xref: /freebsd/usr.sbin/bhyve/mevent.c (revision f5f7c05209ca2c3748fd8b27c5e80ffad49120eb)
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_pipefd[2];
63 static pthread_mutex_t mevent_lmutex = PTHREAD_MUTEX_INITIALIZER;
64 
65 struct mevent {
66 	void	(*me_func)(int, enum ev_type, void *);
67 	int	me_fd;
68 	enum ev_type me_type;
69 	void    *me_param;
70 	int	me_cq;
71 	int	me_state;
72 	int	me_closefd;
73 	LIST_ENTRY(mevent) me_list;
74 };
75 
76 static LIST_HEAD(listhead, mevent) global_head, change_head;
77 
78 static void
79 mevent_qlock(void)
80 {
81 	pthread_mutex_lock(&mevent_lmutex);
82 }
83 
84 static void
85 mevent_qunlock(void)
86 {
87 	pthread_mutex_unlock(&mevent_lmutex);
88 }
89 
90 static void
91 mevent_pipe_read(int fd, enum ev_type type, void *param)
92 {
93 	char buf[MEVENT_MAX];
94 	int status;
95 
96 	/*
97 	 * Drain the pipe read side. The fd is non-blocking so this is
98 	 * safe to do.
99 	 */
100 	do {
101 		status = read(fd, buf, sizeof(buf));
102 	} while (status == MEVENT_MAX);
103 }
104 
105 static void
106 mevent_notify(void)
107 {
108 	char c;
109 
110 	/*
111 	 * If calling from outside the i/o thread, write a byte on the
112 	 * pipe to force the i/o thread to exit the blocking kevent call.
113 	 */
114 	if (mevent_pipefd[1] != 0 && pthread_self() != mevent_tid) {
115 		write(mevent_pipefd[1], &c, 1);
116 	}
117 }
118 
119 static int
120 mevent_kq_filter(struct mevent *mevp)
121 {
122 	int retval;
123 
124 	retval = 0;
125 
126 	if (mevp->me_type == EVF_READ)
127 		retval = EVFILT_READ;
128 
129 	if (mevp->me_type == EVF_WRITE)
130 		retval = EVFILT_WRITE;
131 
132 	return (retval);
133 }
134 
135 static int
136 mevent_kq_flags(struct mevent *mevp)
137 {
138 	int ret;
139 
140 	switch (mevp->me_state) {
141 	case MEV_ENABLE:
142 		ret = EV_ADD;
143 		break;
144 	case MEV_DISABLE:
145 		ret = EV_DISABLE;
146 		break;
147 	case MEV_DEL_PENDING:
148 		ret = EV_DELETE;
149 		break;
150 	}
151 
152 	return (ret);
153 }
154 
155 static int
156 mevent_kq_fflags(struct mevent *mevp)
157 {
158 	/* XXX nothing yet, perhaps EV_EOF for reads ? */
159 	return (0);
160 }
161 
162 static int
163 mevent_build(int mfd, struct kevent *kev)
164 {
165 	struct mevent *mevp, *tmpp;
166 	int i;
167 
168 	i = 0;
169 
170 	mevent_qlock();
171 
172 	LIST_FOREACH_SAFE(mevp, &change_head, me_list, tmpp) {
173 		if (mevp->me_closefd) {
174 			/*
175 			 * A close of the file descriptor will remove the
176 			 * event
177 			 */
178 			close(mevp->me_fd);
179 		} else {
180 			kev[i].ident = mevp->me_fd;
181 			kev[i].filter = mevent_kq_filter(mevp);
182 			kev[i].flags = mevent_kq_flags(mevp);
183 			kev[i].fflags = mevent_kq_fflags(mevp);
184 			kev[i].data = 0;
185 			kev[i].udata = mevp;
186 			i++;
187 		}
188 
189 		mevp->me_cq = 0;
190 		LIST_REMOVE(mevp, me_list);
191 
192 		if (mevp->me_state == MEV_DEL_PENDING) {
193 			free(mevp);
194 		} else {
195 			LIST_INSERT_HEAD(&global_head, mevp, me_list);
196 		}
197 
198 		assert(i < MEVENT_MAX);
199 	}
200 
201 	mevent_qunlock();
202 
203 	return (i);
204 }
205 
206 static void
207 mevent_handle(struct kevent *kev, int numev)
208 {
209 	struct mevent *mevp;
210 	int i;
211 
212 	for (i = 0; i < numev; i++) {
213 		mevp = kev[i].udata;
214 
215 		/* XXX check for EV_ERROR ? */
216 
217 		(*mevp->me_func)(mevp->me_fd, mevp->me_type, mevp->me_param);
218 	}
219 }
220 
221 struct mevent *
222 mevent_add(int fd, enum ev_type type,
223 	   void (*func)(int, enum ev_type, void *), void *param)
224 {
225 	struct mevent *lp, *mevp;
226 
227 	if (fd < 0 || func == NULL) {
228 		return (NULL);
229 	}
230 
231 	mevp = NULL;
232 
233 	mevent_qlock();
234 
235 	/*
236 	 * Verify that the fd/type tuple is not present in any list
237 	 */
238 	LIST_FOREACH(lp, &global_head, me_list) {
239 		if (lp->me_fd == fd && lp->me_type == type) {
240 			goto exit;
241 		}
242 	}
243 
244 	LIST_FOREACH(lp, &change_head, me_list) {
245 		if (lp->me_fd == fd && lp->me_type == type) {
246 			goto exit;
247 		}
248 	}
249 
250 	/*
251 	 * Allocate an entry, populate it, and add it to the change list.
252 	 */
253 	mevp = malloc(sizeof(struct mevent));
254 	if (mevp == NULL) {
255 		goto exit;
256 	}
257 
258 	memset(mevp, 0, sizeof(struct mevent));
259 	mevp->me_fd = fd;
260 	mevp->me_type = type;
261 	mevp->me_func = func;
262 	mevp->me_param = param;
263 
264 	LIST_INSERT_HEAD(&change_head, mevp, me_list);
265 	mevp->me_cq = 1;
266 	mevp->me_state = MEV_ENABLE;
267 	mevent_notify();
268 
269 exit:
270 	mevent_qunlock();
271 
272 	return (mevp);
273 }
274 
275 static int
276 mevent_update(struct mevent *evp, int newstate)
277 {
278 	/*
279 	 * It's not possible to enable/disable a deleted event
280 	 */
281 	if (evp->me_state == MEV_DEL_PENDING)
282 		return (EINVAL);
283 
284 	/*
285 	 * No update needed if state isn't changing
286 	 */
287 	if (evp->me_state == newstate)
288 		return (0);
289 
290 	mevent_qlock();
291 
292 	evp->me_state = newstate;
293 
294 	/*
295 	 * Place the entry onto the changed list if not already there.
296 	 */
297 	if (evp->me_cq == 0) {
298 		evp->me_cq = 1;
299 		LIST_REMOVE(evp, me_list);
300 		LIST_INSERT_HEAD(&change_head, evp, me_list);
301 		mevent_notify();
302 	}
303 
304 	mevent_qunlock();
305 
306 	return (0);
307 }
308 
309 int
310 mevent_enable(struct mevent *evp)
311 {
312 
313 	return (mevent_update(evp, MEV_ENABLE));
314 }
315 
316 int
317 mevent_disable(struct mevent *evp)
318 {
319 
320 	return (mevent_update(evp, MEV_DISABLE));
321 }
322 
323 static int
324 mevent_delete_event(struct mevent *evp, int closefd)
325 {
326 	mevent_qlock();
327 
328 	/*
329          * Place the entry onto the changed list if not already there, and
330 	 * mark as to be deleted.
331          */
332         if (evp->me_cq == 0) {
333 		evp->me_cq = 1;
334 		LIST_REMOVE(evp, me_list);
335 		LIST_INSERT_HEAD(&change_head, evp, me_list);
336 		mevent_notify();
337         }
338 	evp->me_state = MEV_DEL_PENDING;
339 
340 	if (closefd)
341 		evp->me_closefd = 1;
342 
343 	mevent_qunlock();
344 
345 	return (0);
346 }
347 
348 int
349 mevent_delete(struct mevent *evp)
350 {
351 
352 	return (mevent_delete_event(evp, 0));
353 }
354 
355 int
356 mevent_delete_close(struct mevent *evp)
357 {
358 
359 	return (mevent_delete_event(evp, 1));
360 }
361 
362 static void
363 mevent_set_name(void)
364 {
365 	char tname[MAXCOMLEN + 1];
366 
367 	snprintf(tname, sizeof(tname), "%s mevent", vmname);
368 	pthread_set_name_np(mevent_tid, tname);
369 }
370 
371 void
372 mevent_dispatch(void)
373 {
374 	struct kevent changelist[MEVENT_MAX];
375 	struct kevent eventlist[MEVENT_MAX];
376 	struct mevent *pipev;
377 	int mfd;
378 	int numev;
379 	int ret;
380 
381 	mevent_tid = pthread_self();
382 	mevent_set_name();
383 
384 	mfd = kqueue();
385 	assert(mfd > 0);
386 
387 	/*
388 	 * Open the pipe that will be used for other threads to force
389 	 * the blocking kqueue call to exit by writing to it. Set the
390 	 * descriptor to non-blocking.
391 	 */
392 	ret = pipe(mevent_pipefd);
393 	if (ret < 0) {
394 		perror("pipe");
395 		exit(0);
396 	}
397 
398 	/*
399 	 * Add internal event handler for the pipe write fd
400 	 */
401 	pipev = mevent_add(mevent_pipefd[0], EVF_READ, mevent_pipe_read, NULL);
402 	assert(pipev != NULL);
403 
404 	for (;;) {
405 		/*
406 		 * Build changelist if required.
407 		 * XXX the changelist can be put into the blocking call
408 		 * to eliminate the extra syscall. Currently better for
409 		 * debug.
410 		 */
411 		numev = mevent_build(mfd, changelist);
412 		if (numev) {
413 			ret = kevent(mfd, changelist, numev, NULL, 0, NULL);
414 			if (ret == -1) {
415 				perror("Error return from kevent change");
416 			}
417 		}
418 
419 		/*
420 		 * Block awaiting events
421 		 */
422 		ret = kevent(mfd, NULL, 0, eventlist, MEVENT_MAX, NULL);
423 		if (ret == -1) {
424 			perror("Error return from kevent monitor");
425 		}
426 
427 		/*
428 		 * Handle reported events
429 		 */
430 		mevent_handle(eventlist, ret);
431 	}
432 }
433