xref: /freebsd/usr.sbin/bhyve/mevent.c (revision 7431dfd4580e850375fe5478d92ec770344db098)
1 /*-
2  * Copyright (c) 2011 NetApp, Inc.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD$
27  */
28 
29 /*
30  * Micro event library for FreeBSD, designed for a single i/o thread
31  * using kqueue, and having events be persistent by default.
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include <assert.h>
38 #include <errno.h>
39 #include <stdlib.h>
40 #include <stdio.h>
41 #include <string.h>
42 #include <unistd.h>
43 
44 #include <sys/types.h>
45 #include <sys/event.h>
46 #include <sys/time.h>
47 
48 #include <pthread.h>
49 #include <pthread_np.h>
50 
51 #include "mevent.h"
52 
53 #define	MEVENT_MAX	64
54 
55 #define	MEV_ADD		1
56 #define	MEV_ENABLE	2
57 #define	MEV_DISABLE	3
58 #define	MEV_DEL_PENDING	4
59 
60 extern char *vmname;
61 
62 static pthread_t mevent_tid;
63 static int mevent_timid = 43;
64 static int mevent_pipefd[2];
65 static pthread_mutex_t mevent_lmutex = PTHREAD_MUTEX_INITIALIZER;
66 
67 struct mevent {
68 	void	(*me_func)(int, enum ev_type, void *);
69 #define me_msecs me_fd
70 	int	me_fd;
71 	int	me_timid;
72 	enum ev_type me_type;
73 	void    *me_param;
74 	int	me_cq;
75 	int	me_state;
76 	int	me_closefd;
77 	LIST_ENTRY(mevent) me_list;
78 };
79 
80 static LIST_HEAD(listhead, mevent) global_head, change_head;
81 
82 static void
83 mevent_qlock(void)
84 {
85 	pthread_mutex_lock(&mevent_lmutex);
86 }
87 
88 static void
89 mevent_qunlock(void)
90 {
91 	pthread_mutex_unlock(&mevent_lmutex);
92 }
93 
94 static void
95 mevent_pipe_read(int fd, enum ev_type type, void *param)
96 {
97 	char buf[MEVENT_MAX];
98 	int status;
99 
100 	/*
101 	 * Drain the pipe read side. The fd is non-blocking so this is
102 	 * safe to do.
103 	 */
104 	do {
105 		status = read(fd, buf, sizeof(buf));
106 	} while (status == MEVENT_MAX);
107 }
108 
109 static void
110 mevent_notify(void)
111 {
112 	char c;
113 
114 	/*
115 	 * If calling from outside the i/o thread, write a byte on the
116 	 * pipe to force the i/o thread to exit the blocking kevent call.
117 	 */
118 	if (mevent_pipefd[1] != 0 && pthread_self() != mevent_tid) {
119 		write(mevent_pipefd[1], &c, 1);
120 	}
121 }
122 
123 static int
124 mevent_kq_filter(struct mevent *mevp)
125 {
126 	int retval;
127 
128 	retval = 0;
129 
130 	if (mevp->me_type == EVF_READ)
131 		retval = EVFILT_READ;
132 
133 	if (mevp->me_type == EVF_WRITE)
134 		retval = EVFILT_WRITE;
135 
136 	if (mevp->me_type == EVF_TIMER)
137 		retval = EVFILT_TIMER;
138 
139 	if (mevp->me_type == EVF_SIGNAL)
140 		retval = EVFILT_SIGNAL;
141 
142 	return (retval);
143 }
144 
145 static int
146 mevent_kq_flags(struct mevent *mevp)
147 {
148 	int ret;
149 
150 	switch (mevp->me_state) {
151 	case MEV_ADD:
152 		ret = EV_ADD;		/* implicitly enabled */
153 		break;
154 	case MEV_ENABLE:
155 		ret = EV_ENABLE;
156 		break;
157 	case MEV_DISABLE:
158 		ret = EV_DISABLE;
159 		break;
160 	case MEV_DEL_PENDING:
161 		ret = EV_DELETE;
162 		break;
163 	default:
164 		assert(0);
165 		break;
166 	}
167 
168 	return (ret);
169 }
170 
171 static int
172 mevent_kq_fflags(struct mevent *mevp)
173 {
174 	/* XXX nothing yet, perhaps EV_EOF for reads ? */
175 	return (0);
176 }
177 
178 static int
179 mevent_build(int mfd, struct kevent *kev)
180 {
181 	struct mevent *mevp, *tmpp;
182 	int i;
183 
184 	i = 0;
185 
186 	mevent_qlock();
187 
188 	LIST_FOREACH_SAFE(mevp, &change_head, me_list, tmpp) {
189 		if (mevp->me_closefd) {
190 			/*
191 			 * A close of the file descriptor will remove the
192 			 * event
193 			 */
194 			close(mevp->me_fd);
195 		} else {
196 			if (mevp->me_type == EVF_TIMER) {
197 				kev[i].ident = mevp->me_timid;
198 				kev[i].data = mevp->me_msecs;
199 			} else {
200 				kev[i].ident = mevp->me_fd;
201 				kev[i].data = 0;
202 			}
203 			kev[i].filter = mevent_kq_filter(mevp);
204 			kev[i].flags = mevent_kq_flags(mevp);
205 			kev[i].fflags = mevent_kq_fflags(mevp);
206 			kev[i].udata = mevp;
207 			i++;
208 		}
209 
210 		mevp->me_cq = 0;
211 		LIST_REMOVE(mevp, me_list);
212 
213 		if (mevp->me_state == MEV_DEL_PENDING) {
214 			free(mevp);
215 		} else {
216 			LIST_INSERT_HEAD(&global_head, mevp, me_list);
217 		}
218 
219 		assert(i < MEVENT_MAX);
220 	}
221 
222 	mevent_qunlock();
223 
224 	return (i);
225 }
226 
227 static void
228 mevent_handle(struct kevent *kev, int numev)
229 {
230 	struct mevent *mevp;
231 	int i;
232 
233 	for (i = 0; i < numev; i++) {
234 		mevp = kev[i].udata;
235 
236 		/* XXX check for EV_ERROR ? */
237 
238 		(*mevp->me_func)(mevp->me_fd, mevp->me_type, mevp->me_param);
239 	}
240 }
241 
242 struct mevent *
243 mevent_add(int tfd, enum ev_type type,
244 	   void (*func)(int, enum ev_type, void *), void *param)
245 {
246 	struct mevent *lp, *mevp;
247 
248 	if (tfd < 0 || func == NULL) {
249 		return (NULL);
250 	}
251 
252 	mevp = NULL;
253 
254 	mevent_qlock();
255 
256 	/*
257 	 * Verify that the fd/type tuple is not present in any list
258 	 */
259 	LIST_FOREACH(lp, &global_head, me_list) {
260 		if (type != EVF_TIMER && lp->me_fd == tfd &&
261 		    lp->me_type == type) {
262 			goto exit;
263 		}
264 	}
265 
266 	LIST_FOREACH(lp, &change_head, me_list) {
267 		if (type != EVF_TIMER && lp->me_fd == tfd &&
268 		    lp->me_type == type) {
269 			goto exit;
270 		}
271 	}
272 
273 	/*
274 	 * Allocate an entry, populate it, and add it to the change list.
275 	 */
276 	mevp = calloc(1, sizeof(struct mevent));
277 	if (mevp == NULL) {
278 		goto exit;
279 	}
280 
281 	if (type == EVF_TIMER) {
282 		mevp->me_msecs = tfd;
283 		mevp->me_timid = mevent_timid++;
284 	} else
285 		mevp->me_fd = tfd;
286 	mevp->me_type = type;
287 	mevp->me_func = func;
288 	mevp->me_param = param;
289 
290 	LIST_INSERT_HEAD(&change_head, mevp, me_list);
291 	mevp->me_cq = 1;
292 	mevp->me_state = MEV_ADD;
293 	mevent_notify();
294 
295 exit:
296 	mevent_qunlock();
297 
298 	return (mevp);
299 }
300 
301 static int
302 mevent_update(struct mevent *evp, int newstate)
303 {
304 	/*
305 	 * It's not possible to enable/disable a deleted event
306 	 */
307 	if (evp->me_state == MEV_DEL_PENDING)
308 		return (EINVAL);
309 
310 	/*
311 	 * No update needed if state isn't changing
312 	 */
313 	if (evp->me_state == newstate)
314 		return (0);
315 
316 	mevent_qlock();
317 
318 	evp->me_state = newstate;
319 
320 	/*
321 	 * Place the entry onto the changed list if not already there.
322 	 */
323 	if (evp->me_cq == 0) {
324 		evp->me_cq = 1;
325 		LIST_REMOVE(evp, me_list);
326 		LIST_INSERT_HEAD(&change_head, evp, me_list);
327 		mevent_notify();
328 	}
329 
330 	mevent_qunlock();
331 
332 	return (0);
333 }
334 
335 int
336 mevent_enable(struct mevent *evp)
337 {
338 
339 	return (mevent_update(evp, MEV_ENABLE));
340 }
341 
342 int
343 mevent_disable(struct mevent *evp)
344 {
345 
346 	return (mevent_update(evp, MEV_DISABLE));
347 }
348 
349 static int
350 mevent_delete_event(struct mevent *evp, int closefd)
351 {
352 	mevent_qlock();
353 
354 	/*
355          * Place the entry onto the changed list if not already there, and
356 	 * mark as to be deleted.
357          */
358         if (evp->me_cq == 0) {
359 		evp->me_cq = 1;
360 		LIST_REMOVE(evp, me_list);
361 		LIST_INSERT_HEAD(&change_head, evp, me_list);
362 		mevent_notify();
363         }
364 	evp->me_state = MEV_DEL_PENDING;
365 
366 	if (closefd)
367 		evp->me_closefd = 1;
368 
369 	mevent_qunlock();
370 
371 	return (0);
372 }
373 
374 int
375 mevent_delete(struct mevent *evp)
376 {
377 
378 	return (mevent_delete_event(evp, 0));
379 }
380 
381 int
382 mevent_delete_close(struct mevent *evp)
383 {
384 
385 	return (mevent_delete_event(evp, 1));
386 }
387 
388 static void
389 mevent_set_name(void)
390 {
391 
392 	pthread_set_name_np(mevent_tid, "mevent");
393 }
394 
395 void
396 mevent_dispatch(void)
397 {
398 	struct kevent changelist[MEVENT_MAX];
399 	struct kevent eventlist[MEVENT_MAX];
400 	struct mevent *pipev;
401 	int mfd;
402 	int numev;
403 	int ret;
404 
405 	mevent_tid = pthread_self();
406 	mevent_set_name();
407 
408 	mfd = kqueue();
409 	assert(mfd > 0);
410 
411 	/*
412 	 * Open the pipe that will be used for other threads to force
413 	 * the blocking kqueue call to exit by writing to it. Set the
414 	 * descriptor to non-blocking.
415 	 */
416 	ret = pipe(mevent_pipefd);
417 	if (ret < 0) {
418 		perror("pipe");
419 		exit(0);
420 	}
421 
422 	/*
423 	 * Add internal event handler for the pipe write fd
424 	 */
425 	pipev = mevent_add(mevent_pipefd[0], EVF_READ, mevent_pipe_read, NULL);
426 	assert(pipev != NULL);
427 
428 	for (;;) {
429 		/*
430 		 * Build changelist if required.
431 		 * XXX the changelist can be put into the blocking call
432 		 * to eliminate the extra syscall. Currently better for
433 		 * debug.
434 		 */
435 		numev = mevent_build(mfd, changelist);
436 		if (numev) {
437 			ret = kevent(mfd, changelist, numev, NULL, 0, NULL);
438 			if (ret == -1) {
439 				perror("Error return from kevent change");
440 			}
441 		}
442 
443 		/*
444 		 * Block awaiting events
445 		 */
446 		ret = kevent(mfd, NULL, 0, eventlist, MEVENT_MAX, NULL);
447 		if (ret == -1 && errno != EINTR) {
448 			perror("Error return from kevent monitor");
449 		}
450 
451 		/*
452 		 * Handle reported events
453 		 */
454 		mevent_handle(eventlist, ret);
455 	}
456 }
457