xref: /freebsd/usr.sbin/bhyve/mevent.c (revision bc3f5ec90bde2f3a5e4021d133c89793d68b8c73)
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_ENABLE	1
56 #define MEV_DISABLE	2
57 #define MEV_DEL_PENDING	3
58 
59 extern char *vmname;
60 
61 static pthread_t mevent_tid;
62 static int mevent_timid = 43;
63 static int mevent_pipefd[2];
64 static pthread_mutex_t mevent_lmutex = PTHREAD_MUTEX_INITIALIZER;
65 
66 struct mevent {
67 	void	(*me_func)(int, enum ev_type, void *);
68 #define me_msecs me_fd
69 	int	me_fd;
70 	int	me_timid;
71 	enum ev_type me_type;
72 	void    *me_param;
73 	int	me_cq;
74 	int	me_state;
75 	int	me_closefd;
76 	LIST_ENTRY(mevent) me_list;
77 };
78 
79 static LIST_HEAD(listhead, mevent) global_head, change_head;
80 
81 static void
82 mevent_qlock(void)
83 {
84 	pthread_mutex_lock(&mevent_lmutex);
85 }
86 
87 static void
88 mevent_qunlock(void)
89 {
90 	pthread_mutex_unlock(&mevent_lmutex);
91 }
92 
93 static void
94 mevent_pipe_read(int fd, enum ev_type type, void *param)
95 {
96 	char buf[MEVENT_MAX];
97 	int status;
98 
99 	/*
100 	 * Drain the pipe read side. The fd is non-blocking so this is
101 	 * safe to do.
102 	 */
103 	do {
104 		status = read(fd, buf, sizeof(buf));
105 	} while (status == MEVENT_MAX);
106 }
107 
108 static void
109 mevent_notify(void)
110 {
111 	char c;
112 
113 	/*
114 	 * If calling from outside the i/o thread, write a byte on the
115 	 * pipe to force the i/o thread to exit the blocking kevent call.
116 	 */
117 	if (mevent_pipefd[1] != 0 && pthread_self() != mevent_tid) {
118 		write(mevent_pipefd[1], &c, 1);
119 	}
120 }
121 
122 static int
123 mevent_kq_filter(struct mevent *mevp)
124 {
125 	int retval;
126 
127 	retval = 0;
128 
129 	if (mevp->me_type == EVF_READ)
130 		retval = EVFILT_READ;
131 
132 	if (mevp->me_type == EVF_WRITE)
133 		retval = EVFILT_WRITE;
134 
135 	if (mevp->me_type == EVF_TIMER)
136 		retval = EVFILT_TIMER;
137 
138 	if (mevp->me_type == EVF_SIGNAL)
139 		retval = EVFILT_SIGNAL;
140 
141 	return (retval);
142 }
143 
144 static int
145 mevent_kq_flags(struct mevent *mevp)
146 {
147 	int ret;
148 
149 	switch (mevp->me_state) {
150 	case MEV_ENABLE:
151 		ret = EV_ADD;
152 		if (mevp->me_type == EVF_TIMER)
153 			ret |= EV_ENABLE;
154 		break;
155 	case MEV_DISABLE:
156 		ret = EV_DISABLE;
157 		break;
158 	case MEV_DEL_PENDING:
159 		ret = EV_DELETE;
160 		break;
161 	}
162 
163 	return (ret);
164 }
165 
166 static int
167 mevent_kq_fflags(struct mevent *mevp)
168 {
169 	/* XXX nothing yet, perhaps EV_EOF for reads ? */
170 	return (0);
171 }
172 
173 static int
174 mevent_build(int mfd, struct kevent *kev)
175 {
176 	struct mevent *mevp, *tmpp;
177 	int i;
178 
179 	i = 0;
180 
181 	mevent_qlock();
182 
183 	LIST_FOREACH_SAFE(mevp, &change_head, me_list, tmpp) {
184 		if (mevp->me_closefd) {
185 			/*
186 			 * A close of the file descriptor will remove the
187 			 * event
188 			 */
189 			close(mevp->me_fd);
190 		} else {
191 			if (mevp->me_type == EVF_TIMER) {
192 				kev[i].ident = mevp->me_timid;
193 				kev[i].data = mevp->me_msecs;
194 			} else {
195 				kev[i].ident = mevp->me_fd;
196 				kev[i].data = 0;
197 			}
198 			kev[i].filter = mevent_kq_filter(mevp);
199 			kev[i].flags = mevent_kq_flags(mevp);
200 			kev[i].fflags = mevent_kq_fflags(mevp);
201 			kev[i].udata = mevp;
202 			i++;
203 		}
204 
205 		mevp->me_cq = 0;
206 		LIST_REMOVE(mevp, me_list);
207 
208 		if (mevp->me_state == MEV_DEL_PENDING) {
209 			free(mevp);
210 		} else {
211 			LIST_INSERT_HEAD(&global_head, mevp, me_list);
212 		}
213 
214 		assert(i < MEVENT_MAX);
215 	}
216 
217 	mevent_qunlock();
218 
219 	return (i);
220 }
221 
222 static void
223 mevent_handle(struct kevent *kev, int numev)
224 {
225 	struct mevent *mevp;
226 	int i;
227 
228 	for (i = 0; i < numev; i++) {
229 		mevp = kev[i].udata;
230 
231 		/* XXX check for EV_ERROR ? */
232 
233 		(*mevp->me_func)(mevp->me_fd, mevp->me_type, mevp->me_param);
234 	}
235 }
236 
237 struct mevent *
238 mevent_add(int tfd, enum ev_type type,
239 	   void (*func)(int, enum ev_type, void *), void *param)
240 {
241 	struct mevent *lp, *mevp;
242 
243 	if (tfd < 0 || func == NULL) {
244 		return (NULL);
245 	}
246 
247 	mevp = NULL;
248 
249 	mevent_qlock();
250 
251 	/*
252 	 * Verify that the fd/type tuple is not present in any list
253 	 */
254 	LIST_FOREACH(lp, &global_head, me_list) {
255 		if (type != EVF_TIMER && lp->me_fd == tfd &&
256 		    lp->me_type == type) {
257 			goto exit;
258 		}
259 	}
260 
261 	LIST_FOREACH(lp, &change_head, me_list) {
262 		if (type != EVF_TIMER && lp->me_fd == tfd &&
263 		    lp->me_type == type) {
264 			goto exit;
265 		}
266 	}
267 
268 	/*
269 	 * Allocate an entry, populate it, and add it to the change list.
270 	 */
271 	mevp = malloc(sizeof(struct mevent));
272 	if (mevp == NULL) {
273 		goto exit;
274 	}
275 
276 	memset(mevp, 0, sizeof(struct mevent));
277 	if (type == EVF_TIMER) {
278 		mevp->me_msecs = tfd;
279 		mevp->me_timid = mevent_timid++;
280 	} else
281 		mevp->me_fd = tfd;
282 	mevp->me_type = type;
283 	mevp->me_func = func;
284 	mevp->me_param = param;
285 
286 	LIST_INSERT_HEAD(&change_head, mevp, me_list);
287 	mevp->me_cq = 1;
288 	mevp->me_state = MEV_ENABLE;
289 	mevent_notify();
290 
291 exit:
292 	mevent_qunlock();
293 
294 	return (mevp);
295 }
296 
297 static int
298 mevent_update(struct mevent *evp, int newstate)
299 {
300 	/*
301 	 * It's not possible to enable/disable a deleted event
302 	 */
303 	if (evp->me_state == MEV_DEL_PENDING)
304 		return (EINVAL);
305 
306 	/*
307 	 * No update needed if state isn't changing
308 	 */
309 	if (evp->me_state == newstate)
310 		return (0);
311 
312 	mevent_qlock();
313 
314 	evp->me_state = newstate;
315 
316 	/*
317 	 * Place the entry onto the changed list if not already there.
318 	 */
319 	if (evp->me_cq == 0) {
320 		evp->me_cq = 1;
321 		LIST_REMOVE(evp, me_list);
322 		LIST_INSERT_HEAD(&change_head, evp, me_list);
323 		mevent_notify();
324 	}
325 
326 	mevent_qunlock();
327 
328 	return (0);
329 }
330 
331 int
332 mevent_enable(struct mevent *evp)
333 {
334 
335 	return (mevent_update(evp, MEV_ENABLE));
336 }
337 
338 int
339 mevent_disable(struct mevent *evp)
340 {
341 
342 	return (mevent_update(evp, MEV_DISABLE));
343 }
344 
345 static int
346 mevent_delete_event(struct mevent *evp, int closefd)
347 {
348 	mevent_qlock();
349 
350 	/*
351          * Place the entry onto the changed list if not already there, and
352 	 * mark as to be deleted.
353          */
354         if (evp->me_cq == 0) {
355 		evp->me_cq = 1;
356 		LIST_REMOVE(evp, me_list);
357 		LIST_INSERT_HEAD(&change_head, evp, me_list);
358 		mevent_notify();
359         }
360 	evp->me_state = MEV_DEL_PENDING;
361 
362 	if (closefd)
363 		evp->me_closefd = 1;
364 
365 	mevent_qunlock();
366 
367 	return (0);
368 }
369 
370 int
371 mevent_delete(struct mevent *evp)
372 {
373 
374 	return (mevent_delete_event(evp, 0));
375 }
376 
377 int
378 mevent_delete_close(struct mevent *evp)
379 {
380 
381 	return (mevent_delete_event(evp, 1));
382 }
383 
384 static void
385 mevent_set_name(void)
386 {
387 
388 	pthread_set_name_np(mevent_tid, "mevent");
389 }
390 
391 void
392 mevent_dispatch(void)
393 {
394 	struct kevent changelist[MEVENT_MAX];
395 	struct kevent eventlist[MEVENT_MAX];
396 	struct mevent *pipev;
397 	int mfd;
398 	int numev;
399 	int ret;
400 
401 	mevent_tid = pthread_self();
402 	mevent_set_name();
403 
404 	mfd = kqueue();
405 	assert(mfd > 0);
406 
407 	/*
408 	 * Open the pipe that will be used for other threads to force
409 	 * the blocking kqueue call to exit by writing to it. Set the
410 	 * descriptor to non-blocking.
411 	 */
412 	ret = pipe(mevent_pipefd);
413 	if (ret < 0) {
414 		perror("pipe");
415 		exit(0);
416 	}
417 
418 	/*
419 	 * Add internal event handler for the pipe write fd
420 	 */
421 	pipev = mevent_add(mevent_pipefd[0], EVF_READ, mevent_pipe_read, NULL);
422 	assert(pipev != NULL);
423 
424 	for (;;) {
425 		/*
426 		 * Build changelist if required.
427 		 * XXX the changelist can be put into the blocking call
428 		 * to eliminate the extra syscall. Currently better for
429 		 * debug.
430 		 */
431 		numev = mevent_build(mfd, changelist);
432 		if (numev) {
433 			ret = kevent(mfd, changelist, numev, NULL, 0, NULL);
434 			if (ret == -1) {
435 				perror("Error return from kevent change");
436 			}
437 		}
438 
439 		/*
440 		 * Block awaiting events
441 		 */
442 		ret = kevent(mfd, NULL, 0, eventlist, MEVENT_MAX, NULL);
443 		if (ret == -1 && errno != EINTR) {
444 			perror("Error return from kevent monitor");
445 		}
446 
447 		/*
448 		 * Handle reported events
449 		 */
450 		mevent_handle(eventlist, ret);
451 	}
452 }
453