~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/tools/perf/bench/sched-messaging.c

Version: ~ [ linux-6.11.5 ] ~ [ linux-6.10.14 ] ~ [ linux-6.9.12 ] ~ [ linux-6.8.12 ] ~ [ linux-6.7.12 ] ~ [ linux-6.6.58 ] ~ [ linux-6.5.13 ] ~ [ linux-6.4.16 ] ~ [ linux-6.3.13 ] ~ [ linux-6.2.16 ] ~ [ linux-6.1.114 ] ~ [ linux-6.0.19 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.169 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.228 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.284 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.322 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.336 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.337 ] ~ [ linux-4.4.302 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0
  2 /*
  3  *
  4  * sched-messaging.c
  5  *
  6  * messaging: Benchmark for scheduler and IPC mechanisms
  7  *
  8  * Based on hackbench by Rusty Russell <rusty@rustcorp.com.au>
  9  * Ported to perf by Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
 10  *
 11  */
 12 
 13 #include <subcmd/parse-options.h>
 14 #include "bench.h"
 15 
 16 /* Test groups of 20 processes spraying to 20 receivers */
 17 #include <pthread.h>
 18 #include <stdio.h>
 19 #include <stdlib.h>
 20 #include <string.h>
 21 #include <errno.h>
 22 #include <unistd.h>
 23 #include <sys/types.h>
 24 #include <sys/socket.h>
 25 #include <sys/wait.h>
 26 #include <sys/time.h>
 27 #include <poll.h>
 28 #include <limits.h>
 29 #include <err.h>
 30 #include <linux/list.h>
 31 #include <linux/time64.h>
 32 
 33 #define DATASIZE 100
 34 
 35 static bool use_pipes = false;
 36 static unsigned int nr_loops = 100;
 37 static bool thread_mode = false;
 38 static unsigned int num_groups = 10;
 39 static unsigned int total_children = 0;
 40 static struct list_head sender_contexts = LIST_HEAD_INIT(sender_contexts);
 41 static struct list_head receiver_contexts = LIST_HEAD_INIT(receiver_contexts);
 42 
 43 struct sender_context {
 44         struct list_head list;
 45         unsigned int num_fds;
 46         int ready_out;
 47         int wakefd;
 48         int out_fds[];
 49 };
 50 
 51 struct receiver_context {
 52         struct list_head list;
 53         unsigned int num_packets;
 54         int in_fds[2];
 55         int ready_out;
 56         int wakefd;
 57 };
 58 
 59 union messaging_worker {
 60         pthread_t thread;
 61         pid_t pid;
 62 };
 63 
 64 static union messaging_worker *worker_tab;
 65 
 66 static void fdpair(int fds[2])
 67 {
 68         if (use_pipes) {
 69                 if (pipe(fds) == 0)
 70                         return;
 71         } else {
 72                 if (socketpair(AF_UNIX, SOCK_STREAM, 0, fds) == 0)
 73                         return;
 74         }
 75 
 76         err(EXIT_FAILURE, use_pipes ? "pipe()" : "socketpair()");
 77 }
 78 
 79 /* Block until we're ready to go */
 80 static void ready(int ready_out, int wakefd)
 81 {
 82         struct pollfd pollfd = { .fd = wakefd, .events = POLLIN };
 83 
 84         /* Tell them we're ready. */
 85         if (write(ready_out, "R", 1) != 1)
 86                 err(EXIT_FAILURE, "CLIENT: ready write");
 87 
 88         /* Wait for "GO" signal */
 89         if (poll(&pollfd, 1, -1) != 1)
 90                 err(EXIT_FAILURE, "poll");
 91 }
 92 
 93 /* Sender sprays nr_loops messages down each file descriptor */
 94 static void *sender(struct sender_context *ctx)
 95 {
 96         char data[DATASIZE];
 97         unsigned int i, j;
 98 
 99         ready(ctx->ready_out, ctx->wakefd);
100         memset(data, 'S', sizeof(data));
101 
102         /* Now pump to every receiver. */
103         for (i = 0; i < nr_loops; i++) {
104                 for (j = 0; j < ctx->num_fds; j++) {
105                         int ret, done = 0;
106 
107 again:
108                         ret = write(ctx->out_fds[j], data + done,
109                                     sizeof(data) - done);
110                         if (ret < 0)
111                                 err(EXIT_FAILURE, "SENDER: write");
112                         done += ret;
113                         if (done < DATASIZE)
114                                 goto again;
115                 }
116         }
117 
118         return NULL;
119 }
120 
121 
122 /* One receiver per fd */
123 static void *receiver(struct receiver_context* ctx)
124 {
125         unsigned int i;
126 
127         if (!thread_mode)
128                 close(ctx->in_fds[1]);
129 
130         /* Wait for start... */
131         ready(ctx->ready_out, ctx->wakefd);
132 
133         /* Receive them all */
134         for (i = 0; i < ctx->num_packets; i++) {
135                 char data[DATASIZE];
136                 int ret, done = 0;
137 
138 again:
139                 ret = read(ctx->in_fds[0], data + done, DATASIZE - done);
140                 if (ret < 0)
141                         err(EXIT_FAILURE, "SERVER: read");
142                 done += ret;
143                 if (done < DATASIZE)
144                         goto again;
145         }
146 
147         return NULL;
148 }
149 
150 static void create_thread_worker(union messaging_worker *worker,
151                                  void *ctx, void *(*func)(void *))
152 {
153         pthread_attr_t attr;
154         int ret;
155 
156         if (pthread_attr_init(&attr) != 0)
157                 err(EXIT_FAILURE, "pthread_attr_init:");
158 
159 #ifndef __ia64__
160         if (pthread_attr_setstacksize(&attr, PTHREAD_STACK_MIN) != 0)
161                 err(EXIT_FAILURE, "pthread_attr_setstacksize");
162 #endif
163 
164         ret = pthread_create(&worker->thread, &attr, func, ctx);
165         if (ret != 0)
166                 err(EXIT_FAILURE, "pthread_create failed");
167 
168         pthread_attr_destroy(&attr);
169 }
170 
171 static void create_process_worker(union messaging_worker *worker,
172                                   void *ctx, void *(*func)(void *))
173 {
174         /* Fork the receiver. */
175         worker->pid = fork();
176 
177         if (worker->pid == -1) {
178                 err(EXIT_FAILURE, "fork()");
179         } else if (worker->pid == 0) {
180                 (*func) (ctx);
181                 exit(0);
182         }
183 }
184 
185 static void create_worker(union messaging_worker *worker,
186                           void *ctx, void *(*func)(void *))
187 {
188         if (!thread_mode)
189                 return create_process_worker(worker, ctx, func);
190         else
191                 return create_thread_worker(worker, ctx, func);
192 }
193 
194 static void reap_worker(union messaging_worker *worker)
195 {
196         int proc_status;
197         void *thread_status;
198 
199         if (!thread_mode) {
200                 /* process mode */
201                 wait(&proc_status);
202                 if (!WIFEXITED(proc_status))
203                         exit(1);
204         } else {
205                 pthread_join(worker->thread, &thread_status);
206         }
207 }
208 
209 /* One group of senders and receivers */
210 static unsigned int group(union messaging_worker *worker,
211                 unsigned int num_fds,
212                 int ready_out,
213                 int wakefd)
214 {
215         unsigned int i;
216         struct sender_context *snd_ctx = malloc(sizeof(struct sender_context) +
217                                                 num_fds * sizeof(int));
218 
219         if (!snd_ctx)
220                 err(EXIT_FAILURE, "malloc()");
221 
222         list_add(&snd_ctx->list, &sender_contexts);
223         for (i = 0; i < num_fds; i++) {
224                 int fds[2];
225                 struct receiver_context *ctx = malloc(sizeof(*ctx));
226 
227                 if (!ctx)
228                         err(EXIT_FAILURE, "malloc()");
229 
230                 list_add(&ctx->list, &receiver_contexts);
231 
232                 /* Create the pipe between client and server */
233                 fdpair(fds);
234 
235                 ctx->num_packets = num_fds * nr_loops;
236                 ctx->in_fds[0] = fds[0];
237                 ctx->in_fds[1] = fds[1];
238                 ctx->ready_out = ready_out;
239                 ctx->wakefd = wakefd;
240 
241                 create_worker(worker + i, ctx, (void *)receiver);
242 
243                 snd_ctx->out_fds[i] = fds[1];
244                 if (!thread_mode)
245                         close(fds[0]);
246         }
247 
248         /* Now we have all the fds, fork the senders */
249         for (i = 0; i < num_fds; i++) {
250                 snd_ctx->ready_out = ready_out;
251                 snd_ctx->wakefd = wakefd;
252                 snd_ctx->num_fds = num_fds;
253 
254                 create_worker(worker + num_fds + i, snd_ctx, (void *)sender);
255         }
256 
257         /* Close the fds we have left */
258         if (!thread_mode)
259                 for (i = 0; i < num_fds; i++)
260                         close(snd_ctx->out_fds[i]);
261 
262         /* Return number of children to reap */
263         return num_fds * 2;
264 }
265 
266 static void sig_handler(int sig __maybe_unused)
267 {
268         unsigned int i;
269 
270         /*
271          * When exit abnormally, kill all forked child processes.
272          */
273         for (i = 0; i < total_children; i++)
274                 kill(worker_tab[i].pid, SIGKILL);
275 }
276 
277 static const struct option options[] = {
278         OPT_BOOLEAN('p', "pipe", &use_pipes,
279                     "Use pipe() instead of socketpair()"),
280         OPT_BOOLEAN('t', "thread", &thread_mode,
281                     "Be multi thread instead of multi process"),
282         OPT_UINTEGER('g', "group", &num_groups, "Specify number of groups"),
283         OPT_UINTEGER('l', "nr_loops", &nr_loops, "Specify the number of loops to run (default: 100)"),
284         OPT_END()
285 };
286 
287 static const char * const bench_sched_message_usage[] = {
288         "perf bench sched messaging <options>",
289         NULL
290 };
291 
292 int bench_sched_messaging(int argc, const char **argv)
293 {
294         unsigned int i;
295         struct timeval start, stop, diff;
296         unsigned int num_fds = 20;
297         int readyfds[2], wakefds[2];
298         char dummy;
299         struct sender_context *pos, *n;
300 
301         argc = parse_options(argc, argv, options,
302                              bench_sched_message_usage, 0);
303 
304         worker_tab = malloc(num_fds * 2 * num_groups * sizeof(union messaging_worker));
305         if (!worker_tab)
306                 err(EXIT_FAILURE, "main:malloc()");
307 
308         fdpair(readyfds);
309         fdpair(wakefds);
310 
311         if (!thread_mode) {
312                 signal(SIGINT, sig_handler);
313                 signal(SIGTERM, sig_handler);
314         }
315 
316         for (i = 0; i < num_groups; i++)
317                 total_children += group(worker_tab + total_children, num_fds,
318                                         readyfds[1], wakefds[0]);
319 
320         /* Wait for everyone to be ready */
321         for (i = 0; i < total_children; i++)
322                 if (read(readyfds[0], &dummy, 1) != 1)
323                         err(EXIT_FAILURE, "Reading for readyfds");
324 
325         gettimeofday(&start, NULL);
326 
327         /* Kick them off */
328         if (write(wakefds[1], &dummy, 1) != 1)
329                 err(EXIT_FAILURE, "Writing to start them");
330 
331         /* Reap them all */
332         for (i = 0; i < total_children; i++)
333                 reap_worker(worker_tab + i);
334 
335         gettimeofday(&stop, NULL);
336 
337         timersub(&stop, &start, &diff);
338 
339         switch (bench_format) {
340         case BENCH_FORMAT_DEFAULT:
341                 printf("# %d sender and receiver %s per group\n",
342                        num_fds, thread_mode ? "threads" : "processes");
343                 printf("# %d groups == %d %s run\n\n",
344                        num_groups, num_groups * 2 * num_fds,
345                        thread_mode ? "threads" : "processes");
346                 printf(" %14s: %lu.%03lu [sec]\n", "Total time",
347                        (unsigned long) diff.tv_sec,
348                        (unsigned long) (diff.tv_usec / USEC_PER_MSEC));
349                 break;
350         case BENCH_FORMAT_SIMPLE:
351                 printf("%lu.%03lu\n", (unsigned long) diff.tv_sec,
352                        (unsigned long) (diff.tv_usec / USEC_PER_MSEC));
353                 break;
354         default:
355                 /* reaching here is something disaster */
356                 fprintf(stderr, "Unknown format:%d\n", bench_format);
357                 exit(1);
358                 break;
359         }
360 
361         free(worker_tab);
362         list_for_each_entry_safe(pos, n, &sender_contexts, list) {
363                 list_del_init(&pos->list);
364                 free(pos);
365         }
366         list_for_each_entry_safe(pos, n, &receiver_contexts, list) {
367                 list_del_init(&pos->list);
368                 free(pos);
369         }
370         return 0;
371 }
372 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

sflogo.php