1/*	$OpenBSD: poll.c,v 1.2 2002/06/25 15:50:15 mickey Exp $	*/
2
3/*
4 * Copyright 2000-2003 Niels Provos <provos@citi.umich.edu>
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. The name of the author may not be used to endorse or promote products
16 *    derived from this software without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 */
29#ifdef HAVE_CONFIG_H
30#include "config.h"
31#endif
32
33#include <sys/types.h>
34#ifdef HAVE_SYS_TIME_H
35#include <sys/time.h>
36#else
37#include <sys/_time.h>
38#endif
39#include <sys/queue.h>
40#include <sys/tree.h>
41#include <poll.h>
42#include <signal.h>
43#include <stdio.h>
44#include <stdlib.h>
45#include <string.h>
46#include <unistd.h>
47#include <errno.h>
48#ifdef CHECK_INVARIANTS
49#include <assert.h>
50#endif
51
52#include "event.h"
53#include "event-internal.h"
54#include "evsignal.h"
55#include "log.h"
56
57extern volatile sig_atomic_t evsignal_caught;
58
59struct pollop {
60	int event_count;		/* Highest number alloc */
61	int nfds;                       /* Size of event_* */
62	int fd_count;                   /* Size of idxplus1_by_fd */
63	struct pollfd *event_set;
64	struct event **event_r_back;
65	struct event **event_w_back;
66	int *idxplus1_by_fd; /* Index into event_set by fd; we add 1 so
67			      * that 0 (which is easy to memset) can mean
68			      * "no entry." */
69};
70
71void *poll_init	(void);
72int poll_add		(void *, struct event *);
73int poll_del		(void *, struct event *);
74int poll_recalc		(struct event_base *, void *, int);
75int poll_dispatch	(struct event_base *, void *, struct timeval *);
76void poll_dealloc	(void *);
77
78const struct eventop pollops = {
79	"poll",
80	poll_init,
81	poll_add,
82	poll_del,
83	poll_recalc,
84	poll_dispatch,
85	poll_dealloc
86};
87
88void *
89poll_init(void)
90{
91	struct pollop *pollop;
92
93	/* Disable poll when this environment variable is set */
94	if (getenv("EVENT_NOPOLL"))
95		return (NULL);
96
97	if (!(pollop = calloc(1, sizeof(struct pollop))))
98		return (NULL);
99
100	evsignal_init();
101
102	return (pollop);
103}
104
105/*
106 * Called with the highest fd that we know about.  If it is 0, completely
107 * recalculate everything.
108 */
109
110int
111poll_recalc(struct event_base *base, void *arg, int max)
112{
113	return (0);
114}
115
116#ifdef CHECK_INVARIANTS
117static void
118poll_check_ok(struct pollop *pop)
119{
120	int i, idx;
121	struct event *ev;
122
123	for (i = 0; i < pop->fd_count; ++i) {
124		idx = pop->idxplus1_by_fd[i]-1;
125		if (idx < 0)
126			continue;
127		assert(pop->event_set[idx].fd == i);
128		if (pop->event_set[idx].events & POLLIN) {
129			ev = pop->event_r_back[idx];
130			assert(ev);
131			assert(ev->ev_events & EV_READ);
132			assert(ev->ev_fd == i);
133		}
134		if (pop->event_set[idx].events & POLLOUT) {
135			ev = pop->event_w_back[idx];
136			assert(ev);
137			assert(ev->ev_events & EV_WRITE);
138			assert(ev->ev_fd == i);
139		}
140	}
141	for (i = 0; i < pop->nfds; ++i) {
142		struct pollfd *pfd = &pop->event_set[i];
143		assert(pop->idxplus1_by_fd[pfd->fd] == i+1);
144	}
145}
146#else
147#define poll_check_ok(pop)
148#endif
149
150int
151poll_dispatch(struct event_base *base, void *arg, struct timeval *tv)
152{
153	int res, i, sec, nfds;
154	struct pollop *pop = arg;
155
156	poll_check_ok(pop);
157	sec = tv->tv_sec * 1000 + (tv->tv_usec + 999) / 1000;
158	nfds = pop->nfds;
159	res = poll(pop->event_set, nfds, sec);
160
161	if (res == -1) {
162		if (errno != EINTR) {
163                        event_warn("poll");
164			return (-1);
165		}
166
167		evsignal_process();
168		return (0);
169	} else if (evsignal_caught)
170		evsignal_process();
171
172	event_debug(("%s: poll reports %d", __func__, res));
173
174	if (res == 0)
175		return (0);
176
177	for (i = 0; i < nfds; i++) {
178		int what = pop->event_set[i].revents;
179		struct event *r_ev = NULL, *w_ev = NULL;
180		if (!what)
181			continue;
182
183		res = 0;
184
185		/* If the file gets closed notify */
186		if (what & (POLLHUP|POLLERR))
187			what |= POLLIN|POLLOUT;
188		if (what & POLLIN) {
189			res |= EV_READ;
190			r_ev = pop->event_r_back[i];
191		}
192		if (what & POLLOUT) {
193			res |= EV_WRITE;
194			w_ev = pop->event_w_back[i];
195		}
196		if (res == 0)
197			continue;
198
199		if (r_ev && (res & r_ev->ev_events)) {
200			if (!(r_ev->ev_events & EV_PERSIST))
201				event_del(r_ev);
202			event_active(r_ev, res & r_ev->ev_events, 1);
203		}
204		if (w_ev && w_ev != r_ev && (res & w_ev->ev_events)) {
205			if (!(w_ev->ev_events & EV_PERSIST))
206				event_del(w_ev);
207			event_active(w_ev, res & w_ev->ev_events, 1);
208		}
209	}
210
211	return (0);
212}
213
214int
215poll_add(void *arg, struct event *ev)
216{
217	struct pollop *pop = arg;
218	struct pollfd *pfd = NULL;
219	int i;
220
221	if (ev->ev_events & EV_SIGNAL)
222		return (evsignal_add(ev));
223	if (!(ev->ev_events & (EV_READ|EV_WRITE)))
224		return (0);
225
226	poll_check_ok(pop);
227	if (pop->nfds + 1 >= pop->event_count) {
228		struct pollfd *tmp_event_set;
229		struct event **tmp_event_r_back;
230		struct event **tmp_event_w_back;
231		int tmp_event_count;
232
233		if (pop->event_count < 32)
234			tmp_event_count = 32;
235		else
236			tmp_event_count = pop->event_count * 2;
237
238		/* We need more file descriptors */
239		tmp_event_set = realloc(pop->event_set,
240				 tmp_event_count * sizeof(struct pollfd));
241		if (tmp_event_set == NULL) {
242			event_warn("realloc");
243			return (-1);
244		}
245		pop->event_set = tmp_event_set;
246
247		tmp_event_r_back = realloc(pop->event_r_back,
248			    tmp_event_count * sizeof(struct event *));
249		if (tmp_event_r_back == NULL) {
250			/* event_set overallocated; that's okay. */
251			event_warn("realloc");
252			return (-1);
253		}
254		pop->event_r_back = tmp_event_r_back;
255
256		tmp_event_w_back = realloc(pop->event_w_back,
257			    tmp_event_count * sizeof(struct event *));
258		if (tmp_event_w_back == NULL) {
259			/* event_set and event_r_back overallocated; that's
260			 * okay. */
261			event_warn("realloc");
262			return (-1);
263		}
264		pop->event_w_back = tmp_event_w_back;
265
266		pop->event_count = tmp_event_count;
267	}
268	if (ev->ev_fd >= pop->fd_count) {
269		int *tmp_idxplus1_by_fd;
270		int new_count;
271		if (pop->fd_count < 32)
272			new_count = 32;
273		else
274			new_count = pop->fd_count * 2;
275		while (new_count <= ev->ev_fd)
276			new_count *= 2;
277		tmp_idxplus1_by_fd =
278			realloc(pop->idxplus1_by_fd, new_count * sizeof(int));
279		if (tmp_idxplus1_by_fd == NULL) {
280			event_warn("realloc");
281			return (-1);
282		}
283		pop->idxplus1_by_fd = tmp_idxplus1_by_fd;
284		memset(pop->idxplus1_by_fd + pop->fd_count,
285		       0, sizeof(int)*(new_count - pop->fd_count));
286		pop->fd_count = new_count;
287	}
288
289	i = pop->idxplus1_by_fd[ev->ev_fd] - 1;
290	if (i >= 0) {
291		pfd = &pop->event_set[i];
292	} else {
293		i = pop->nfds++;
294		pfd = &pop->event_set[i];
295		pfd->events = 0;
296		pfd->fd = ev->ev_fd;
297		pop->event_w_back[i] = pop->event_r_back[i] = NULL;
298		pop->idxplus1_by_fd[ev->ev_fd] = i + 1;
299	}
300
301	pfd->revents = 0;
302	if (ev->ev_events & EV_WRITE) {
303		pfd->events |= POLLOUT;
304		pop->event_w_back[i] = ev;
305	}
306	if (ev->ev_events & EV_READ) {
307		pfd->events |= POLLIN;
308		pop->event_r_back[i] = ev;
309	}
310	poll_check_ok(pop);
311
312	return (0);
313}
314
315/*
316 * Nothing to be done here.
317 */
318
319int
320poll_del(void *arg, struct event *ev)
321{
322	struct pollop *pop = arg;
323	struct pollfd *pfd = NULL;
324	int i;
325
326	if (ev->ev_events & EV_SIGNAL)
327		return (evsignal_del(ev));
328
329	if (!(ev->ev_events & (EV_READ|EV_WRITE)))
330		return (0);
331
332	poll_check_ok(pop);
333	i = pop->idxplus1_by_fd[ev->ev_fd] - 1;
334	if (i < 0)
335		return (-1);
336
337	/* Do we still want to read or write? */
338	pfd = &pop->event_set[i];
339	if (ev->ev_events & EV_READ) {
340		pfd->events &= ~POLLIN;
341		pop->event_r_back[i] = NULL;
342	}
343	if (ev->ev_events & EV_WRITE) {
344		pfd->events &= ~POLLOUT;
345		pop->event_w_back[i] = NULL;
346	}
347	poll_check_ok(pop);
348	if (pfd->events)
349		/* Another event cares about that fd. */
350		return (0);
351
352	/* Okay, so we aren't interested in that fd anymore. */
353	pop->idxplus1_by_fd[ev->ev_fd] = 0;
354
355	--pop->nfds;
356	if (i != pop->nfds) {
357		/*
358		 * Shift the last pollfd down into the now-unoccupied
359		 * position.
360		 */
361		memcpy(&pop->event_set[i], &pop->event_set[pop->nfds],
362		       sizeof(struct pollfd));
363		pop->event_r_back[i] = pop->event_r_back[pop->nfds];
364		pop->event_w_back[i] = pop->event_w_back[pop->nfds];
365		pop->idxplus1_by_fd[pop->event_set[i].fd] = i + 1;
366	}
367
368	poll_check_ok(pop);
369	return (0);
370}
371
372void
373poll_dealloc(void *arg)
374{
375	struct pollop *pop = arg;
376
377	if (pop->event_set)
378		free(pop->event_set);
379	if (pop->event_r_back)
380		free(pop->event_r_back);
381	if (pop->event_w_back)
382		free(pop->event_w_back);
383	if (pop->idxplus1_by_fd)
384		free(pop->idxplus1_by_fd);
385
386	memset(pop, 0, sizeof(struct pollop));
387	free(pop);
388}
389