projects / fio.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
engines/sg: improve error handling
[fio.git] / client.c
... / ...
CommitLineData
1#include <stdio.h>
2#include <stdlib.h>
3#include <unistd.h>
4#include <errno.h>
5#include <fcntl.h>
6#include <poll.h>
7#include <sys/types.h>
8#include <sys/stat.h>
9#include <sys/socket.h>
10#include <sys/un.h>
11#include <netinet/in.h>
12#include <arpa/inet.h>
13#include <netdb.h>
14#include <signal.h>
15#ifdef CONFIG_ZLIB
16#include <zlib.h>
17#endif
18
19#include "fio.h"
20#include "client.h"
21#include "server.h"
22#include "flist.h"
23#include "hash.h"
24#include "verify-state.h"
25
26static void handle_du(struct fio_client *client, struct fio_net_cmd *cmd);
27static void handle_ts(struct fio_client *client, struct fio_net_cmd *cmd);
28static void handle_gs(struct fio_client *client, struct fio_net_cmd *cmd);
29static void handle_probe(struct fio_client *client, struct fio_net_cmd *cmd);
30static void handle_text(struct fio_client *client, struct fio_net_cmd *cmd);
31static void handle_stop(struct fio_client *client);
32static void handle_start(struct fio_client *client, struct fio_net_cmd *cmd);
33
34static void convert_text(struct fio_net_cmd *cmd);
35static void client_display_thread_status(struct jobs_eta *je);
36
37struct client_ops fio_client_ops = {
38 .text = handle_text,
39 .disk_util = handle_du,
40 .thread_status = handle_ts,
41 .group_stats = handle_gs,
42 .stop = handle_stop,
43 .start = handle_start,
44 .eta = client_display_thread_status,
45 .probe = handle_probe,
46 .eta_msec = FIO_CLIENT_DEF_ETA_MSEC,
47 .client_type = FIO_CLIENT_TYPE_CLI,
48};
49
50static struct timespec eta_ts;
51
52static FLIST_HEAD(client_list);
53static FLIST_HEAD(eta_list);
54
55static FLIST_HEAD(arg_list);
56
57struct thread_stat client_ts;
58struct group_run_stats client_gs;
59int sum_stat_clients;
60
61static int sum_stat_nr;
62static struct json_object *root = NULL;
63static struct json_object *job_opt_object = NULL;
64static struct json_array *clients_array = NULL;
65static struct json_array *du_array = NULL;
66
67static int error_clients;
68
69#define FIO_CLIENT_HASH_BITS 7
70#define FIO_CLIENT_HASH_SZ (1 << FIO_CLIENT_HASH_BITS)
71#define FIO_CLIENT_HASH_MASK (FIO_CLIENT_HASH_SZ - 1)
72static struct flist_head client_hash[FIO_CLIENT_HASH_SZ];
73
74static struct cmd_iolog_pdu *convert_iolog(struct fio_net_cmd *, bool *);
75
76static void fio_client_add_hash(struct fio_client *client)
77{
78 int bucket = hash_long(client->fd, FIO_CLIENT_HASH_BITS);
79
80 bucket &= FIO_CLIENT_HASH_MASK;
81 flist_add(&client->hash_list, &client_hash[bucket]);
82}
83
84static void fio_client_remove_hash(struct fio_client *client)
85{
86 if (!flist_empty(&client->hash_list))
87 flist_del_init(&client->hash_list);
88}
89
90static void fio_init fio_client_hash_init(void)
91{
92 int i;
93
94 for (i = 0; i < FIO_CLIENT_HASH_SZ; i++)
95 INIT_FLIST_HEAD(&client_hash[i]);
96}
97
98static int read_data(int fd, void *data, size_t size)
99{
100 ssize_t ret;
101
102 while (size) {
103 ret = read(fd, data, size);
104 if (ret < 0) {
105 if (errno == EAGAIN || errno == EINTR)
106 continue;
107 break;
108 } else if (!ret)
109 break;
110 else {
111 data += ret;
112 size -= ret;
113 }
114 }
115
116 if (size)
117 return EAGAIN;
118
119 return 0;
120}
121
122static int read_ini_data(int fd, void *data, size_t size)
123{
124 char *p = data;
125 int ret = 0;
126 FILE *fp;
127 int dupfd;
128
129 dupfd = dup(fd);
130 if (dupfd < 0)
131 return errno;
132
133 fp = fdopen(dupfd, "r");
134 if (!fp) {
135 ret = errno;
136 close(dupfd);
137 goto out;
138 }
139
140 while (1) {
141 ssize_t len;
142 char buf[OPT_LEN_MAX+1], *sub;
143
144 if (!fgets(buf, sizeof(buf), fp)) {
145 if (ferror(fp)) {
146 if (errno == EAGAIN || errno == EINTR)
147 continue;
148 ret = errno;
149 }
150 break;
151 }
152
153 sub = fio_option_dup_subs(buf);
154 len = strlen(sub);
155 if (len + 1 > size) {
156 log_err("fio: no space left to read data\n");
157 free(sub);
158 ret = ENOSPC;
159 break;
160 }
161
162 memcpy(p, sub, len);
163 free(sub);
164 p += len;
165 *p = '\0';
166 size -= len;
167 }
168
169 fclose(fp);
170out:
171 return ret;
172}
173
174static void fio_client_json_init(void)
175{
176 char time_buf[32];
177 time_t time_p;
178
179 if (!(output_format & FIO_OUTPUT_JSON))
180 return;
181
182 time(&time_p);
183 os_ctime_r((const time_t *) &time_p, time_buf, sizeof(time_buf));
184 time_buf[strlen(time_buf) - 1] = '\0';
185
186 root = json_create_object();
187 json_object_add_value_string(root, "fio version", fio_version_string);
188 json_object_add_value_int(root, "timestamp", time_p);
189 json_object_add_value_string(root, "time", time_buf);
190
191 job_opt_object = json_create_object();
192 json_object_add_value_object(root, "global options", job_opt_object);
193 clients_array = json_create_array();
194 json_object_add_value_array(root, "client_stats", clients_array);
195 du_array = json_create_array();
196 json_object_add_value_array(root, "disk_util", du_array);
197}
198
199static void fio_client_json_fini(void)
200{
201 if (!(output_format & FIO_OUTPUT_JSON))
202 return;
203
204 log_info("\n");
205 json_print_object(root, NULL);
206 log_info("\n");
207 json_free_object(root);
208 root = NULL;
209 clients_array = NULL;
210 du_array = NULL;
211}
212
213static struct fio_client *find_client_by_fd(int fd)
214{
215 int bucket = hash_long(fd, FIO_CLIENT_HASH_BITS) & FIO_CLIENT_HASH_MASK;
216 struct fio_client *client;
217 struct flist_head *entry;
218
219 flist_for_each(entry, &client_hash[bucket]) {
220 client = flist_entry(entry, struct fio_client, hash_list);
221
222 if (client->fd == fd) {
223 client->refs++;
224 return client;
225 }
226 }
227
228 return NULL;
229}
230
231void fio_put_client(struct fio_client *client)
232{
233 if (--client->refs)
234 return;
235
236 free(client->hostname);
237 if (client->argv)
238 free(client->argv);
239 if (client->name)
240 free(client->name);
241 while (client->nr_files) {
242 struct client_file *cf = &client->files[--client->nr_files];
243
244 free(cf->file);
245 }
246 if (client->files)
247 free(client->files);
248 if (client->opt_lists)
249 free(client->opt_lists);
250
251 if (!client->did_stat)
252 sum_stat_clients--;
253
254 if (client->error)
255 error_clients++;
256
257 free(client);
258}
259
260static int fio_client_dec_jobs_eta(struct client_eta *eta, client_eta_op eta_fn)
261{
262 if (!--eta->pending) {
263 eta_fn(&eta->eta);
264 free(eta);
265 return 0;
266 }
267
268 return 1;
269}
270
271static void fio_drain_client_text(struct fio_client *client)
272{
273 do {
274 struct fio_net_cmd *cmd;
275
276 cmd = fio_net_recv_cmd(client->fd, false);
277 if (!cmd)
278 break;
279
280 if (cmd->opcode == FIO_NET_CMD_TEXT) {
281 convert_text(cmd);
282 client->ops->text(client, cmd);
283 }
284
285 free(cmd);
286 } while (1);
287}
288
289static void remove_client(struct fio_client *client)
290{
291 assert(client->refs);
292
293 dprint(FD_NET, "client: removed <%s>\n", client->hostname);
294
295 fio_drain_client_text(client);
296
297 if (!flist_empty(&client->list))
298 flist_del_init(&client->list);
299
300 fio_client_remove_hash(client);
301
302 if (!flist_empty(&client->eta_list)) {
303 flist_del_init(&client->eta_list);
304 fio_client_dec_jobs_eta(client->eta_in_flight, client->ops->eta);
305 }
306
307 close(client->fd);
308 client->fd = -1;
309
310 if (client->ops->removed)
311 client->ops->removed(client);
312
313 nr_clients--;
314 fio_put_client(client);
315}
316
317struct fio_client *fio_get_client(struct fio_client *client)
318{
319 client->refs++;
320 return client;
321}
322
323static void __fio_client_add_cmd_option(struct fio_client *client,
324 const char *opt)
325{
326 int index;
327
328 index = client->argc++;
329 client->argv = realloc(client->argv, sizeof(char *) * client->argc);
330 client->argv[index] = strdup(opt);
331 dprint(FD_NET, "client: add cmd %d: %s\n", index, opt);
332}
333
334void fio_client_add_cmd_option(void *cookie, const char *opt)
335{
336 struct fio_client *client = cookie;
337 struct flist_head *entry;
338
339 if (!client || !opt)
340 return;
341
342 __fio_client_add_cmd_option(client, opt);
343
344 /*
345 * Duplicate arguments to shared client group
346 */
347 flist_for_each(entry, &arg_list) {
348 client = flist_entry(entry, struct fio_client, arg_list);
349
350 __fio_client_add_cmd_option(client, opt);
351 }
352}
353
354struct fio_client *fio_client_add_explicit(struct client_ops *ops,
355 const char *hostname, int type,
356 int port)
357{
358 struct fio_client *client;
359
360 client = malloc(sizeof(*client));
361 memset(client, 0, sizeof(*client));
362
363 INIT_FLIST_HEAD(&client->list);
364 INIT_FLIST_HEAD(&client->hash_list);
365 INIT_FLIST_HEAD(&client->arg_list);
366 INIT_FLIST_HEAD(&client->eta_list);
367 INIT_FLIST_HEAD(&client->cmd_list);
368
369 client->hostname = strdup(hostname);
370
371 if (type == Fio_client_socket)
372 client->is_sock = true;
373 else {
374 int ipv6;
375
376 ipv6 = type == Fio_client_ipv6;
377 if (fio_server_parse_host(hostname, ipv6,
378 &client->addr.sin_addr,
379 &client->addr6.sin6_addr))
380 goto err;
381
382 client->port = port;
383 }
384
385 client->fd = -1;
386 client->ops = ops;
387 client->refs = 1;
388 client->type = ops->client_type;
389
390 __fio_client_add_cmd_option(client, "fio");
391
392 flist_add(&client->list, &client_list);
393 nr_clients++;
394 dprint(FD_NET, "client: added <%s>\n", client->hostname);
395 return client;
396err:
397 free(client);
398 return NULL;
399}
400
401int fio_client_add_ini_file(void *cookie, const char *ini_file, bool remote)
402{
403 struct fio_client *client = cookie;
404 struct client_file *cf;
405 size_t new_size;
406 void *new_files;
407
408 if (!client)
409 return 1;
410
411 dprint(FD_NET, "client <%s>: add ini %s\n", client->hostname, ini_file);
412
413 new_size = (client->nr_files + 1) * sizeof(struct client_file);
414 new_files = realloc(client->files, new_size);
415 if (!new_files)
416 return 1;
417
418 client->files = new_files;
419 cf = &client->files[client->nr_files];
420 cf->file = strdup(ini_file);
421 cf->remote = remote;
422 client->nr_files++;
423 return 0;
424}
425
426int fio_client_add(struct client_ops *ops, const char *hostname, void **cookie)
427{
428 struct fio_client *existing = *cookie;
429 struct fio_client *client;
430
431 if (existing) {
432 /*
433 * We always add our "exec" name as the option, hence 1
434 * means empty.
435 */
436 if (existing->argc == 1)
437 flist_add_tail(&existing->arg_list, &arg_list);
438 else {
439 while (!flist_empty(&arg_list))
440 flist_del_init(arg_list.next);
441 }
442 }
443
444 client = malloc(sizeof(*client));
445 memset(client, 0, sizeof(*client));
446
447 INIT_FLIST_HEAD(&client->list);
448 INIT_FLIST_HEAD(&client->hash_list);
449 INIT_FLIST_HEAD(&client->arg_list);
450 INIT_FLIST_HEAD(&client->eta_list);
451 INIT_FLIST_HEAD(&client->cmd_list);
452
453 if (fio_server_parse_string(hostname, &client->hostname,
454 &client->is_sock, &client->port,
455 &client->addr.sin_addr,
456 &client->addr6.sin6_addr,
457 &client->ipv6))
458 return -1;
459
460 client->fd = -1;
461 client->ops = ops;
462 client->refs = 1;
463 client->type = ops->client_type;
464
465 __fio_client_add_cmd_option(client, "fio");
466
467 flist_add(&client->list, &client_list);
468 nr_clients++;
469 dprint(FD_NET, "client: added <%s>\n", client->hostname);
470 *cookie = client;
471 return 0;
472}
473
474static const char *server_name(struct fio_client *client, char *buf,
475 size_t bufsize)
476{
477 const char *from;
478
479 if (client->ipv6)
480 from = inet_ntop(AF_INET6, (struct sockaddr *) &client->addr6.sin6_addr, buf, bufsize);
481 else if (client->is_sock)
482 from = "sock";
483 else
484 from = inet_ntop(AF_INET, (struct sockaddr *) &client->addr.sin_addr, buf, bufsize);
485
486 return from;
487}
488
489static void probe_client(struct fio_client *client)
490{
491 struct cmd_client_probe_pdu pdu;
492 const char *sname;
493 uint64_t tag;
494 char buf[64];
495
496 dprint(FD_NET, "client: send probe\n");
497
498#ifdef CONFIG_ZLIB
499 pdu.flags = __le64_to_cpu(FIO_PROBE_FLAG_ZLIB);
500#else
501 pdu.flags = 0;
502#endif
503
504 sname = server_name(client, buf, sizeof(buf));
505 memset(pdu.server, 0, sizeof(pdu.server));
506 strncpy((char *) pdu.server, sname, sizeof(pdu.server) - 1);
507
508 fio_net_send_cmd(client->fd, FIO_NET_CMD_PROBE, &pdu, sizeof(pdu), &tag, &client->cmd_list);
509}
510
511static int fio_client_connect_ip(struct fio_client *client)
512{
513 struct sockaddr *addr;
514 socklen_t socklen;
515 int fd, domain;
516
517 if (client->ipv6) {
518 client->addr6.sin6_family = AF_INET6;
519 client->addr6.sin6_port = htons(client->port);
520 domain = AF_INET6;
521 addr = (struct sockaddr *) &client->addr6;
522 socklen = sizeof(client->addr6);
523 } else {
524 client->addr.sin_family = AF_INET;
525 client->addr.sin_port = htons(client->port);
526 domain = AF_INET;
527 addr = (struct sockaddr *) &client->addr;
528 socklen = sizeof(client->addr);
529 }
530
531 fd = socket(domain, SOCK_STREAM, 0);
532 if (fd < 0) {
533 int ret = -errno;
534
535 log_err("fio: socket: %s\n", strerror(errno));
536 return ret;
537 }
538
539 if (connect(fd, addr, socklen) < 0) {
540 int ret = -errno;
541
542 log_err("fio: connect: %s\n", strerror(errno));
543 log_err("fio: failed to connect to %s:%u\n", client->hostname,
544 client->port);
545 close(fd);
546 return ret;
547 }
548
549 return fd;
550}
551
552static int fio_client_connect_sock(struct fio_client *client)
553{
554 struct sockaddr_un *addr = &client->addr_un;
555 socklen_t len;
556 int fd;
557
558 memset(addr, 0, sizeof(*addr));
559 addr->sun_family = AF_UNIX;
560 strncpy(addr->sun_path, client->hostname, sizeof(addr->sun_path) - 1);
561
562 fd = socket(AF_UNIX, SOCK_STREAM, 0);
563 if (fd < 0) {
564 int ret = -errno;
565
566 log_err("fio: socket: %s\n", strerror(errno));
567 return ret;
568 }
569
570 len = sizeof(addr->sun_family) + strlen(addr->sun_path) + 1;
571 if (connect(fd, (struct sockaddr *) addr, len) < 0) {
572 int ret = -errno;
573
574 log_err("fio: connect; %s\n", strerror(errno));
575 close(fd);
576 return ret;
577 }
578
579 return fd;
580}
581
582int fio_client_connect(struct fio_client *client)
583{
584 int fd;
585
586 dprint(FD_NET, "client: connect to host %s\n", client->hostname);
587
588 if (client->is_sock)
589 fd = fio_client_connect_sock(client);
590 else
591 fd = fio_client_connect_ip(client);
592
593 dprint(FD_NET, "client: %s connected %d\n", client->hostname, fd);
594
595 if (fd < 0)
596 return fd;
597
598 client->fd = fd;
599 fio_client_add_hash(client);
600 client->state = Client_connected;
601
602 probe_client(client);
603 return 0;
604}
605
606int fio_client_terminate(struct fio_client *client)
607{
608 return fio_net_send_quit(client->fd);
609}
610
611static void fio_clients_terminate(void)
612{
613 struct flist_head *entry;
614 struct fio_client *client;
615
616 dprint(FD_NET, "client: terminate clients\n");
617
618 flist_for_each(entry, &client_list) {
619 client = flist_entry(entry, struct fio_client, list);
620 fio_client_terminate(client);
621 }
622}
623
624static void sig_int(int sig)
625{
626 dprint(FD_NET, "client: got signal %d\n", sig);
627 fio_clients_terminate();
628}
629
630static void client_signal_handler(void)
631{
632 struct sigaction act;
633
634 memset(&act, 0, sizeof(act));
635 act.sa_handler = sig_int;
636 act.sa_flags = SA_RESTART;
637 sigaction(SIGINT, &act, NULL);
638
639 memset(&act, 0, sizeof(act));
640 act.sa_handler = sig_int;
641 act.sa_flags = SA_RESTART;
642 sigaction(SIGTERM, &act, NULL);
643
644/* Windows uses SIGBREAK as a quit signal from other applications */
645#ifdef WIN32
646 memset(&act, 0, sizeof(act));
647 act.sa_handler = sig_int;
648 act.sa_flags = SA_RESTART;
649 sigaction(SIGBREAK, &act, NULL);
650#endif
651
652 memset(&act, 0, sizeof(act));
653 act.sa_handler = sig_show_status;
654 act.sa_flags = SA_RESTART;
655 sigaction(SIGUSR1, &act, NULL);
656}
657
658static int send_client_cmd_line(struct fio_client *client)
659{
660 struct cmd_single_line_pdu *cslp;
661 struct cmd_line_pdu *clp;
662 unsigned long offset;
663 unsigned int *lens;
664 void *pdu;
665 size_t mem;
666 int i, ret;
667
668 dprint(FD_NET, "client: send cmdline %d\n", client->argc);
669
670 lens = malloc(client->argc * sizeof(unsigned int));
671
672 /*
673 * Find out how much mem we need
674 */
675 for (i = 0, mem = 0; i < client->argc; i++) {
676 lens[i] = strlen(client->argv[i]) + 1;
677 mem += lens[i];
678 }
679
680 /*
681 * We need one cmd_line_pdu, and argc number of cmd_single_line_pdu
682 */
683 mem += sizeof(*clp) + (client->argc * sizeof(*cslp));
684
685 pdu = malloc(mem);
686 clp = pdu;
687 offset = sizeof(*clp);
688
689 for (i = 0; i < client->argc; i++) {
690 uint16_t arg_len = lens[i];
691
692 cslp = pdu + offset;
693 strcpy((char *) cslp->text, client->argv[i]);
694 cslp->len = cpu_to_le16(arg_len);
695 offset += sizeof(*cslp) + arg_len;
696 }
697
698 free(lens);
699 clp->lines = cpu_to_le16(client->argc);
700 clp->client_type = __cpu_to_le16(client->type);
701 ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_JOBLINE, pdu, mem, NULL, NULL);
702 free(pdu);
703 return ret;
704}
705
706int fio_clients_connect(void)
707{
708 struct fio_client *client;
709 struct flist_head *entry, *tmp;
710 int ret;
711
712#ifdef WIN32
713 WSADATA wsd;
714 WSAStartup(MAKEWORD(2, 2), &wsd);
715#endif
716
717 dprint(FD_NET, "client: connect all\n");
718
719 client_signal_handler();
720
721 flist_for_each_safe(entry, tmp, &client_list) {
722 client = flist_entry(entry, struct fio_client, list);
723
724 ret = fio_client_connect(client);
725 if (ret) {
726 remove_client(client);
727 continue;
728 }
729
730 if (client->argc > 1)
731 send_client_cmd_line(client);
732 }
733
734 return !nr_clients;
735}
736
737int fio_start_client(struct fio_client *client)
738{
739 dprint(FD_NET, "client: start %s\n", client->hostname);
740 return fio_net_send_simple_cmd(client->fd, FIO_NET_CMD_RUN, 0, NULL);
741}
742
743int fio_start_all_clients(void)
744{
745 struct fio_client *client;
746 struct flist_head *entry, *tmp;
747 int ret;
748
749 dprint(FD_NET, "client: start all\n");
750
751 fio_client_json_init();
752
753 flist_for_each_safe(entry, tmp, &client_list) {
754 client = flist_entry(entry, struct fio_client, list);
755
756 ret = fio_start_client(client);
757 if (ret) {
758 remove_client(client);
759 continue;
760 }
761 }
762
763 return flist_empty(&client_list);
764}
765
766static int __fio_client_send_remote_ini(struct fio_client *client,
767 const char *filename)
768{
769 struct cmd_load_file_pdu *pdu;
770 size_t p_size;
771 int ret;
772
773 dprint(FD_NET, "send remote ini %s to %s\n", filename, client->hostname);
774
775 p_size = sizeof(*pdu) + strlen(filename) + 1;
776 pdu = malloc(p_size);
777 memset(pdu, 0, p_size);
778 pdu->name_len = strlen(filename);
779 strcpy((char *) pdu->file, filename);
780 pdu->client_type = cpu_to_le16((uint16_t) client->type);
781
782 client->sent_job = true;
783 ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_LOAD_FILE, pdu, p_size,NULL, NULL);
784 free(pdu);
785 return ret;
786}
787
788/*
789 * Send file contents to server backend. We could use sendfile(), but to remain
790 * more portable lets just read/write the darn thing.
791 */
792static int __fio_client_send_local_ini(struct fio_client *client,
793 const char *filename)
794{
795 struct cmd_job_pdu *pdu;
796 size_t p_size;
797 struct stat sb;
798 char *p;
799 void *buf;
800 off_t len;
801 int fd, ret;
802
803 dprint(FD_NET, "send ini %s to %s\n", filename, client->hostname);
804
805 fd = open(filename, O_RDONLY);
806 if (fd < 0) {
807 ret = -errno;
808 log_err("fio: job file <%s> open: %s\n", filename, strerror(errno));
809 return ret;
810 }
811
812 if (fstat(fd, &sb) < 0) {
813 ret = -errno;
814 log_err("fio: job file stat: %s\n", strerror(errno));
815 close(fd);
816 return ret;
817 }
818
819 /*
820 * Add extra space for variable expansion, but doesn't guarantee.
821 */
822 sb.st_size += OPT_LEN_MAX;
823 p_size = sb.st_size + sizeof(*pdu);
824 pdu = malloc(p_size);
825 buf = pdu->buf;
826
827 len = sb.st_size;
828 p = buf;
829 if (read_ini_data(fd, p, len)) {
830 log_err("fio: failed reading job file %s\n", filename);
831 close(fd);
832 free(pdu);
833 return 1;
834 }
835
836 pdu->buf_len = __cpu_to_le32(sb.st_size);
837 pdu->client_type = cpu_to_le32(client->type);
838
839 client->sent_job = true;
840 ret = fio_net_send_cmd(client->fd, FIO_NET_CMD_JOB, pdu, p_size, NULL, NULL);
841 free(pdu);
842 close(fd);
843 return ret;
844}
845
846int fio_client_send_ini(struct fio_client *client, const char *filename,
847 bool remote)
848{
849 int ret;
850
851 if (!remote)
852 ret = __fio_client_send_local_ini(client, filename);
853 else
854 ret = __fio_client_send_remote_ini(client, filename);
855
856 if (!ret)
857 client->sent_job = true;
858
859 return ret;
860}
861
862static int fio_client_send_cf(struct fio_client *client,
863 struct client_file *cf)
864{
865 return fio_client_send_ini(client, cf->file, cf->remote);
866}
867
868int fio_clients_send_ini(const char *filename)
869{
870 struct fio_client *client;
871 struct flist_head *entry, *tmp;
872
873 flist_for_each_safe(entry, tmp, &client_list) {
874 bool failed = false;
875
876 client = flist_entry(entry, struct fio_client, list);
877
878 if (client->nr_files) {
879 int i;
880
881 for (i = 0; i < client->nr_files; i++) {
882 struct client_file *cf;
883
884 cf = &client->files[i];
885
886 if (fio_client_send_cf(client, cf)) {
887 failed = true;
888 remove_client(client);
889 break;
890 }
891 }
892 }
893 if (client->sent_job || failed)
894 continue;
895 if (!filename || fio_client_send_ini(client, filename, 0))
896 remove_client(client);
897 }
898
899 return !nr_clients;
900}
901
902int fio_client_update_options(struct fio_client *client,
903 struct thread_options *o, uint64_t *tag)
904{
905 struct cmd_add_job_pdu pdu;
906
907 pdu.thread_number = cpu_to_le32(client->thread_number);
908 pdu.groupid = cpu_to_le32(client->groupid);
909 convert_thread_options_to_net(&pdu.top, o);
910
911 return fio_net_send_cmd(client->fd, FIO_NET_CMD_UPDATE_JOB, &pdu, sizeof(pdu), tag, &client->cmd_list);
912}
913
914static void convert_io_stat(struct io_stat *dst, struct io_stat *src)
915{
916 dst->max_val = le64_to_cpu(src->max_val);
917 dst->min_val = le64_to_cpu(src->min_val);
918 dst->samples = le64_to_cpu(src->samples);
919
920 /*
921 * Floats arrive as IEEE 754 encoded uint64_t, convert back to double
922 */
923 dst->mean.u.f = fio_uint64_to_double(le64_to_cpu(dst->mean.u.i));
924 dst->S.u.f = fio_uint64_to_double(le64_to_cpu(dst->S.u.i));
925}
926
927static void convert_ts(struct thread_stat *dst, struct thread_stat *src)
928{
929 int i, j;
930
931 dst->error = le32_to_cpu(src->error);
932 dst->thread_number = le32_to_cpu(src->thread_number);
933 dst->groupid = le32_to_cpu(src->groupid);
934 dst->pid = le32_to_cpu(src->pid);
935 dst->members = le32_to_cpu(src->members);
936 dst->unified_rw_rep = le32_to_cpu(src->unified_rw_rep);
937
938 for (i = 0; i < DDIR_RWDIR_CNT; i++) {
939 convert_io_stat(&dst->clat_stat[i], &src->clat_stat[i]);
940 convert_io_stat(&dst->slat_stat[i], &src->slat_stat[i]);
941 convert_io_stat(&dst->lat_stat[i], &src->lat_stat[i]);
942 convert_io_stat(&dst->bw_stat[i], &src->bw_stat[i]);
943 convert_io_stat(&dst->iops_stat[i], &src->iops_stat[i]);
944 }
945
946 dst->usr_time = le64_to_cpu(src->usr_time);
947 dst->sys_time = le64_to_cpu(src->sys_time);
948 dst->ctx = le64_to_cpu(src->ctx);
949 dst->minf = le64_to_cpu(src->minf);
950 dst->majf = le64_to_cpu(src->majf);
951 dst->clat_percentiles = le32_to_cpu(src->clat_percentiles);
952 dst->lat_percentiles = le32_to_cpu(src->lat_percentiles);
953 dst->percentile_precision = le64_to_cpu(src->percentile_precision);
954
955 for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++) {
956 fio_fp64_t *fps = &src->percentile_list[i];
957 fio_fp64_t *fpd = &dst->percentile_list[i];
958
959 fpd->u.f = fio_uint64_to_double(le64_to_cpu(fps->u.i));
960 }
961
962 for (i = 0; i < FIO_IO_U_MAP_NR; i++) {
963 dst->io_u_map[i] = le64_to_cpu(src->io_u_map[i]);
964 dst->io_u_submit[i] = le64_to_cpu(src->io_u_submit[i]);
965 dst->io_u_complete[i] = le64_to_cpu(src->io_u_complete[i]);
966 }
967
968 for (i = 0; i < FIO_IO_U_LAT_N_NR; i++)
969 dst->io_u_lat_n[i] = le64_to_cpu(src->io_u_lat_n[i]);
970 for (i = 0; i < FIO_IO_U_LAT_U_NR; i++)
971 dst->io_u_lat_u[i] = le64_to_cpu(src->io_u_lat_u[i]);
972 for (i = 0; i < FIO_IO_U_LAT_M_NR; i++)
973 dst->io_u_lat_m[i] = le64_to_cpu(src->io_u_lat_m[i]);
974
975 for (i = 0; i < DDIR_RWDIR_CNT; i++)
976 for (j = 0; j < FIO_IO_U_PLAT_NR; j++)
977 dst->io_u_plat[i][j] = le64_to_cpu(src->io_u_plat[i][j]);
978
979 for (i = 0; i < DDIR_RWDIR_CNT; i++) {
980 dst->total_io_u[i] = le64_to_cpu(src->total_io_u[i]);
981 dst->short_io_u[i] = le64_to_cpu(src->short_io_u[i]);
982 dst->drop_io_u[i] = le64_to_cpu(src->drop_io_u[i]);
983 }
984
985 dst->total_submit = le64_to_cpu(src->total_submit);
986 dst->total_complete = le64_to_cpu(src->total_complete);
987
988 for (i = 0; i < DDIR_RWDIR_CNT; i++) {
989 dst->io_bytes[i] = le64_to_cpu(src->io_bytes[i]);
990 dst->runtime[i] = le64_to_cpu(src->runtime[i]);
991 }
992
993 dst->total_run_time = le64_to_cpu(src->total_run_time);
994 dst->continue_on_error = le16_to_cpu(src->continue_on_error);
995 dst->total_err_count = le64_to_cpu(src->total_err_count);
996 dst->first_error = le32_to_cpu(src->first_error);
997 dst->kb_base = le32_to_cpu(src->kb_base);
998 dst->unit_base = le32_to_cpu(src->unit_base);
999
1000 dst->sig_figs = le32_to_cpu(src->sig_figs);
1001
1002 dst->latency_depth = le32_to_cpu(src->latency_depth);
1003 dst->latency_target = le64_to_cpu(src->latency_target);
1004 dst->latency_window = le64_to_cpu(src->latency_window);
1005 dst->latency_percentile.u.f = fio_uint64_to_double(le64_to_cpu(src->latency_percentile.u.i));
1006
1007 dst->nr_block_infos = le64_to_cpu(src->nr_block_infos);
1008 for (i = 0; i < dst->nr_block_infos; i++)
1009 dst->block_infos[i] = le32_to_cpu(src->block_infos[i]);
1010
1011 dst->ss_dur = le64_to_cpu(src->ss_dur);
1012 dst->ss_state = le32_to_cpu(src->ss_state);
1013 dst->ss_head = le32_to_cpu(src->ss_head);
1014 dst->ss_limit.u.f = fio_uint64_to_double(le64_to_cpu(src->ss_limit.u.i));
1015 dst->ss_slope.u.f = fio_uint64_to_double(le64_to_cpu(src->ss_slope.u.i));
1016 dst->ss_deviation.u.f = fio_uint64_to_double(le64_to_cpu(src->ss_deviation.u.i));
1017 dst->ss_criterion.u.f = fio_uint64_to_double(le64_to_cpu(src->ss_criterion.u.i));
1018
1019 if (dst->ss_state & FIO_SS_DATA) {
1020 for (i = 0; i < dst->ss_dur; i++ ) {
1021 dst->ss_iops_data[i] = le64_to_cpu(src->ss_iops_data[i]);
1022 dst->ss_bw_data[i] = le64_to_cpu(src->ss_bw_data[i]);
1023 }
1024 }
1025}
1026
1027static void convert_gs(struct group_run_stats *dst, struct group_run_stats *src)
1028{
1029 int i;
1030
1031 for (i = 0; i < DDIR_RWDIR_CNT; i++) {
1032 dst->max_run[i] = le64_to_cpu(src->max_run[i]);
1033 dst->min_run[i] = le64_to_cpu(src->min_run[i]);
1034 dst->max_bw[i] = le64_to_cpu(src->max_bw[i]);
1035 dst->min_bw[i] = le64_to_cpu(src->min_bw[i]);
1036 dst->iobytes[i] = le64_to_cpu(src->iobytes[i]);
1037 dst->agg[i] = le64_to_cpu(src->agg[i]);
1038 }
1039
1040 dst->kb_base = le32_to_cpu(src->kb_base);
1041 dst->unit_base = le32_to_cpu(src->unit_base);
1042 dst->sig_figs = le32_to_cpu(src->sig_figs);
1043 dst->groupid = le32_to_cpu(src->groupid);
1044 dst->unified_rw_rep = le32_to_cpu(src->unified_rw_rep);
1045}
1046
1047static void json_object_add_client_info(struct json_object *obj,
1048 struct fio_client *client)
1049{
1050 const char *hostname = client->hostname ? client->hostname : "";
1051
1052 json_object_add_value_string(obj, "hostname", hostname);
1053 json_object_add_value_int(obj, "port", client->port);
1054}
1055
1056static void handle_ts(struct fio_client *client, struct fio_net_cmd *cmd)
1057{
1058 struct cmd_ts_pdu *p = (struct cmd_ts_pdu *) cmd->payload;
1059 struct flist_head *opt_list = NULL;
1060 struct json_object *tsobj;
1061
1062 if (output_format & FIO_OUTPUT_TERSE)
1063 return;
1064
1065 if (client->opt_lists && p->ts.thread_number <= client->jobs)
1066 opt_list = &client->opt_lists[p->ts.thread_number - 1];
1067
1068 tsobj = show_thread_status(&p->ts, &p->rs, opt_list, NULL);
1069 client->did_stat = true;
1070 if (tsobj) {
1071 json_object_add_client_info(tsobj, client);
1072 json_array_add_value_object(clients_array, tsobj);
1073 }
1074
1075 if (sum_stat_clients <= 1)
1076 return;
1077
1078 sum_thread_stats(&client_ts, &p->ts, sum_stat_nr == 1);
1079 sum_group_stats(&client_gs, &p->rs);
1080
1081 client_ts.members++;
1082 client_ts.thread_number = p->ts.thread_number;
1083 client_ts.groupid = p->ts.groupid;
1084 client_ts.unified_rw_rep = p->ts.unified_rw_rep;
1085 client_ts.sig_figs = p->ts.sig_figs;
1086
1087 if (++sum_stat_nr == sum_stat_clients) {
1088 strcpy(client_ts.name, "All clients");
1089 tsobj = show_thread_status(&client_ts, &client_gs, NULL, NULL);
1090 if (tsobj) {
1091 json_object_add_client_info(tsobj, client);
1092 json_array_add_value_object(clients_array, tsobj);
1093 }
1094 }
1095}
1096
1097static void handle_gs(struct fio_client *client, struct fio_net_cmd *cmd)
1098{
1099 struct group_run_stats *gs = (struct group_run_stats *) cmd->payload;
1100
1101 if (output_format & FIO_OUTPUT_TERSE)
1102 return;
1103
1104 if (output_format & FIO_OUTPUT_NORMAL)
1105 show_group_stats(gs, NULL);
1106}
1107
1108static void handle_job_opt(struct fio_client *client, struct fio_net_cmd *cmd)
1109{
1110 struct cmd_job_option *pdu = (struct cmd_job_option *) cmd->payload;
1111 struct print_option *p;
1112
1113 if (!job_opt_object)
1114 return;
1115
1116 pdu->global = le16_to_cpu(pdu->global);
1117 pdu->truncated = le16_to_cpu(pdu->truncated);
1118 pdu->groupid = le32_to_cpu(pdu->groupid);
1119
1120 p = malloc(sizeof(*p));
1121 p->name = strdup((char *) pdu->name);
1122 if (pdu->value[0] != '\0')
1123 p->value = strdup((char *) pdu->value);
1124 else
1125 p->value = NULL;
1126
1127 if (pdu->global) {
1128 const char *pos = "";
1129
1130 if (p->value)
1131 pos = p->value;
1132
1133 json_object_add_value_string(job_opt_object, p->name, pos);
1134 } else if (client->opt_lists) {
1135 struct flist_head *opt_list = &client->opt_lists[pdu->groupid];
1136
1137 flist_add_tail(&p->list, opt_list);
1138 }
1139}
1140
1141static void handle_text(struct fio_client *client, struct fio_net_cmd *cmd)
1142{
1143 struct cmd_text_pdu *pdu = (struct cmd_text_pdu *) cmd->payload;
1144 const char *buf = (const char *) pdu->buf;
1145 const char *name;
1146 int fio_unused ret;
1147
1148 name = client->name ? client->name : client->hostname;
1149
1150 if (!client->skip_newline && !(output_format & FIO_OUTPUT_TERSE))
1151 fprintf(f_out, "<%s> ", name);
1152 ret = fwrite(buf, pdu->buf_len, 1, f_out);
1153 fflush(f_out);
1154 client->skip_newline = strchr(buf, '\n') == NULL;
1155}
1156
1157static void convert_agg(struct disk_util_agg *agg)
1158{
1159 int i;
1160
1161 for (i = 0; i < 2; i++) {
1162 agg->ios[i] = le64_to_cpu(agg->ios[i]);
1163 agg->merges[i] = le64_to_cpu(agg->merges[i]);
1164 agg->sectors[i] = le64_to_cpu(agg->sectors[i]);
1165 agg->ticks[i] = le64_to_cpu(agg->ticks[i]);
1166 }
1167
1168 agg->io_ticks = le64_to_cpu(agg->io_ticks);
1169 agg->time_in_queue = le64_to_cpu(agg->time_in_queue);
1170 agg->slavecount = le32_to_cpu(agg->slavecount);
1171 agg->max_util.u.f = fio_uint64_to_double(le64_to_cpu(agg->max_util.u.i));
1172}
1173
1174static void convert_dus(struct disk_util_stat *dus)
1175{
1176 int i;
1177
1178 for (i = 0; i < 2; i++) {
1179 dus->s.ios[i] = le64_to_cpu(dus->s.ios[i]);
1180 dus->s.merges[i] = le64_to_cpu(dus->s.merges[i]);
1181 dus->s.sectors[i] = le64_to_cpu(dus->s.sectors[i]);
1182 dus->s.ticks[i] = le64_to_cpu(dus->s.ticks[i]);
1183 }
1184
1185 dus->s.io_ticks = le64_to_cpu(dus->s.io_ticks);
1186 dus->s.time_in_queue = le64_to_cpu(dus->s.time_in_queue);
1187 dus->s.msec = le64_to_cpu(dus->s.msec);
1188}
1189
1190static void handle_du(struct fio_client *client, struct fio_net_cmd *cmd)
1191{
1192 struct cmd_du_pdu *du = (struct cmd_du_pdu *) cmd->payload;
1193
1194 if (output_format & FIO_OUTPUT_TERSE)
1195 return;
1196
1197 if (!client->disk_stats_shown) {
1198 client->disk_stats_shown = true;
1199 if (!(output_format & FIO_OUTPUT_JSON))
1200 log_info("\nDisk stats (read/write):\n");
1201 }
1202
1203 if (output_format & FIO_OUTPUT_JSON) {
1204 struct json_object *duobj;
1205 json_array_add_disk_util(&du->dus, &du->agg, du_array);
1206 duobj = json_array_last_value_object(du_array);
1207 json_object_add_client_info(duobj, client);
1208 }
1209 if (output_format & FIO_OUTPUT_NORMAL)
1210 print_disk_util(&du->dus, &du->agg, 0, NULL);
1211}
1212
1213static void convert_jobs_eta(struct jobs_eta *je)
1214{
1215 int i;
1216
1217 je->nr_running = le32_to_cpu(je->nr_running);
1218 je->nr_ramp = le32_to_cpu(je->nr_ramp);
1219 je->nr_pending = le32_to_cpu(je->nr_pending);
1220 je->nr_setting_up = le32_to_cpu(je->nr_setting_up);
1221 je->files_open = le32_to_cpu(je->files_open);
1222
1223 for (i = 0; i < DDIR_RWDIR_CNT; i++) {
1224 je->m_rate[i] = le64_to_cpu(je->m_rate[i]);
1225 je->t_rate[i] = le64_to_cpu(je->t_rate[i]);
1226 je->m_iops[i] = le32_to_cpu(je->m_iops[i]);
1227 je->t_iops[i] = le32_to_cpu(je->t_iops[i]);
1228 je->rate[i] = le64_to_cpu(je->rate[i]);
1229 je->iops[i] = le32_to_cpu(je->iops[i]);
1230 }
1231
1232 je->elapsed_sec = le64_to_cpu(je->elapsed_sec);
1233 je->eta_sec = le64_to_cpu(je->eta_sec);
1234 je->nr_threads = le32_to_cpu(je->nr_threads);
1235 je->is_pow2 = le32_to_cpu(je->is_pow2);
1236 je->unit_base = le32_to_cpu(je->unit_base);
1237 je->sig_figs = le32_to_cpu(je->sig_figs);
1238}
1239
1240void fio_client_sum_jobs_eta(struct jobs_eta *dst, struct jobs_eta *je)
1241{
1242 int i;
1243
1244 dst->nr_running += je->nr_running;
1245 dst->nr_ramp += je->nr_ramp;
1246 dst->nr_pending += je->nr_pending;
1247 dst->nr_setting_up += je->nr_setting_up;
1248 dst->files_open += je->files_open;
1249
1250 for (i = 0; i < DDIR_RWDIR_CNT; i++) {
1251 dst->m_rate[i] += je->m_rate[i];
1252 dst->t_rate[i] += je->t_rate[i];
1253 dst->m_iops[i] += je->m_iops[i];
1254 dst->t_iops[i] += je->t_iops[i];
1255 dst->rate[i] += je->rate[i];
1256 dst->iops[i] += je->iops[i];
1257 }
1258
1259 dst->elapsed_sec += je->elapsed_sec;
1260
1261 if (je->eta_sec > dst->eta_sec)
1262 dst->eta_sec = je->eta_sec;
1263
1264 dst->nr_threads += je->nr_threads;
1265
1266 /*
1267 * This wont be correct for multiple strings, but at least it
1268 * works for the basic cases.
1269 */
1270 strcpy((char *) dst->run_str, (char *) je->run_str);
1271}
1272
1273static bool remove_reply_cmd(struct fio_client *client, struct fio_net_cmd *cmd)
1274{
1275 struct fio_net_cmd_reply *reply = NULL;
1276 struct flist_head *entry;
1277
1278 flist_for_each(entry, &client->cmd_list) {
1279 reply = flist_entry(entry, struct fio_net_cmd_reply, list);
1280
1281 if (cmd->tag == (uintptr_t) reply)
1282 break;
1283
1284 reply = NULL;
1285 }
1286
1287 if (!reply) {
1288 log_err("fio: client: unable to find matching tag (%llx)\n", (unsigned long long) cmd->tag);
1289 return false;
1290 }
1291
1292 flist_del(&reply->list);
1293 cmd->tag = reply->saved_tag;
1294 free(reply);
1295 return true;
1296}
1297
1298int fio_client_wait_for_reply(struct fio_client *client, uint64_t tag)
1299{
1300 do {
1301 struct fio_net_cmd_reply *reply = NULL;
1302 struct flist_head *entry;
1303
1304 flist_for_each(entry, &client->cmd_list) {
1305 reply = flist_entry(entry, struct fio_net_cmd_reply, list);
1306
1307 if (tag == (uintptr_t) reply)
1308 break;
1309
1310 reply = NULL;
1311 }
1312
1313 if (!reply)
1314 break;
1315
1316 usleep(1000);
1317 } while (1);
1318
1319 return 0;
1320}
1321
1322static void handle_eta(struct fio_client *client, struct fio_net_cmd *cmd)
1323{
1324 struct jobs_eta *je = (struct jobs_eta *) cmd->payload;
1325 struct client_eta *eta = (struct client_eta *) (uintptr_t) cmd->tag;
1326
1327 dprint(FD_NET, "client: got eta tag %p, %d\n", eta, eta->pending);
1328
1329 assert(client->eta_in_flight == eta);
1330
1331 client->eta_in_flight = NULL;
1332 flist_del_init(&client->eta_list);
1333 client->eta_timeouts = 0;
1334
1335 if (client->ops->jobs_eta)
1336 client->ops->jobs_eta(client, je);
1337
1338 fio_client_sum_jobs_eta(&eta->eta, je);
1339 fio_client_dec_jobs_eta(eta, client->ops->eta);
1340}
1341
1342static void client_flush_hist_samples(FILE *f, int hist_coarseness, void *samples,
1343 uint64_t sample_size)
1344{
1345 struct io_sample *s;
1346 int log_offset;
1347 uint64_t i, j, nr_samples;
1348 struct io_u_plat_entry *entry;
1349 uint64_t *io_u_plat;
1350
1351 int stride = 1 << hist_coarseness;
1352
1353 if (!sample_size)
1354 return;
1355
1356 s = __get_sample(samples, 0, 0);
1357 log_offset = (s->__ddir & LOG_OFFSET_SAMPLE_BIT) != 0;
1358
1359 nr_samples = sample_size / __log_entry_sz(log_offset);
1360
1361 for (i = 0; i < nr_samples; i++) {
1362
1363 s = (struct io_sample *)((char *)__get_sample(samples, log_offset, i) +
1364 i * sizeof(struct io_u_plat_entry));
1365
1366 entry = s->data.plat_entry;
1367 io_u_plat = entry->io_u_plat;
1368
1369 fprintf(f, "%lu, %u, %llu, ", (unsigned long) s->time,
1370 io_sample_ddir(s), (unsigned long long) s->bs);
1371 for (j = 0; j < FIO_IO_U_PLAT_NR - stride; j += stride) {
1372 fprintf(f, "%llu, ", (unsigned long long)hist_sum(j, stride, io_u_plat, NULL));
1373 }
1374 fprintf(f, "%llu\n", (unsigned long long)
1375 hist_sum(FIO_IO_U_PLAT_NR - stride, stride, io_u_plat, NULL));
1376
1377 }
1378}
1379
1380static int fio_client_handle_iolog(struct fio_client *client,
1381 struct fio_net_cmd *cmd)
1382{
1383 struct cmd_iolog_pdu *pdu = NULL;
1384 bool store_direct;
1385 char *log_pathname = NULL;
1386 int ret = 0;
1387
1388 pdu = convert_iolog(cmd, &store_direct);
1389 if (!pdu) {
1390 log_err("fio: failed converting IO log\n");
1391 ret = 1;
1392 goto out;
1393 }
1394
1395 /* allocate buffer big enough for next sprintf() call */
1396 log_pathname = malloc(10 + strlen((char *)pdu->name) +
1397 strlen(client->hostname));
1398 if (!log_pathname) {
1399 log_err("fio: memory allocation of unique pathname failed\n");
1400 ret = -1;
1401 goto out;
1402 }
1403 /* generate a unique pathname for the log file using hostname */
1404 sprintf(log_pathname, "%s.%s", pdu->name, client->hostname);
1405
1406 if (store_direct) {
1407 ssize_t wrote;
1408 size_t sz;
1409 int fd;
1410
1411 fd = open((const char *) log_pathname,
1412 O_WRONLY | O_CREAT | O_TRUNC, 0644);
1413 if (fd < 0) {
1414 log_err("fio: open log %s: %s\n",
1415 log_pathname, strerror(errno));
1416 ret = 1;
1417 goto out;
1418 }
1419
1420 sz = cmd->pdu_len - sizeof(*pdu);
1421 wrote = write(fd, pdu->samples, sz);
1422 close(fd);
1423
1424 if (wrote != sz) {
1425 log_err("fio: short write on compressed log\n");
1426 ret = 1;
1427 goto out;
1428 }
1429
1430 ret = 0;
1431 } else {
1432 FILE *f;
1433 f = fopen((const char *) log_pathname, "w");
1434 if (!f) {
1435 log_err("fio: fopen log %s : %s\n",
1436 log_pathname, strerror(errno));
1437 ret = 1;
1438 goto out;
1439 }
1440
1441 if (pdu->log_type == IO_LOG_TYPE_HIST) {
1442 client_flush_hist_samples(f, pdu->log_hist_coarseness, pdu->samples,
1443 pdu->nr_samples * sizeof(struct io_sample));
1444 } else {
1445 flush_samples(f, pdu->samples,
1446 pdu->nr_samples * sizeof(struct io_sample));
1447 }
1448 fclose(f);
1449 ret = 0;
1450 }
1451
1452out:
1453 if (pdu && pdu != (void *) cmd->payload)
1454 free(pdu);
1455
1456 if (log_pathname)
1457 free(log_pathname);
1458
1459 return ret;
1460}
1461
1462static void handle_probe(struct fio_client *client, struct fio_net_cmd *cmd)
1463{
1464 struct cmd_probe_reply_pdu *probe = (struct cmd_probe_reply_pdu *) cmd->payload;
1465 const char *os, *arch;
1466 char bit[16];
1467
1468 if (output_format & FIO_OUTPUT_TERSE)
1469 return;
1470
1471 os = fio_get_os_string(probe->os);
1472 if (!os)
1473 os = "unknown";
1474
1475 arch = fio_get_arch_string(probe->arch);
1476 if (!arch)
1477 os = "unknown";
1478
1479 sprintf(bit, "%d-bit", probe->bpp * 8);
1480 probe->flags = le64_to_cpu(probe->flags);
1481
1482 if (!(output_format & FIO_OUTPUT_JSON))
1483 log_info("hostname=%s, be=%u, %s, os=%s, arch=%s, fio=%s, flags=%lx\n",
1484 probe->hostname, probe->bigendian, bit, os, arch,
1485 probe->fio_version, (unsigned long) probe->flags);
1486
1487 if (!client->name)
1488 client->name = strdup((char *) probe->hostname);
1489}
1490
1491static void handle_start(struct fio_client *client, struct fio_net_cmd *cmd)
1492{
1493 struct cmd_start_pdu *pdu = (struct cmd_start_pdu *) cmd->payload;
1494
1495 client->state = Client_started;
1496 client->jobs = le32_to_cpu(pdu->jobs);
1497 client->nr_stat = le32_to_cpu(pdu->stat_outputs);
1498
1499 if (client->jobs) {
1500 int i;
1501
1502 if (client->opt_lists)
1503 free(client->opt_lists);
1504
1505 client->opt_lists = malloc(client->jobs * sizeof(struct flist_head));
1506 for (i = 0; i < client->jobs; i++)
1507 INIT_FLIST_HEAD(&client->opt_lists[i]);
1508 }
1509
1510 sum_stat_clients += client->nr_stat;
1511}
1512
1513static void handle_stop(struct fio_client *client)
1514{
1515 if (client->error)
1516 log_info("client <%s>: exited with error %d\n", client->hostname, client->error);
1517}
1518
1519static void convert_stop(struct fio_net_cmd *cmd)
1520{
1521 struct cmd_end_pdu *pdu = (struct cmd_end_pdu *) cmd->payload;
1522
1523 pdu->error = le32_to_cpu(pdu->error);
1524}
1525
1526static void convert_text(struct fio_net_cmd *cmd)
1527{
1528 struct cmd_text_pdu *pdu = (struct cmd_text_pdu *) cmd->payload;
1529
1530 pdu->level = le32_to_cpu(pdu->level);
1531 pdu->buf_len = le32_to_cpu(pdu->buf_len);
1532 pdu->log_sec = le64_to_cpu(pdu->log_sec);
1533 pdu->log_usec = le64_to_cpu(pdu->log_usec);
1534}
1535
1536static struct cmd_iolog_pdu *convert_iolog_gz(struct fio_net_cmd *cmd,
1537 struct cmd_iolog_pdu *pdu)
1538{
1539#ifdef CONFIG_ZLIB
1540 struct cmd_iolog_pdu *ret;
1541 z_stream stream;
1542 uint64_t nr_samples;
1543 size_t total;
1544 char *p;
1545
1546 stream.zalloc = Z_NULL;
1547 stream.zfree = Z_NULL;
1548 stream.opaque = Z_NULL;
1549 stream.avail_in = 0;
1550 stream.next_in = Z_NULL;
1551
1552 if (inflateInit(&stream) != Z_OK)
1553 return NULL;
1554
1555 /*
1556 * Get header first, it's not compressed
1557 */
1558 nr_samples = le64_to_cpu(pdu->nr_samples);
1559
1560 if (pdu->log_type == IO_LOG_TYPE_HIST)
1561 total = nr_samples * (__log_entry_sz(le32_to_cpu(pdu->log_offset)) +
1562 sizeof(struct io_u_plat_entry));
1563 else
1564 total = nr_samples * __log_entry_sz(le32_to_cpu(pdu->log_offset));
1565 ret = malloc(total + sizeof(*pdu));
1566 ret->nr_samples = nr_samples;
1567
1568 memcpy(ret, pdu, sizeof(*pdu));
1569
1570 p = (char *) ret + sizeof(*pdu);
1571
1572 stream.avail_in = cmd->pdu_len - sizeof(*pdu);
1573 stream.next_in = (void *)((char *) pdu + sizeof(*pdu));
1574 while (stream.avail_in) {
1575 unsigned int this_chunk = 65536;
1576 unsigned int this_len;
1577 int err;
1578
1579 if (this_chunk > total)
1580 this_chunk = total;
1581
1582 stream.avail_out = this_chunk;
1583 stream.next_out = (void *)p;
1584 err = inflate(&stream, Z_NO_FLUSH);
1585 /* may be Z_OK, or Z_STREAM_END */
1586 if (err < 0) {
1587 log_err("fio: inflate error %d\n", err);
1588 free(ret);
1589 ret = NULL;
1590 goto err;
1591 }
1592
1593 this_len = this_chunk - stream.avail_out;
1594 p += this_len;
1595 total -= this_len;
1596 }
1597
1598err:
1599 inflateEnd(&stream);
1600 return ret;
1601#else
1602 return NULL;
1603#endif
1604}
1605
1606/*
1607 * This has been compressed on the server side, since it can be big.
1608 * Uncompress here.
1609 */
1610static struct cmd_iolog_pdu *convert_iolog(struct fio_net_cmd *cmd,
1611 bool *store_direct)
1612{
1613 struct cmd_iolog_pdu *pdu = (struct cmd_iolog_pdu *) cmd->payload;
1614 struct cmd_iolog_pdu *ret;
1615 uint64_t i;
1616 int compressed;
1617 void *samples;
1618
1619 *store_direct = false;
1620
1621 /*
1622 * Convert if compressed and we support it. If it's not
1623 * compressed, we need not do anything.
1624 */
1625 compressed = le32_to_cpu(pdu->compressed);
1626 if (compressed == XMIT_COMPRESSED) {
1627#ifndef CONFIG_ZLIB
1628 log_err("fio: server sent compressed data by mistake\n");
1629 return NULL;
1630#endif
1631 ret = convert_iolog_gz(cmd, pdu);
1632 if (!ret) {
1633 log_err("fio: failed decompressing log\n");
1634 return NULL;
1635 }
1636 } else if (compressed == STORE_COMPRESSED) {
1637 *store_direct = true;
1638 ret = pdu;
1639 } else
1640 ret = pdu;
1641
1642 ret->nr_samples = le64_to_cpu(ret->nr_samples);
1643 ret->thread_number = le32_to_cpu(ret->thread_number);
1644 ret->log_type = le32_to_cpu(ret->log_type);
1645 ret->compressed = le32_to_cpu(ret->compressed);
1646 ret->log_offset = le32_to_cpu(ret->log_offset);
1647 ret->log_hist_coarseness = le32_to_cpu(ret->log_hist_coarseness);
1648
1649 if (*store_direct)
1650 return ret;
1651
1652 samples = &ret->samples[0];
1653 for (i = 0; i < ret->nr_samples; i++) {
1654 struct io_sample *s;
1655
1656 s = __get_sample(samples, ret->log_offset, i);
1657 if (ret->log_type == IO_LOG_TYPE_HIST)
1658 s = (struct io_sample *)((char *)s + sizeof(struct io_u_plat_entry) * i);
1659
1660 s->time = le64_to_cpu(s->time);
1661 s->data.val = le64_to_cpu(s->data.val);
1662 s->__ddir = le32_to_cpu(s->__ddir);
1663 s->bs = le64_to_cpu(s->bs);
1664
1665 if (ret->log_offset) {
1666 struct io_sample_offset *so = (void *) s;
1667
1668 so->offset = le64_to_cpu(so->offset);
1669 }
1670
1671 if (ret->log_type == IO_LOG_TYPE_HIST) {
1672 s->data.plat_entry = (struct io_u_plat_entry *)(((char *)s) + sizeof(*s));
1673 s->data.plat_entry->list.next = NULL;
1674 s->data.plat_entry->list.prev = NULL;
1675 }
1676 }
1677
1678 return ret;
1679}
1680
1681static void sendfile_reply(int fd, struct cmd_sendfile_reply *rep,
1682 size_t size, uint64_t tag)
1683{
1684 rep->error = cpu_to_le32(rep->error);
1685 fio_net_send_cmd(fd, FIO_NET_CMD_SENDFILE, rep, size, &tag, NULL);
1686}
1687
1688static int fio_send_file(struct fio_client *client, struct cmd_sendfile *pdu,
1689 uint64_t tag)
1690{
1691 struct cmd_sendfile_reply *rep;
1692 struct stat sb;
1693 size_t size;
1694 int fd;
1695
1696 size = sizeof(*rep);
1697 rep = malloc(size);
1698
1699 if (stat((char *)pdu->path, &sb) < 0) {
1700fail:
1701 rep->error = errno;
1702 sendfile_reply(client->fd, rep, size, tag);
1703 free(rep);
1704 return 1;
1705 }
1706
1707 size += sb.st_size;
1708 rep = realloc(rep, size);
1709 rep->size = cpu_to_le32((uint32_t) sb.st_size);
1710
1711 fd = open((char *)pdu->path, O_RDONLY);
1712 if (fd == -1 )
1713 goto fail;
1714
1715 rep->error = read_data(fd, &rep->data, sb.st_size);
1716 sendfile_reply(client->fd, rep, size, tag);
1717 free(rep);
1718 close(fd);
1719 return 0;
1720}
1721
1722int fio_handle_client(struct fio_client *client)
1723{
1724 struct client_ops *ops = client->ops;
1725 struct fio_net_cmd *cmd;
1726 int size;
1727
1728 dprint(FD_NET, "client: handle %s\n", client->hostname);
1729
1730 cmd = fio_net_recv_cmd(client->fd, true);
1731 if (!cmd)
1732 return 0;
1733
1734 dprint(FD_NET, "client: got cmd op %s from %s (pdu=%u)\n",
1735 fio_server_op(cmd->opcode), client->hostname, cmd->pdu_len);
1736
1737 client->last_cmd = cmd->opcode;
1738
1739 switch (cmd->opcode) {
1740 case FIO_NET_CMD_QUIT:
1741 if (ops->quit)
1742 ops->quit(client, cmd);
1743 remove_client(client);
1744 break;
1745 case FIO_NET_CMD_TEXT:
1746 convert_text(cmd);
1747 ops->text(client, cmd);
1748 break;
1749 case FIO_NET_CMD_DU: {
1750 struct cmd_du_pdu *du = (struct cmd_du_pdu *) cmd->payload;
1751
1752 convert_dus(&du->dus);
1753 convert_agg(&du->agg);
1754
1755 ops->disk_util(client, cmd);
1756 break;
1757 }
1758 case FIO_NET_CMD_TS: {
1759 struct cmd_ts_pdu *p = (struct cmd_ts_pdu *) cmd->payload;
1760
1761 dprint(FD_NET, "client: ts->ss_state = %u\n", (unsigned int) le32_to_cpu(p->ts.ss_state));
1762 if (le32_to_cpu(p->ts.ss_state) & FIO_SS_DATA) {
1763 dprint(FD_NET, "client: received steadystate ring buffers\n");
1764
1765 size = le64_to_cpu(p->ts.ss_dur);
1766 p->ts.ss_iops_data = (uint64_t *) ((struct cmd_ts_pdu *)cmd->payload + 1);
1767 p->ts.ss_bw_data = p->ts.ss_iops_data + size;
1768 }
1769
1770 convert_ts(&p->ts, &p->ts);
1771 convert_gs(&p->rs, &p->rs);
1772
1773 ops->thread_status(client, cmd);
1774 break;
1775 }
1776 case FIO_NET_CMD_GS: {
1777 struct group_run_stats *gs = (struct group_run_stats *) cmd->payload;
1778
1779 convert_gs(gs, gs);
1780
1781 ops->group_stats(client, cmd);
1782 break;
1783 }
1784 case FIO_NET_CMD_ETA: {
1785 struct jobs_eta *je = (struct jobs_eta *) cmd->payload;
1786
1787 if (!remove_reply_cmd(client, cmd))
1788 break;
1789 convert_jobs_eta(je);
1790 handle_eta(client, cmd);
1791 break;
1792 }
1793 case FIO_NET_CMD_PROBE:
1794 remove_reply_cmd(client, cmd);
1795 ops->probe(client, cmd);
1796 break;
1797 case FIO_NET_CMD_SERVER_START:
1798 client->state = Client_running;
1799 if (ops->job_start)
1800 ops->job_start(client, cmd);
1801 break;
1802 case FIO_NET_CMD_START: {
1803 struct cmd_start_pdu *pdu = (struct cmd_start_pdu *) cmd->payload;
1804
1805 pdu->jobs = le32_to_cpu(pdu->jobs);
1806 ops->start(client, cmd);
1807 break;
1808 }
1809 case FIO_NET_CMD_STOP: {
1810 struct cmd_end_pdu *pdu = (struct cmd_end_pdu *) cmd->payload;
1811
1812 convert_stop(cmd);
1813 client->state = Client_stopped;
1814 client->error = le32_to_cpu(pdu->error);
1815 client->signal = le32_to_cpu(pdu->signal);
1816 ops->stop(client);
1817 break;
1818 }
1819 case FIO_NET_CMD_ADD_JOB: {
1820 struct cmd_add_job_pdu *pdu = (struct cmd_add_job_pdu *) cmd->payload;
1821
1822 client->thread_number = le32_to_cpu(pdu->thread_number);
1823 client->groupid = le32_to_cpu(pdu->groupid);
1824
1825 if (ops->add_job)
1826 ops->add_job(client, cmd);
1827 break;
1828 }
1829 case FIO_NET_CMD_IOLOG:
1830 fio_client_handle_iolog(client, cmd);
1831 break;
1832 case FIO_NET_CMD_UPDATE_JOB:
1833 ops->update_job(client, cmd);
1834 remove_reply_cmd(client, cmd);
1835 break;
1836 case FIO_NET_CMD_VTRIGGER: {
1837 struct all_io_list *pdu = (struct all_io_list *) cmd->payload;
1838 char buf[128];
1839 int off = 0;
1840
1841 if (aux_path) {
1842 strcpy(buf, aux_path);
1843 off = strlen(buf);
1844 }
1845
1846 __verify_save_state(pdu, server_name(client, &buf[off], sizeof(buf) - off));
1847 exec_trigger(trigger_cmd);
1848 break;
1849 }
1850 case FIO_NET_CMD_SENDFILE: {
1851 struct cmd_sendfile *pdu = (struct cmd_sendfile *) cmd->payload;
1852 fio_send_file(client, pdu, cmd->tag);
1853 break;
1854 }
1855 case FIO_NET_CMD_JOB_OPT: {
1856 handle_job_opt(client, cmd);
1857 break;
1858 }
1859 default:
1860 log_err("fio: unknown client op: %s\n", fio_server_op(cmd->opcode));
1861 break;
1862 }
1863
1864 free(cmd);
1865 return 1;
1866}
1867
1868int fio_clients_send_trigger(const char *cmd)
1869{
1870 struct flist_head *entry;
1871 struct fio_client *client;
1872 size_t slen;
1873
1874 dprint(FD_NET, "client: send vtrigger: %s\n", cmd);
1875
1876 if (!cmd)
1877 slen = 0;
1878 else
1879 slen = strlen(cmd);
1880
1881 flist_for_each(entry, &client_list) {
1882 struct cmd_vtrigger_pdu *pdu;
1883
1884 client = flist_entry(entry, struct fio_client, list);
1885
1886 pdu = malloc(sizeof(*pdu) + slen);
1887 pdu->len = cpu_to_le16((uint16_t) slen);
1888 if (slen)
1889 memcpy(pdu->cmd, cmd, slen);
1890 fio_net_send_cmd(client->fd, FIO_NET_CMD_VTRIGGER, pdu,
1891 sizeof(*pdu) + slen, NULL, NULL);
1892 free(pdu);
1893 }
1894
1895 return 0;
1896}
1897
1898static void request_client_etas(struct client_ops *ops)
1899{
1900 struct fio_client *client;
1901 struct flist_head *entry;
1902 struct client_eta *eta;
1903 int skipped = 0;
1904
1905 if (eta_print == FIO_ETA_NEVER)
1906 return;
1907
1908 dprint(FD_NET, "client: request eta (%d)\n", nr_clients);
1909
1910 eta = calloc(1, sizeof(*eta) + __THREAD_RUNSTR_SZ(REAL_MAX_JOBS));
1911 eta->pending = nr_clients;
1912
1913 flist_for_each(entry, &client_list) {
1914 client = flist_entry(entry, struct fio_client, list);
1915
1916 if (!flist_empty(&client->eta_list)) {
1917 skipped++;
1918 continue;
1919 }
1920 if (client->state != Client_running)
1921 continue;
1922
1923 assert(!client->eta_in_flight);
1924 flist_add_tail(&client->eta_list, &eta_list);
1925 client->eta_in_flight = eta;
1926 fio_net_send_simple_cmd(client->fd, FIO_NET_CMD_SEND_ETA,
1927 (uintptr_t) eta, &client->cmd_list);
1928 }
1929
1930 while (skipped--) {
1931 if (!fio_client_dec_jobs_eta(eta, ops->eta))
1932 break;
1933 }
1934
1935 dprint(FD_NET, "client: requested eta tag %p\n", eta);
1936}
1937
1938/*
1939 * A single SEND_ETA timeout isn't fatal. Attempt to recover.
1940 */
1941static int handle_cmd_timeout(struct fio_client *client,
1942 struct fio_net_cmd_reply *reply)
1943{
1944 uint16_t reply_opcode = reply->opcode;
1945
1946 flist_del(&reply->list);
1947 free(reply);
1948
1949 if (reply_opcode != FIO_NET_CMD_SEND_ETA)
1950 return 1;
1951
1952 log_info("client <%s>: timeout on SEND_ETA\n", client->hostname);
1953
1954 flist_del_init(&client->eta_list);
1955 if (client->eta_in_flight) {
1956 fio_client_dec_jobs_eta(client->eta_in_flight, client->ops->eta);
1957 client->eta_in_flight = NULL;
1958 }
1959
1960 /*
1961 * If we fail 5 in a row, give up...
1962 */
1963 if (client->eta_timeouts++ > 5)
1964 return 1;
1965
1966 return 0;
1967}
1968
1969static int client_check_cmd_timeout(struct fio_client *client,
1970 struct timespec *now)
1971{
1972 struct fio_net_cmd_reply *reply;
1973 struct flist_head *entry, *tmp;
1974 int ret = 0;
1975
1976 flist_for_each_safe(entry, tmp, &client->cmd_list) {
1977 unsigned int op;
1978
1979 reply = flist_entry(entry, struct fio_net_cmd_reply, list);
1980
1981 if (mtime_since(&reply->ts, now) < FIO_NET_CLIENT_TIMEOUT)
1982 continue;
1983
1984 op = reply->opcode;
1985 if (!handle_cmd_timeout(client, reply))
1986 continue;
1987
1988 log_err("fio: client %s, timeout on cmd %s\n", client->hostname,
1989 fio_server_op(op));
1990 ret = 1;
1991 }
1992
1993 return flist_empty(&client->cmd_list) && ret;
1994}
1995
1996static int fio_check_clients_timed_out(void)
1997{
1998 struct fio_client *client;
1999 struct flist_head *entry, *tmp;
2000 struct timespec ts;
2001 int ret = 0;
2002
2003 fio_gettime(&ts, NULL);
2004
2005 flist_for_each_safe(entry, tmp, &client_list) {
2006 client = flist_entry(entry, struct fio_client, list);
2007
2008 if (flist_empty(&client->cmd_list))
2009 continue;
2010
2011 if (!client_check_cmd_timeout(client, &ts))
2012 continue;
2013
2014 if (client->ops->timed_out)
2015 client->ops->timed_out(client);
2016 else
2017 log_err("fio: client %s timed out\n", client->hostname);
2018
2019 if (client->last_cmd != FIO_NET_CMD_VTRIGGER)
2020 client->error = ETIMEDOUT;
2021 else
2022 log_info("fio: ignoring timeout due to vtrigger\n");
2023 remove_client(client);
2024 ret = 1;
2025 }
2026
2027 return ret;
2028}
2029
2030int fio_handle_clients(struct client_ops *ops)
2031{
2032 struct pollfd *pfds;
2033 int i, ret = 0, retval = 0;
2034
2035 fio_gettime(&eta_ts, NULL);
2036
2037 pfds = malloc(nr_clients * sizeof(struct pollfd));
2038
2039 init_thread_stat(&client_ts);
2040 init_group_run_stat(&client_gs);
2041
2042 while (!exit_backend && nr_clients) {
2043 struct flist_head *entry, *tmp;
2044 struct fio_client *client;
2045
2046 i = 0;
2047 flist_for_each_safe(entry, tmp, &client_list) {
2048 client = flist_entry(entry, struct fio_client, list);
2049
2050 if (!client->sent_job && !client->ops->stay_connected &&
2051 flist_empty(&client->cmd_list)) {
2052 remove_client(client);
2053 continue;
2054 }
2055
2056 pfds[i].fd = client->fd;
2057 pfds[i].events = POLLIN;
2058 i++;
2059 }
2060
2061 if (!nr_clients)
2062 break;
2063
2064 assert(i == nr_clients);
2065
2066 do {
2067 struct timespec ts;
2068 int timeout;
2069
2070 fio_gettime(&ts, NULL);
2071 if (eta_time_within_slack(mtime_since(&eta_ts, &ts))) {
2072 request_client_etas(ops);
2073 memcpy(&eta_ts, &ts, sizeof(ts));
2074
2075 if (fio_check_clients_timed_out())
2076 break;
2077 }
2078
2079 check_trigger_file();
2080
2081 timeout = min(100u, ops->eta_msec);
2082
2083 ret = poll(pfds, nr_clients, timeout);
2084 if (ret < 0) {
2085 if (errno == EINTR)
2086 continue;
2087 log_err("fio: poll clients: %s\n", strerror(errno));
2088 break;
2089 } else if (!ret)
2090 continue;
2091 } while (ret <= 0);
2092
2093 for (i = 0; i < nr_clients; i++) {
2094 if (!(pfds[i].revents & POLLIN))
2095 continue;
2096
2097 client = find_client_by_fd(pfds[i].fd);
2098 if (!client) {
2099 log_err("fio: unknown client fd %ld\n", (long) pfds[i].fd);
2100 continue;
2101 }
2102 if (!fio_handle_client(client)) {
2103 log_info("client: host=%s disconnected\n",
2104 client->hostname);
2105 remove_client(client);
2106 retval = 1;
2107 } else if (client->error)
2108 retval = 1;
2109 fio_put_client(client);
2110 }
2111 }
2112
2113 fio_client_json_fini();
2114
2115 free(pfds);
2116 return retval || error_clients;
2117}
2118
2119static void client_display_thread_status(struct jobs_eta *je)
2120{
2121 if (!(output_format & FIO_OUTPUT_JSON))
2122 display_thread_status(je);
2123}
fio - Flexible IO Tester