X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=client.c;h=366db71615306a98b6303b2f468682444b71f815;hp=928f7d38be47aa19ba5cdebc207dc268ca986695;hb=0279b88017b5d21f0fcbb1b57481339735c41797;hpb=a89d5319371a3907df7eb4fa13cef899b09ec374

diff --git a/client.c b/client.c
index 928f7d38..366db716 100644
--- a/client.c
+++ b/client.c
@@ -23,6 +23,7 @@
 #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);
@@ -58,9 +59,12 @@ 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)
 #define FIO_CLIENT_HASH_MASK	(FIO_CLIENT_HASH_SZ - 1)
@@ -88,12 +92,49 @@ 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)
 {
-	if (output_format != FIO_OUTPUT_JSON)
+	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();
@@ -102,9 +143,11 @@ static void fio_client_json_init(void)
 
 static void fio_client_json_fini(void)
 {
-	if (output_format != FIO_OUTPUT_JSON)
+	if (!(output_format & FIO_OUTPUT_JSON))
 		return;
-	json_print_object(root);
+
+	log_info("\n");
+	json_print_object(root, NULL);
 	log_info("\n");
 	json_free_object(root);
 	root = NULL;
@@ -147,13 +190,29 @@ void fio_put_client(struct fio_client *client)
 	}
 	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);
@@ -271,6 +330,9 @@ int fio_client_add_ini_file(void *cookie, const char *ini_file, int remote)
 	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);
@@ -334,10 +396,27 @@ 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");
 
@@ -347,6 +426,10 @@ static void probe_client(struct fio_client *client)
 	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);
 }
 
@@ -469,11 +552,6 @@ static void sig_int(int sig)
 	fio_clients_terminate();
 }
 
-static void sig_show_status(int sig)
-{
-	show_running_run_stats();
-}
-
 static void client_signal_handler(void)
 {
 	struct sigaction act;
@@ -651,15 +729,13 @@ static int __fio_client_send_local_ini(struct fio_client *client,
 
 	fd = open(filename, O_RDONLY);
 	if (fd < 0) {
-		int ret = -errno;
-
+		ret = -errno;
 		log_err("fio: job file <%s> open: %s\n", filename, strerror(errno));
 		return ret;
 	}
 
 	if (fstat(fd, &sb) < 0) {
-		int ret = -errno;
-
+		ret = -errno;
 		log_err("fio: job file stat: %s\n", strerror(errno));
 		close(fd);
 		return ret;
@@ -671,21 +747,7 @@ static int __fio_client_send_local_ini(struct fio_client *client,
 
 	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(pdu);
@@ -804,6 +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];
@@ -852,6 +915,10 @@ static void convert_ts(struct thread_stat *dst, struct thread_stat *src)
 	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)
@@ -885,9 +952,13 @@ static void json_object_add_client_info(struct json_object *obj,
 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;
 
-	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];
+
+	tsobj = show_thread_status(&p->ts, &p->rs, opt_list, NULL);
 	client->did_stat = 1;
 	if (tsobj) {
 		json_object_add_client_info(tsobj, client);
@@ -897,7 +968,7 @@ static void handle_ts(struct fio_client *client, struct fio_net_cmd *cmd)
 	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++;
@@ -907,7 +978,7 @@ static void handle_ts(struct fio_client *client, struct fio_net_cmd *cmd)
 
 	if (++sum_stat_nr == sum_stat_clients) {
 		strcpy(client_ts.name, "All clients");
-		tsobj = 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);
@@ -919,7 +990,37 @@ 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;
+
+	pdu->global = le16_to_cpu(pdu->global);
+	pdu->groupid = le16_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)
@@ -943,16 +1044,16 @@ 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)
@@ -960,14 +1061,14 @@ static void convert_dus(struct disk_util_stat *dus)
 	int i;
 
 	for (i = 0; i < 2; i++) {
-		dus->s.ios[i]		= le32_to_cpu(dus->s.ios[i]);
-		dus->s.merges[i]	= le32_to_cpu(dus->s.merges[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]		= le32_to_cpu(dus->s.ticks[i]);
+		dus->s.ticks[i]		= le64_to_cpu(dus->s.ticks[i]);
 	}
 
-	dus->s.io_ticks		= le32_to_cpu(dus->s.io_ticks);
-	dus->s.time_in_queue	= le32_to_cpu(dus->s.time_in_queue);
+	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);
 }
 
@@ -980,13 +1081,16 @@ static void handle_du(struct fio_client *client, struct fio_net_cmd *cmd)
 		log_info("\nDisk stats (read/write):\n");
 	}
 
-	if (output_format == FIO_OUTPUT_JSON) {
+	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);
-	} else
-		print_disk_util(&du->dus, &du->agg, output_format == FIO_OUTPUT_TERSE);
+	}
+	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)
@@ -1048,14 +1152,6 @@ void fio_client_sum_jobs_eta(struct jobs_eta *dst, struct jobs_eta *je)
 	strcpy((char *) dst->run_str, (char *) je->run_str);
 }
 
-void fio_client_dec_jobs_eta(struct client_eta *eta, client_eta_op eta_fn)
-{
-	if (!--eta->pending) {
-		eta_fn(&eta->eta);
-		free(eta);
-	}
-}
-
 static void remove_reply_cmd(struct fio_client *client, struct fio_net_cmd *cmd)
 {
 	struct fio_net_cmd_reply *reply = NULL;
@@ -1115,6 +1211,7 @@ 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);
@@ -1156,6 +1253,17 @@ static void handle_start(struct fio_client *client, struct fio_net_cmd *cmd)
 	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;
 }
 
@@ -1302,6 +1410,47 @@ static struct cmd_iolog_pdu *convert_iolog(struct fio_net_cmd *cmd)
 	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;
@@ -1321,12 +1470,10 @@ int fio_handle_client(struct fio_client *client)
 		if (ops->quit)
 			ops->quit(client, cmd);
 		remove_client(client);
-		free(cmd);
 		break;
 	case FIO_NET_CMD_TEXT:
 		convert_text(cmd);
 		ops->text(client, cmd);
-		free(cmd);
 		break;
 	case FIO_NET_CMD_DU: {
 		struct cmd_du_pdu *du = (struct cmd_du_pdu *) cmd->payload;
@@ -1335,7 +1482,6 @@ 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: {
@@ -1345,7 +1491,6 @@ 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: {
@@ -1354,7 +1499,6 @@ 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: {
@@ -1363,26 +1507,22 @@ 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;
 		if (ops->job_start)
 			ops->job_start(client, cmd);
-		free(cmd);
 		break;
 	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);
-		free(cmd);
 		break;
 		}
 	case FIO_NET_CMD_STOP: {
@@ -1393,7 +1533,6 @@ int fio_handle_client(struct fio_client *client)
 		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: {
@@ -1404,7 +1543,6 @@ int fio_handle_client(struct fio_client *client)
 
 		if (ops->add_job)
 			ops->add_job(client, cmd);
-		free(cmd);
 		break;
 		}
 	case FIO_NET_CMD_IOLOG:
@@ -1414,22 +1552,73 @@ int fio_handle_client(struct fio_client *client)
 			pdu = convert_iolog(cmd);
 			ops->iolog(client, pdu);
 		}
-		free(cmd);
 		break;
 	case FIO_NET_CMD_UPDATE_JOB:
 		ops->update_job(client, cmd);
 		remove_reply_cmd(client, cmd);
-		free(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;
@@ -1459,12 +1648,42 @@ 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)
 {
@@ -1478,6 +1697,9 @@ static int client_check_cmd_timeout(struct fio_client *client,
 		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(reply->opcode));
 		flist_del(&reply->list);
@@ -1511,6 +1733,7 @@ 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;
 	}
@@ -1556,6 +1779,7 @@ int fio_handle_clients(struct client_ops *ops)
 
 		do {
 			struct timeval tv;
+			int timeout;
 
 			fio_gettime(&tv, NULL);
 			if (mtime_since(&eta_tv, &tv) >= 900) {
@@ -1566,7 +1790,11 @@ int fio_handle_clients(struct client_ops *ops)
 					break;
 			}
 
-			ret = poll(pfds, nr_clients, ops->eta_msec);
+			check_trigger_file();
+
+			timeout = min(100u, ops->eta_msec);
+
+			ret = poll(pfds, nr_clients, timeout);
 			if (ret < 0) {
 				if (errno == EINTR)
 					continue;
@@ -1599,5 +1827,5 @@ int fio_handle_clients(struct client_ops *ops)
 	fio_client_json_fini();
 
 	free(pfds);
-	return retval;
+	return retval || error_clients;
 }