks[i]	= le64_to_cpu(agg->ticks[i]);
 	}
 
-	agg->io_ticks		= le32_to_cpu(agg->io_ticks);
-	agg->time_in_queue	= le32_to_cpu(agg->time_in_queue);
+	agg->io_ticks		= le64_to_cpu(agg->io_ticks);
+	agg->time_in_queue	= le64_to_cpu(agg->time_in_queue);
 	agg->slavecount		= le32_to_cpu(agg->slavecount);
-	agg->max_util.u.f	= fio_uint64_to_double(__le64_to_cpu(agg->max_util.u.i));
+	agg->max_util.u.f	= fio_uint64_to_double(le64_to_cpu(agg->max_util.u.i));
 }
 
 static void convert_dus(struct disk_util_stat *dus)
@@ -726,15 +1065,15 @@ static void convert_dus(struct disk_util_stat *dus)
 	int i;
 
 	for (i = 0; i < 2; i++) {
-		dus->ios[i]	= le32_to_cpu(dus->ios[i]);
-		dus->merges[i]	= le32_to_cpu(dus->merges[i]);
-		dus->sectors[i]	= le64_to_cpu(dus->sectors[i]);
-		dus->ticks[i]	= le32_to_cpu(dus->ticks[i]);
+		dus->s.ios[i]		= le64_to_cpu(dus->s.ios[i]);
+		dus->s.merges[i]	= le64_to_cpu(dus->s.merges[i]);
+		dus->s.sectors[i]	= le64_to_cpu(dus->s.sectors[i]);
+		dus->s.ticks[i]		= le64_to_cpu(dus->s.ticks[i]);
 	}
 
-	dus->io_ticks		= le32_to_cpu(dus->io_ticks);
-	dus->time_in_queue	= le32_to_cpu(dus->time_in_queue);
-	dus->msec		= le64_to_cpu(dus->msec);
+	dus->s.io_ticks		= le64_to_cpu(dus->s.io_ticks);
+	dus->s.time_in_queue	= le64_to_cpu(dus->s.time_in_queue);
+	dus->s.msec		= le64_to_cpu(dus->s.msec);
 }
 
 static void handle_du(struct fio_client *client, struct fio_net_cmd *cmd)
@@ -746,7 +1085,16 @@ static void handle_du(struct fio_client *client, struct fio_net_cmd *cmd)
 		log_info("\nDisk stats (read/write):\n");
 	}
 
-	print_disk_util(&du->dus, &du->agg, terse_output);
+	if (output_format & FIO_OUTPUT_JSON) {
+		struct json_object *duobj;
+		json_array_add_disk_util(&du->dus, &du->agg, du_array);
+		duobj = json_array_last_value_object(du_array);
+		json_object_add_client_info(duobj, client);
+	}
+	if (output_format & FIO_OUTPUT_TERSE)
+		print_disk_util(&du->dus, &du->agg, 1, NULL);
+	if (output_format & FIO_OUTPUT_NORMAL)
+		print_disk_util(&du->dus, &du->agg, 0, NULL);
 }
 
 static void convert_jobs_eta(struct jobs_eta *je)
@@ -756,13 +1104,14 @@ static void convert_jobs_eta(struct jobs_eta *je)
 	je->nr_running		= le32_to_cpu(je->nr_running);
 	je->nr_ramp		= le32_to_cpu(je->nr_ramp);
 	je->nr_pending		= le32_to_cpu(je->nr_pending);
+	je->nr_setting_up	= le32_to_cpu(je->nr_setting_up);
 	je->files_open		= le32_to_cpu(je->files_open);
 
-	for (i = 0; i < 2; i++) {
-		je->m_rate[i]		= le32_to_cpu(je->m_rate[i]);
-		je->t_rate[i]		= le32_to_cpu(je->t_rate[i]);
-		je->m_iops[i]		= le32_to_cpu(je->m_iops[i]);
-		je->t_iops[i]		= le32_to_cpu(je->t_iops[i]);
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
+		je->m_rate[i]	= le32_to_cpu(je->m_rate[i]);
+		je->t_rate[i]	= le32_to_cpu(je->t_rate[i]);
+		je->m_iops[i]	= le32_to_cpu(je->m_iops[i]);
+		je->t_iops[i]	= le32_to_cpu(je->t_iops[i]);
 		je->rate[i]	= le32_to_cpu(je->rate[i]);
 		je->iops[i]	= le32_to_cpu(je->iops[i]);
 	}
@@ -770,6 +1119,8 @@ static void convert_jobs_eta(struct jobs_eta *je)
 	je->elapsed_sec		= le64_to_cpu(je->elapsed_sec);
 	je->eta_sec		= le64_to_cpu(je->eta_sec);
 	je->nr_threads		= le32_to_cpu(je->nr_threads);
+	je->is_pow2		= le32_to_cpu(je->is_pow2);
+	je->unit_base		= le32_to_cpu(je->unit_base);
 }
 
 void fio_client_sum_jobs_eta(struct jobs_eta *dst, struct jobs_eta *je)
@@ -779,9 +1130,10 @@ void fio_client_sum_jobs_eta(struct jobs_eta *dst, struct jobs_eta *je)
 	dst->nr_running		+= je->nr_running;
 	dst->nr_ramp		+= je->nr_ramp;
 	dst->nr_pending		+= je->nr_pending;
+	dst->nr_setting_up	+= je->nr_setting_up;
 	dst->files_open		+= je->files_open;
 
-	for (i = 0; i < 2; i++) {
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 		dst->m_rate[i]	+= je->m_rate[i];
 		dst->t_rate[i]	+= je->t_rate[i];
 		dst->m_iops[i]	+= je->m_iops[i];
@@ -796,39 +1148,60 @@ void fio_client_sum_jobs_eta(struct jobs_eta *dst, struct jobs_eta *je)
 		dst->eta_sec = je->eta_sec;
 
 	dst->nr_threads		+= je->nr_threads;
-	/* we need to handle je->run_str too ... */
-}
 
-void fio_client_dec_jobs_eta(struct client_eta *eta, client_eta_op eta_fn)
-{
-	if (!--eta->pending) {
-		eta_fn(&eta->eta);
-		free(eta);
-	}
+	/*
+	 * This wont be correct for multiple strings, but at least it
+	 * works for the basic cases.
+	 */
+	strcpy((char *) dst->run_str, (char *) je->run_str);
 }
 
 static void remove_reply_cmd(struct fio_client *client, struct fio_net_cmd *cmd)
 {
-	struct fio_net_int_cmd *icmd = NULL;
+	struct fio_net_cmd_reply *reply = NULL;
 	struct flist_head *entry;
 
 	flist_for_each(entry, &client->cmd_list) {
-		icmd = flist_entry(entry, struct fio_net_int_cmd, list);
+		reply = flist_entry(entry, struct fio_net_cmd_reply, list);
 
-		if (cmd->tag == (uintptr_t) icmd)
+		if (cmd->tag == (uintptr_t) reply)
 			break;
 
-		icmd = NULL;
+		reply = NULL;
 	}
 
-	if (!icmd) {
-		log_err("fio: client: unable to find matching tag\n");
+	if (!reply) {
+		log_err("fio: client: unable to find matching tag (%llx)\n", (unsigned long long) cmd->tag);
 		return;
 	}
 
-	flist_del(&icmd->list);
-	cmd->tag = icmd->saved_tag;
-	free(icmd);
+	flist_del(&reply->list);
+	cmd->tag = reply->saved_tag;
+	free(reply);
+}
+
+int fio_client_wait_for_reply(struct fio_client *client, uint64_t tag)
+{
+	do {
+		struct fio_net_cmd_reply *reply = NULL;
+		struct flist_head *entry;
+
+		flist_for_each(entry, &client->cmd_list) {
+			reply = flist_entry(entry, struct fio_net_cmd_reply, list);
+
+			if (tag == (uintptr_t) reply)
+				break;
+
+			reply = NULL;
+		}
+
+		if (!reply)
+			break;
+
+		usleep(1000);
+	} while (1);
+
+	return 0;
 }
 
 static void handle_eta(struct fio_client *client, struct fio_net_cmd *cmd)
@@ -842,6 +1215,7 @@ static void handle_eta(struct fio_client *client, struct fio_net_cmd *cmd)
 
 	client->eta_in_flight = NULL;
 	flist_del_init(&client->eta_list);
+	client->eta_timeouts = 0;
 
 	if (client->ops->jobs_eta)
 		client->ops->jobs_eta(client, je);
@@ -852,7 +1226,7 @@ static void handle_eta(struct fio_client *client, struct fio_net_cmd *cmd)
 
 static void handle_probe(struct fio_client *client, struct fio_net_cmd *cmd)
 {
-	struct cmd_probe_pdu *probe = (struct cmd_probe_pdu *) cmd->payload;
+	struct cmd_probe_reply_pdu *probe = (struct cmd_probe_reply_pdu *) cmd->payload;
 	const char *os, *arch;
 	char bit[16];
 
@@ -865,10 +1239,11 @@ static void handle_probe(struct fio_client *client, struct fio_net_cmd *cmd)
 		os = "unknown";
 
 	sprintf(bit, "%d-bit", probe->bpp * 8);
+	probe->flags = le64_to_cpu(probe->flags);
 
-	log_info("hostname=%s, be=%u, %s, os=%s, arch=%s, fio=%u.%u.%u\n",
+	log_info("hostname=%s, be=%u, %s, os=%s, arch=%s, fio=%s, flags=%lx\n",
 		probe->hostname, probe->bigendian, bit, os, arch,
-		probe->fio_major, probe->fio_minor, probe->fio_patch);
+		probe->fio_version, (unsigned long) probe->flags);
 
 	if (!client->name)
 		client->name = strdup((char *) probe->hostname);
@@ -880,6 +1255,20 @@ static void handle_start(struct fio_client *client, struct fio_net_cmd *cmd)
 
 	client->state = Client_started;
 	client->jobs = le32_to_cpu(pdu->jobs);
+	client->nr_stat = le32_to_cpu(pdu->stat_outputs);
+
+	if (client->jobs) {
+		int i;
+
+		if (client->opt_lists)
+			free(client->opt_lists);
+
+		client->opt_lists = malloc(client->jobs * sizeof(struct flist_head));
+		for (i = 0; i < client->jobs; i++)
+			INIT_FLIST_HEAD(&client->opt_lists[i]);
+	}
+
+	sum_stat_clients += client->nr_stat;
 }
 
 static void handle_stop(struct fio_client *client, struct fio_net_cmd *cmd)
@@ -905,6 +1294,167 @@ static void convert_text(struct fio_net_cmd *cmd)
 	pdu->log_usec	= le64_to_cpu(pdu->log_usec);
 }
 
+static struct cmd_iolog_pdu *convert_iolog_gz(struct fio_net_cmd *cmd,
+					      struct cmd_iolog_pdu *pdu)
+{
+#ifdef CONFIG_ZLIB
+	struct cmd_iolog_pdu *ret;
+	z_stream stream;
+	uint32_t nr_samples;
+	size_t total;
+	void *p;
+
+	stream.zalloc = Z_NULL;
+	stream.zfree = Z_NULL;
+	stream.opaque = Z_NULL;
+	stream.avail_in = 0;
+	stream.next_in = Z_NULL;
+
+	if (inflateInit(&stream) != Z_OK)
+		return NULL;
+
+	/*
+	 * Get header first, it's not compressed
+	 */
+	nr_samples = le64_to_cpu(pdu->nr_samples);
+
+	total = nr_samples * __log_entry_sz(le32_to_cpu(pdu->log_offset));
+	ret = malloc(total + sizeof(*pdu));
+	ret->nr_samples = nr_samples;
+
+	memcpy(ret, pdu, sizeof(*pdu));
+
+	p = (void *) ret + sizeof(*pdu);
+
+	stream.avail_in = cmd->pdu_len - sizeof(*pdu);
+	stream.next_in = (void *) pdu + sizeof(*pdu);
+	while (stream.avail_in) {
+		unsigned int this_chunk = 65536;
+		unsigned int this_len;
+		int err;
+
+		if (this_chunk > total)
+			this_chunk = total;
+
+		stream.avail_out = this_chunk;
+		stream.next_out = p;
+		err = inflate(&stream, Z_NO_FLUSH);
+		/* may be Z_OK, or Z_STREAM_END */
+		if (err < 0) {
+			log_err("fio: inflate error %d\n", err);
+			free(ret);
+			ret = NULL;
+			goto err;
+		}
+
+		this_len = this_chunk - stream.avail_out;
+		p += this_len;
+		total -= this_len;
+	}
+
+err:
+	inflateEnd(&stream);
+	return ret;
+#else
+	return NULL;
+#endif
+}
+
+/*
+ * This has been compressed on the server side, since it can be big.
+ * Uncompress here.
+ */
+static struct cmd_iolog_pdu *convert_iolog(struct fio_net_cmd *cmd)
+{
+	struct cmd_iolog_pdu *pdu = (struct cmd_iolog_pdu *) cmd->payload;
+	struct cmd_iolog_pdu *ret;
+	uint64_t i;
+	void *samples;
+
+	/*
+	 * Convert if compressed and we support it. If it's not
+	 * compressed, we need not do anything.
+	 */
+	if (le32_to_cpu(pdu->compressed)) {
+#ifndef CONFIG_ZLIB
+		log_err("fio: server sent compressed data by mistake\n");
+		return NULL;
+#endif
+		ret = convert_iolog_gz(cmd, pdu);
+		if (!ret) {
+			log_err("fio: failed decompressing log\n");
+			return NULL;
+		}
+	} else
+		ret = pdu;
+
+	ret->nr_samples		= le64_to_cpu(ret->nr_samples);
+	ret->thread_number	= le32_to_cpu(ret->thread_number);
+	ret->log_type		= le32_to_cpu(ret->log_type);
+	ret->compressed		= le32_to_cpu(ret->compressed);
+	ret->log_offset		= le32_to_cpu(ret->log_offset);
+
+	samples = &ret->samples[0];
+	for (i = 0; i < ret->nr_samples; i++) {
+		struct io_sample *s;
+
+		s = __get_sample(samples, ret->log_offset, i);
+		s->time		= le64_to_cpu(s->time);
+		s->val		= le64_to_cpu(s->val);
+		s->__ddir	= le32_to_cpu(s->__ddir);
+		s->bs		= le32_to_cpu(s->bs);
+
+		if (ret->log_offset) {
+			struct io_sample_offset *so = (void *) s;
+
+			so->offset = le64_to_cpu(so->offset);
+		}
+	}
+
+	return ret;
+}
+
+static void sendfile_reply(int fd, struct cmd_sendfile_reply *rep,
+			   size_t size, uint64_t tag)
+{
+	rep->error = cpu_to_le32(rep->error);
+	fio_net_send_cmd(fd, FIO_NET_CMD_SENDFILE, rep, size, &tag, NULL);
+}
+
+static int send_file(struct fio_client *client, struct cmd_sendfile *pdu,
+		     uint64_t tag)
+{
+	struct cmd_sendfile_reply *rep;
+	struct stat sb;
+	size_t size;
+	int fd;
+
+	size = sizeof(*rep);
+	rep = malloc(size);
+
+	if (stat((char *)pdu->path, &sb) < 0) {
+fail:
+		rep->error = errno;
+		sendfile_reply(client->fd, rep, size, tag);
+		free(rep);
+		return 1;
+	}
+
+	size += sb.st_size;
+	rep = realloc(rep, size);
+	rep->size = cpu_to_le32((uint32_t) sb.st_size);
+
+	fd = open((char *)pdu->path, O_RDONLY);
+	if (fd == -1 )
+		goto fail;
+
+	rep->error = read_data(fd, &rep->data, sb.st_size);
+	sendfile_reply(client->fd, rep, size, tag);
+	free(rep);
+	close(fd);
+	return 0;
+}
+
 int fio_handle_client(struct fio_client *client)
 {
 	struct client_ops *ops = client->ops;
@@ -922,14 +1472,12 @@ int fio_handle_client(struct fio_client *client)
 	switch (cmd->opcode) {
 	case FIO_NET_CMD_QUIT:
 		if (ops->quit)
-			ops->quit(client);
+			ops->quit(client, cmd);
 		remove_client(client);
-		free(cmd);
 		break;
 	case FIO_NET_CMD_TEXT:
 		convert_text(cmd);
-		ops->text_op(client, cmd);
-		free(cmd);
+		ops->text(client, cmd);
 		break;
 	case FIO_NET_CMD_DU: {
 		struct cmd_du_pdu *du = (struct cmd_du_pdu *) cmd->payload;
@@ -938,7 +1486,6 @@ int fio_handle_client(struct fio_client *client)
 		convert_agg(&du->agg);
 
 		ops->disk_util(client, cmd);
-		free(cmd);
 		break;
 		}
 	case FIO_NET_CMD_TS: {
@@ -948,7 +1495,6 @@ int fio_handle_client(struct fio_client *client)
 		convert_gs(&p->rs, &p->rs);
 
 		ops->thread_status(client, cmd);
-		free(cmd);
 		break;
 		}
 	case FIO_NET_CMD_GS: {
@@ -957,7 +1503,6 @@ int fio_handle_client(struct fio_client *client)
 		convert_gs(gs, gs);
 
 		ops->group_stats(client, cmd);
-		free(cmd);
 		break;
 		}
 	case FIO_NET_CMD_ETA: {
@@ -966,46 +1511,118 @@ int fio_handle_client(struct fio_client *client)
 		remove_reply_cmd(client, cmd);
 		convert_jobs_eta(je);
 		handle_eta(client, cmd);
-		free(cmd);
 		break;
 		}
 	case FIO_NET_CMD_PROBE:
 		remove_reply_cmd(client, cmd);
 		ops->probe(client, cmd);
-		free(cmd);
 		break;
 	case FIO_NET_CMD_SERVER_START:
 		client->state = Client_running;
-		free(cmd);
+		if (ops->job_start)
+			ops->job_start(client, cmd);
 		break;
-	case FIO_NET_CMD_START:
-		handle_start(client, cmd);
-		free(cmd);
+	case FIO_NET_CMD_START: {
+		struct cmd_start_pdu *pdu = (struct cmd_start_pdu *) cmd->payload;
+
+		pdu->jobs = le32_to_cpu(pdu->jobs);
+		ops->start(client, cmd);
 		break;
+		}
 	case FIO_NET_CMD_STOP: {
 		struct cmd_end_pdu *pdu = (struct cmd_end_pdu *) cmd->payload;
 
 		convert_stop(cmd);
 		client->state = Client_stopped;
-		client->error = pdu->error;
+		client->error = le32_to_cpu(pdu->error);
+		client->signal = le32_to_cpu(pdu->signal);
 		ops->stop(client, cmd);
-		free(cmd);
 		break;
 		}
-	case FIO_NET_CMD_ADD_JOB:
+	case FIO_NET_CMD_ADD_JOB: {
+		struct cmd_add_job_pdu *pdu = (struct cmd_add_job_pdu *) cmd->payload;
+
+		client->thread_number = le32_to_cpu(pdu->thread_number);
+		client->groupid = le32_to_cpu(pdu->groupid);
+
 		if (ops->add_job)
 			ops->add_job(client, cmd);
-		free(cmd);
 		break;
+		}
+	case FIO_NET_CMD_IOLOG:
+		if (ops->iolog) {
+			struct cmd_iolog_pdu *pdu;
+
+			pdu = convert_iolog(cmd);
+			ops->iolog(client, pdu);
+		}
+		break;
+	case FIO_NET_CMD_UPDATE_JOB:
+		ops->update_job(client, cmd);
+		remove_reply_cmd(client, cmd);
+		break;
+	case FIO_NET_CMD_VTRIGGER: {
+		struct all_io_list *pdu = (struct all_io_list *) cmd->payload;
+		char buf[128];
+		int off = 0;
+
+		if (aux_path) {
+			strcpy(buf, aux_path);
+			off = strlen(buf);
+		}
+
+		__verify_save_state(pdu, server_name(client, &buf[off], sizeof(buf) - off));
+		exec_trigger(trigger_cmd);
+		break;
+		}
+	case FIO_NET_CMD_SENDFILE: {
+		struct cmd_sendfile *pdu = (struct cmd_sendfile *) cmd->payload;
+		send_file(client, pdu, cmd->tag);
+		break;
+		}
+	case FIO_NET_CMD_JOB_OPT: {
+		handle_job_opt(client, cmd);
+		break;
+	}
 	default:
 		log_err("fio: unknown client op: %s\n", fio_server_op(cmd->opcode));
-		free(cmd);
 		break;
 	}
 
+	free(cmd);
 	return 1;
 }
 
+int fio_clients_send_trigger(const char *cmd)
+{
+	struct flist_head *entry;
+	struct fio_client *client;
+	size_t slen;
+
+	dprint(FD_NET, "client: send vtrigger: %s\n", cmd);
+
+	if (!cmd)
+		slen = 0;
+	else
+		slen = strlen(cmd);
+
+	flist_for_each(entry, &client_list) {
+		struct cmd_vtrigger_pdu *pdu;
+
+		client = flist_entry(entry, struct fio_client, list);
+
+		pdu = malloc(sizeof(*pdu) + slen);
+		pdu->len = cpu_to_le16((uint16_t) slen);
+		if (slen)
+			memcpy(pdu->cmd, cmd, slen);
+		fio_net_send_cmd(client->fd, FIO_NET_CMD_VTRIGGER, pdu,
+					sizeof(*pdu) + slen, NULL, NULL);
+		free(pdu);
+	}
+
+	return 0;
+}
+
 static void request_client_etas(struct client_ops *ops)
 {
 	struct fio_client *client;
@@ -1015,8 +1632,7 @@ static void request_client_etas(struct client_ops *ops)
 
 	dprint(FD_NET, "client: request eta (%d)\n", nr_clients);
 
-	eta = malloc(sizeof(*eta));
-	memset(&eta->eta, 0, sizeof(eta->eta));
+	eta = calloc(1, sizeof(*eta) + __THREAD_RUNSTR_SZ(REAL_MAX_JOBS));
 	eta->pending = nr_clients;
 
 	flist_for_each(entry, &client_list) {
@@ -1036,29 +1652,62 @@ static void request_client_etas(struct client_ops *ops)
 					(uintptr_t) eta, &client->cmd_list);
 	}
 
-	while (skipped--)
-		fio_client_dec_jobs_eta(eta, ops->eta);
+	while (skipped--) {
+		if (!fio_client_dec_jobs_eta(eta, ops->eta))
+			break;
+	}
 
 	dprint(FD_NET, "client: requested eta tag %p\n", eta);
 }
 
+/*
+ * A single SEND_ETA timeout isn't fatal. Attempt to recover.
+ */
+static int handle_cmd_timeout(struct fio_client *client,
+			      struct fio_net_cmd_reply *reply)
+{
+	if (reply->opcode != FIO_NET_CMD_SEND_ETA)
+		return 1;
+
+	log_info("client <%s>: timeout on SEND_ETA\n", client->hostname);
+	flist_del(&reply->list);
+	free(reply);
+
+	flist_del_init(&client->eta_list);
+	if (client->eta_in_flight) {
+		fio_client_dec_jobs_eta(client->eta_in_flight, client->ops->eta);
+		client->eta_in_flight = NULL;
+	}
+
+	/*
+	 * If we fail 5 in a row, give up...
+	 */
+	if (client->eta_timeouts++ > 5)
+		return 1;
+
+	return 0;
+}
+
 static int client_check_cmd_timeout(struct fio_client *client,
 				    struct timeval *now)
 {
-	struct fio_net_int_cmd *cmd;
+	struct fio_net_cmd_reply *reply;
 	struct flist_head *entry, *tmp;
 	int ret = 0;
 
 	flist_for_each_safe(entry, tmp, &client->cmd_list) {
-		cmd = flist_entry(entry, struct fio_net_int_cmd, list);
+		reply = flist_entry(entry, struct fio_net_cmd_reply, list);
 
-		if (mtime_since(&cmd->tv, now) < FIO_NET_CLIENT_TIMEOUT)
+		if (mtime_since(&reply->tv, now) < FIO_NET_CLIENT_TIMEOUT)
+			continue;
+
+		if (!handle_cmd_timeout(client, reply))
 			continue;
 
 		log_err("fio: client %s, timeout on cmd %s\n", client->hostname,
-						fio_server_op(cmd->cmd.opcode));
-		flist_del(&cmd->list);
-		free(cmd);
+						fio_server_op(reply->opcode));
+		flist_del(&reply->list);
+		free(reply);
 		ret = 1;
 	}
 
@@ -1072,7 +1721,7 @@ static int fio_check_clients_timed_out(void)
 	struct timeval tv;
 	int ret = 0;
 
-	gettimeofday(&tv, NULL);
+	fio_gettime(&tv, NULL);
 
 	flist_for_each_safe(entry, tmp, &client_list) {
 		client = flist_entry(entry, struct fio_client, list);
@@ -1088,6 +1737,7 @@ static int fio_check_clients_timed_out(void)
 		else
 			log_err("fio: client %s timed out\n", client->hostname);
 
+		client->error = ETIMEDOUT;
 		remove_client(client);
 		ret = 1;
 	}
@@ -1100,11 +1750,10 @@ int fio_handle_clients(struct client_ops *ops)
 	struct pollfd *pfds;
 	int i, ret = 0, retval = 0;
 
-	gettimeofday(&eta_tv, NULL);
+	fio_gettime(&eta_tv, NULL);
 
 	pfds = malloc(nr_clients * sizeof(struct pollfd));
 
-	sum_stat_clients = nr_clients;
 	init_thread_stat(&client_ts);
 	init_group_run_stat(&client_gs);
 
@@ -1134,8 +1783,9 @@ int fio_handle_clients(struct client_ops *ops)
 
 		do {
 			struct timeval tv;
+			int timeout;
 
-			gettimeofday(&tv, NULL);
+			fio_gettime(&tv, NULL);
 			if (mtime_since(&eta_tv, &tv) >= 900) {
 				request_client_etas(ops);
 				memcpy(&eta_tv, &tv, sizeof(tv));
@@ -1144,7 +1794,11 @@ int fio_handle_clients(struct client_ops *ops)
 					break;
 			}
 
-			ret = poll(pfds, nr_clients, 100);
+			check_trigger_file();
+
+			timeout = min(100u, ops->eta_msec);
+
+			ret = poll(pfds, nr_clients, timeout);
 			if (ret < 0) {
 				if (errno == EINTR)
 					continue;
@@ -1160,7 +1814,7 @@ int fio_handle_clients(struct client_ops *ops)
 
 			client = find_client_by_fd(pfds[i].fd);
 			if (!client) {
-				log_err("fio: unknown client fd %d\n", pfds[i].fd);
+				log_err("fio: unknown client fd %ld\n", (long) pfds[i].fd);
 				continue;
 			}
 			if (!fio_handle_client(client)) {
@@ -1170,10 +1824,12 @@ int fio_handle_clients(struct client_ops *ops)
 				retval = 1;
 			} else if (client->error)
 				retval = 1;
-			put_client(client);
+			fio_put_client(client);
 		}
 	}
 
+	fio_client_json_fini();
+
 	free(pfds);
-	return retval;
+	return retval || error_clients;
 }