[PATCH] Pretty up output on large SMP machines
[blktrace.git] / blktrace.c
CommitLineData
d0ca268b
JA
1/*
2 * block queue tracing application
3 *
d956a2cd
JA
4 * Copyright (C) 2005 Jens Axboe <axboe@suse.de>
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19 *
d0ca268b
JA
20 */
21#include <pthread.h>
22#include <sys/types.h>
23#include <sys/stat.h>
24#include <unistd.h>
25#include <locale.h>
26#include <signal.h>
27#include <fcntl.h>
28#include <string.h>
29#include <sys/ioctl.h>
b9d4294e 30#include <sys/param.h>
d0ca268b
JA
31#include <stdio.h>
32#include <stdlib.h>
33#include <sched.h>
d39c04ca
AB
34#include <ctype.h>
35#include <getopt.h>
d0ca268b
JA
36
37#include "blktrace.h"
38
39#define BUF_SIZE (128 *1024)
40#define BUF_NR (4)
41
d39c04ca
AB
42#define DECLARE_MASK_MAP(mask) { BLK_TC_##mask, #mask, "BLK_TC_"#mask }
43#define COMPARE_MASK_MAP(mmp, str) \
75da3c6a
NS
44 (!strcasecmp((mmp)->short_form, (str)) || \
45 !strcasecmp((mmp)->long_form, (str)))
d39c04ca
AB
46
47#define VALID_SET(x) ((1 <= (x)) && ((x) < (1 << BLK_TC_SHIFT)))
48
49struct mask_map {
50 int mask;
51 char *short_form;
52 char *long_form;
53};
54
55struct mask_map mask_maps[] = {
5c86134e
JA
56 DECLARE_MASK_MAP(READ),
57 DECLARE_MASK_MAP(WRITE),
58 DECLARE_MASK_MAP(BARRIER),
59 DECLARE_MASK_MAP(SYNC),
60 DECLARE_MASK_MAP(QUEUE),
61 DECLARE_MASK_MAP(REQUEUE),
62 DECLARE_MASK_MAP(ISSUE),
63 DECLARE_MASK_MAP(COMPLETE),
64 DECLARE_MASK_MAP(FS),
65 DECLARE_MASK_MAP(PC),
d39c04ca
AB
66};
67
bc39777c 68#define S_OPTS "d:a:A:r:o:k"
d5396421 69static struct option l_opts[] = {
5c86134e 70 {
d39c04ca
AB
71 .name = "dev",
72 .has_arg = 1,
73 .flag = NULL,
74 .val = 'd'
75 },
5c86134e 76 {
d39c04ca
AB
77 .name = "act-mask",
78 .has_arg = 1,
79 .flag = NULL,
80 .val = 'a'
81 },
5c86134e 82 {
d39c04ca
AB
83 .name = "set-mask",
84 .has_arg = 1,
85 .flag = NULL,
86 .val = 'A'
87 },
5c86134e 88 {
5270dddd
JA
89 .name = "relay",
90 .has_arg = 1,
91 .flag = NULL,
92 .val = 'r'
93 },
d5396421
JA
94 {
95 .name = "output",
96 .has_arg = 1,
97 .flag = NULL,
98 .val = 'o'
99 },
bc39777c
JA
100 {
101 .name = "kill",
102 .has_arg = 0,
103 .flag = NULL,
104 .val = 'k'
105 },
d39c04ca
AB
106 {
107 .name = NULL,
108 .has_arg = 0,
109 .flag = NULL,
110 .val = 0
111 }
112};
113
d0ca268b
JA
114struct thread_information {
115 int cpu;
116 pthread_t thread;
b9d4294e
JA
117
118 int fd;
119 char fn[MAXPATHLEN + 64];
120
d5396421
JA
121 pthread_mutex_t *fd_lock;
122 int ofd;
123
d0ca268b
JA
124 unsigned long events_processed;
125};
126
5270dddd 127static char *relay_path;
d0ca268b
JA
128
129#define is_done() (*(volatile int *)(&done))
130static volatile int done;
131
132static int devfd, ncpus;
133static struct thread_information *thread_information;
134static char *buts_name_p;
d39c04ca 135static char *dev;
d5396421 136static char *output_name;
5c86134e 137static int act_mask = ~0U;
830fd65c 138static int trace_started;
bc39777c 139static int kill_running_trace;
d39c04ca 140
d5396421
JA
141static pthread_mutex_t stdout_mutex = PTHREAD_MUTEX_INITIALIZER;
142
72ca8801
NS
143static void exit_trace(int status);
144
1f79c4a0 145static int find_mask_map(char *string)
d39c04ca 146{
5c86134e
JA
147 int i;
148
149 for (i = 0; i < sizeof(mask_maps)/sizeof(mask_maps[0]); i++)
75da3c6a 150 if (COMPARE_MASK_MAP(&mask_maps[i], string))
5c86134e 151 return mask_maps[i].mask;
d39c04ca 152
d39c04ca
AB
153 return -1;
154}
d0ca268b 155
3aabcd89 156static int start_trace(char *dev)
d0ca268b
JA
157{
158 struct blk_user_trace_setup buts;
159
1f79c4a0 160 memset(&buts, 0, sizeof(buts));
d0ca268b
JA
161 buts.buf_size = BUF_SIZE;
162 buts.buf_nr = BUF_NR;
d39c04ca 163 buts.act_mask = act_mask;
d0ca268b 164
d0ca268b
JA
165 if (ioctl(devfd, BLKSTARTTRACE, &buts) < 0) {
166 perror("BLKSTARTTRACE");
167 return 1;
168 }
169
830fd65c 170 trace_started = 1;
d0ca268b
JA
171 buts_name_p = strdup(buts.name);
172 return 0;
173}
174
3aabcd89 175static void stop_trace(void)
d0ca268b 176{
bc39777c 177 if (trace_started || kill_running_trace) {
707b0914
JA
178 if (ioctl(devfd, BLKSTOPTRACE) < 0)
179 perror("BLKSTOPTRACE");
d0ca268b 180
707b0914
JA
181 trace_started = 0;
182 }
d0ca268b
JA
183}
184
d5396421 185static void extract_data(struct thread_information *tip, char *ofn, int nb)
d0ca268b
JA
186{
187 int ret, bytes_left;
87b72777 188 unsigned char *buf, *p;
d0ca268b 189
87b72777 190 buf = malloc(nb);
d0ca268b
JA
191 p = buf;
192 bytes_left = nb;
193 while (bytes_left > 0) {
b9d4294e 194 ret = read(tip->fd, p, bytes_left);
3aabcd89
JA
195 if (!ret)
196 usleep(1000);
197 else if (ret < 0) {
b9d4294e 198 perror(tip->fn);
d0ca268b 199 fprintf(stderr, "Thread %d extract_data %s failed\n",
b9d4294e 200 tip->cpu, tip->fn);
87b72777 201 free(buf);
76718bcd 202 exit_trace(1);
3aabcd89 203 } else {
d0ca268b
JA
204 p += ret;
205 bytes_left -= ret;
206 }
207 }
208
d5396421 209 ret = write(tip->ofd, buf, nb);
d0ca268b
JA
210 if (ret != nb) {
211 perror(ofn);
b9d4294e 212 fprintf(stderr,"Thread %d extract_data %s failed\n", tip->cpu, ofn);
87b72777 213 free(buf);
76718bcd 214 exit_trace(1);
d0ca268b 215 }
87b72777
JA
216
217 free(buf);
d0ca268b
JA
218}
219
d5396421
JA
220static inline void tip_fd_unlock(struct thread_information *tip)
221{
222 if (tip->fd_lock)
223 pthread_mutex_unlock(tip->fd_lock);
224}
225
226static inline void tip_fd_lock(struct thread_information *tip)
227{
228 if (tip->fd_lock)
229 pthread_mutex_lock(tip->fd_lock);
230}
231
3aabcd89 232static void *extract(void *arg)
d0ca268b
JA
233{
234 struct thread_information *tip = arg;
d5396421
JA
235 int ret, pdu_len;
236 char dp[64];
d0ca268b
JA
237 struct blk_io_trace t;
238 pid_t pid = getpid();
239 cpu_set_t cpu_mask;
240
241 CPU_ZERO(&cpu_mask);
b9d4294e 242 CPU_SET((tip->cpu), &cpu_mask);
d0ca268b
JA
243
244 if (sched_setaffinity(pid, sizeof(cpu_mask), &cpu_mask) == -1) {
245 perror("sched_setaffinity");
76718bcd 246 exit_trace(1);
d0ca268b
JA
247 }
248
b9d4294e
JA
249 snprintf(tip->fn, sizeof(tip->fn),
250 "%s/block/%s/trace%d", relay_path, buts_name_p, tip->cpu);
251 tip->fd = open(tip->fn, O_RDONLY);
252 if (tip->fd < 0) {
253 perror(tip->fn);
5c86134e
JA
254 fprintf(stderr,"Thread %d failed open of %s\n", tip->cpu,
255 tip->fn);
76718bcd 256 exit_trace(1);
d0ca268b
JA
257 }
258
259 while (!is_done()) {
b9d4294e 260 ret = read(tip->fd, &t, sizeof(t));
d0ca268b
JA
261 if (ret != sizeof(t)) {
262 if (ret < 0) {
b9d4294e 263 perror(tip->fn);
d0ca268b 264 fprintf(stderr,"Thread %d failed read of %s\n",
b9d4294e 265 tip->cpu, tip->fn);
76718bcd 266 exit_trace(1);
d0ca268b 267 } else if (ret > 0) {
8fc0abbc 268 fprintf(stderr,"Thread %d misread %s %d,%d\n",
b9d4294e 269 tip->cpu, tip->fn, ret, (int)sizeof(t));
76718bcd 270 exit_trace(1);
d0ca268b
JA
271 } else {
272 usleep(10000);
273 continue;
274 }
275 }
276
277 if (verify_trace(&t))
76718bcd 278 exit_trace(1);
d0ca268b 279
18ada3d4
JA
280 pdu_len = t.pdu_len;
281
6fe4709e
JA
282 trace_to_be(&t);
283
d5396421
JA
284 tip_fd_lock(tip);
285
286 ret = write(tip->ofd, &t, sizeof(t));
d0ca268b 287 if (ret < 0) {
d5396421
JA
288 fprintf(stderr,"Thread %d failed write\n", tip->cpu);
289 tip_fd_unlock(tip);
76718bcd 290 exit_trace(1);
d0ca268b
JA
291 }
292
18ada3d4 293 if (pdu_len)
d5396421
JA
294 extract_data(tip, dp, pdu_len);
295
296 tip_fd_unlock(tip);
87b72777 297
d0ca268b
JA
298 tip->events_processed++;
299 }
300
301 return NULL;
302}
303
3aabcd89 304static int start_threads(void)
d0ca268b
JA
305{
306 struct thread_information *tip;
d5396421 307 char op[64];
d0ca268b
JA
308 int i;
309
310 ncpus = sysconf(_SC_NPROCESSORS_ONLN);
311 if (ncpus < 0) {
312 fprintf(stderr, "sysconf(_SC_NPROCESSORS_ONLN) failed\n");
1f79c4a0 313 return 0;
d0ca268b 314 }
d0ca268b
JA
315
316 thread_information = malloc(ncpus * sizeof(struct thread_information));
317 for (i = 0, tip = thread_information; i < ncpus; i++, tip++) {
d5396421 318 tip->fd_lock = NULL;
d0ca268b
JA
319 tip->cpu = i;
320 tip->events_processed = 0;
321
d5396421 322 if (!strcmp(output_name, "-")) {
1f79c4a0 323 tip->ofd = dup(STDOUT_FILENO);
d5396421
JA
324 tip->fd_lock = &stdout_mutex;
325 } else {
326 sprintf(op, "%s_out.%d", output_name, tip->cpu);
327 tip->ofd = open(op, O_CREAT|O_TRUNC|O_WRONLY, 0644);
328 }
329
330 if (tip->ofd < 0) {
331 perror(op);
1f79c4a0 332 return 0;
d5396421
JA
333 }
334
d0ca268b
JA
335 if (pthread_create(&tip->thread, NULL, extract, tip)) {
336 perror( "pthread_create");
337 return 0;
338 }
339 }
340
341 return ncpus;
342}
343
72ca8801
NS
344static void close_thread(struct thread_information *tip)
345{
346 if (tip->fd != -1)
347 close(tip->fd);
348 if (tip->ofd != -1)
349 close(tip->ofd);
350 tip->fd = tip->ofd = -1;
351}
352
3aabcd89
JA
353static void stop_threads(void)
354{
355 struct thread_information *tip = thread_information;
356 int i;
357
358 for (i = 0; i < ncpus; i++, tip++) {
359 int ret;
360
361 if (pthread_join(tip->thread, (void *) &ret))
362 perror("thread_join");
72ca8801 363 close_thread(tip);
3aabcd89
JA
364 }
365}
366
72ca8801
NS
367static void stop_tracing(void)
368{
369 struct thread_information *tip = thread_information;
370 int i;
371
372 for (i = 0; i < ncpus; i++, tip++)
373 close_thread(tip);
374 stop_trace();
375}
376
377static void exit_trace(int status)
378{
379 stop_tracing();
380 exit(status);
381}
382
1f79c4a0 383static void show_stats(void)
d0ca268b
JA
384{
385 int i;
386 struct thread_information *tip;
387 unsigned long events_processed = 0;
388
d5396421
JA
389 if (!strcmp(output_name, "-"))
390 return;
391
d0ca268b
JA
392 for (i = 0, tip = thread_information; i < ncpus; i++, tip++) {
393 printf("CPU%3d: %20ld events\n",
394 tip->cpu, tip->events_processed);
395 events_processed += tip->events_processed;
396 }
397
398 printf("Total: %20ld events\n", events_processed);
399}
400
1f79c4a0 401static void handle_sigint(int sig)
d0ca268b 402{
d0ca268b
JA
403 done = 1;
404}
405
406int main(int argc, char *argv[])
407{
5270dddd 408 static char default_relay_path[] = "/relay";
d0ca268b 409 struct stat st;
d39c04ca
AB
410 int i, c;
411 int act_mask_tmp = 0;
412
413 while ((c = getopt_long(argc, argv, S_OPTS, l_opts, NULL)) >= 0) {
414 switch (c) {
415 case 'a':
416 i = find_mask_map(optarg);
417 if (i < 0) {
418 fprintf(stderr,"Invalid action mask %s\n",
419 optarg);
420 return 4;
421 }
422 act_mask_tmp |= i;
423 break;
424
425 case 'A':
426 if ((sscanf(optarg, "%x", &i) != 1) || !VALID_SET(i)) {
427 fprintf(stderr,
428 "Invalid set action mask %s/0x%x\n",
429 optarg, i);
430 return 4;
431 }
432 act_mask_tmp = i;
433 break;
d0ca268b 434
d39c04ca
AB
435 case 'd':
436 dev = strdup(optarg);
437 break;
438
5270dddd
JA
439 case 'r':
440 relay_path = optarg;
441 break;
442
d5396421
JA
443 case 'o':
444 output_name = strdup(optarg);
445 break;
bc39777c
JA
446 case 'k':
447 kill_running_trace = 1;
448 break;
d5396421 449
d39c04ca
AB
450 default:
451 fprintf(stderr,"Usage: %s -d <dev> "
452 "[-a <trace> [-a <trace>]]\n", argv[0]);
453 return 4;
454 }
455 }
456
457 if ((dev == NULL) || (optind < argc)) {
458 fprintf(stderr,"Usage: %s -d <dev> "
459 "[-a <trace> [-a <trace>]]\n", argv[0]);
460 return 4;
461 }
462
5270dddd
JA
463 if (!relay_path)
464 relay_path = default_relay_path;
465
d5396421 466 if (act_mask_tmp != 0)
d39c04ca 467 act_mask = act_mask_tmp;
d0ca268b
JA
468
469 if (stat(relay_path, &st) < 0) {
d39c04ca 470 fprintf(stderr,"%s does not appear to be mounted\n",
d0ca268b
JA
471 relay_path);
472 return 2;
473 }
474
bc39777c
JA
475 devfd = open(dev, O_RDONLY);
476 if (devfd < 0) {
477 perror(dev);
478 return 3;
479 }
480
481 if (kill_running_trace) {
482 stop_trace();
483 exit(0);
484 }
485
d39c04ca 486 if (start_trace(dev)) {
5270dddd 487 close(devfd);
d39c04ca 488 fprintf(stderr, "Failed to start trace on %s\n", dev);
bc39777c 489 return 4;
d0ca268b
JA
490 }
491
492 setlocale(LC_NUMERIC, "en_US");
493
d5396421
JA
494 if (!output_name)
495 output_name = strdup(buts_name_p);
496
d0ca268b
JA
497 i = start_threads();
498 if (!i) {
499 fprintf(stderr, "Failed to start worker threads\n");
500 stop_trace();
bc39777c 501 return 5;
d0ca268b
JA
502 }
503
d0ca268b
JA
504 signal(SIGINT, handle_sigint);
505 signal(SIGHUP, handle_sigint);
506 signal(SIGTERM, handle_sigint);
507
72ca8801 508 atexit(stop_tracing);
830fd65c 509
d0ca268b
JA
510 while (!is_done())
511 sleep(1);
512
b9d4294e 513 stop_threads();
3a9995b9 514 stop_trace();
d0ca268b 515 show_stats();
bc39777c 516 close(devfd);
d0ca268b
JA
517
518 return 0;
519}
520