summaryrefslogtreecommitdiff
path: root/lib/tevent/tevent_epoll.c
blob: 61a8525d2bcfb11216890a063c94f8a7cc9e415a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
/* 
   Unix SMB/CIFS implementation.

   main select loop and event handling - epoll implementation

   Copyright (C) Andrew Tridgell	2003-2005
   Copyright (C) Stefan Metzmacher	2005-2009

     ** NOTE! The following LGPL license applies to the tevent
     ** library. This does NOT imply that all of Samba is released
     ** under the LGPL

   This library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 3 of the License, or (at your option) any later version.

   This library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/

#include "replace.h"
#include "system/filesys.h"
#include "system/select.h"
#include "tevent.h"
#include "tevent_internal.h"
#include "tevent_util.h"

struct epoll_event_context {
	/* a pointer back to the generic event_context */
	struct tevent_context *ev;

	/* when using epoll this is the handle from epoll_create */
	int epoll_fd;

	pid_t pid;

	bool (*panic_fallback)(struct tevent_context *ev, bool replay);
};

#ifdef TEST_PANIC_FALLBACK
static int epoll_wait_panic_fallback(int epfd,
				struct epoll_event *events,
				int maxevents,
				int timeout)
{
	/* 50% of the time, fail... */
	if ((random() % 2) == 0) {
		errno = EINVAL;
		return -1;
	}

	return epoll_wait(epfd, events, maxevents, timeout);
}

#define epoll_wait epoll_wait_panic_fallback
#endif

/*
  called to set the panic fallback function.
*/
_PRIVATE_ bool tevent_epoll_set_panic_fallback(struct tevent_context *ev,
				bool (*panic_fallback)(struct tevent_context *ev,
						       bool replay))
{
	struct epoll_event_context *epoll_ev;

	if (ev->additional_data == NULL) {
		return false;
	}

	epoll_ev = talloc_get_type(ev->additional_data,
				struct epoll_event_context);
	if (epoll_ev == NULL) {
		return false;
	}
	epoll_ev->panic_fallback = panic_fallback;
	return true;
}

/*
  called when a epoll call fails
*/
static void epoll_panic(struct epoll_event_context *epoll_ev,
			const char *reason, bool replay)
{
	struct tevent_context *ev = epoll_ev->ev;

	if (epoll_ev->panic_fallback == NULL) {
		tevent_debug(ev, TEVENT_DEBUG_FATAL,
			"%s (%s) replay[%u] - calling abort()\n",
			reason, strerror(errno), (unsigned)replay);
		abort();
	}

	tevent_debug(ev, TEVENT_DEBUG_WARNING,
		     "%s (%s) replay[%u] - calling panic_fallback\n",
		     reason, strerror(errno), (unsigned)replay);

	if (!epoll_ev->panic_fallback(ev, replay)) {
		/* Fallback failed. */
		tevent_debug(ev, TEVENT_DEBUG_FATAL,
			"%s (%s) replay[%u] - calling abort()\n",
			reason, strerror(errno), (unsigned)replay);
		abort();
	}
}

/*
  map from TEVENT_FD_* to EPOLLIN/EPOLLOUT
*/
static uint32_t epoll_map_flags(uint16_t flags)
{
	uint32_t ret = 0;
	if (flags & TEVENT_FD_READ) ret |= (EPOLLIN | EPOLLERR | EPOLLHUP);
	if (flags & TEVENT_FD_WRITE) ret |= (EPOLLOUT | EPOLLERR | EPOLLHUP);
	return ret;
}

/*
 free the epoll fd
*/
static int epoll_ctx_destructor(struct epoll_event_context *epoll_ev)
{
	close(epoll_ev->epoll_fd);
	epoll_ev->epoll_fd = -1;
	return 0;
}

/*
 init the epoll fd
*/
static int epoll_init_ctx(struct epoll_event_context *epoll_ev)
{
	epoll_ev->epoll_fd = epoll_create(64);
	if (epoll_ev->epoll_fd == -1) {
		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
			     "Failed to create epoll handle.\n");
		return -1;
	}

	if (!ev_set_close_on_exec(epoll_ev->epoll_fd)) {
		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_WARNING,
			     "Failed to set close-on-exec, file descriptor may be leaked to children.\n");
	}

	epoll_ev->pid = getpid();
	talloc_set_destructor(epoll_ev, epoll_ctx_destructor);

	return 0;
}

static void epoll_add_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde);

/*
  reopen the epoll handle when our pid changes
  see http://junkcode.samba.org/ftp/unpacked/junkcode/epoll_fork.c for an 
  demonstration of why this is needed
 */
static void epoll_check_reopen(struct epoll_event_context *epoll_ev)
{
	struct tevent_fd *fde;

	if (epoll_ev->pid == getpid()) {
		return;
	}

	close(epoll_ev->epoll_fd);
	epoll_ev->epoll_fd = epoll_create(64);
	if (epoll_ev->epoll_fd == -1) {
		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
			     "Failed to recreate epoll handle after fork\n");
		return;
	}

	if (!ev_set_close_on_exec(epoll_ev->epoll_fd)) {
		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_WARNING,
			     "Failed to set close-on-exec, file descriptor may be leaked to children.\n");
	}

	epoll_ev->pid = getpid();
	for (fde=epoll_ev->ev->fd_events;fde;fde=fde->next) {
		epoll_add_event(epoll_ev, fde);
	}
}

#define EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT	(1<<0)
#define EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR	(1<<1)
#define EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR	(1<<2)

/*
 add the epoll event to the given fd_event
*/
static void epoll_add_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
{
	struct epoll_event event;

	if (epoll_ev->epoll_fd == -1) return;

	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;

	/* if we don't want events yet, don't add an epoll_event */
	if (fde->flags == 0) return;

	ZERO_STRUCT(event);
	event.events = epoll_map_flags(fde->flags);
	event.data.ptr = fde;
	if (epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_ADD, fde->fd, &event) != 0) {
		epoll_panic(epoll_ev, "EPOLL_CTL_ADD failed", false);
		return;
	}
	fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;

	/* only if we want to read we want to tell the event handler about errors */
	if (fde->flags & TEVENT_FD_READ) {
		fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
	}
}

/*
 delete the epoll event for given fd_event
*/
static void epoll_del_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
{
	struct epoll_event event;

	if (epoll_ev->epoll_fd == -1) return;

	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;

	/* if there's no epoll_event, we don't need to delete it */
	if (!(fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT)) return;

	ZERO_STRUCT(event);
	event.events = epoll_map_flags(fde->flags);
	event.data.ptr = fde;
	if (epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_DEL, fde->fd, &event) != 0) {
		tevent_debug(epoll_ev->ev, TEVENT_DEBUG_FATAL,
			     "epoll_del_event failed! probable early close bug (%s)\n",
			     strerror(errno));
	}
	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT;
}

/*
 change the epoll event to the given fd_event
*/
static void epoll_mod_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
{
	struct epoll_event event;
	if (epoll_ev->epoll_fd == -1) return;

	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;

	ZERO_STRUCT(event);
	event.events = epoll_map_flags(fde->flags);
	event.data.ptr = fde;
	if (epoll_ctl(epoll_ev->epoll_fd, EPOLL_CTL_MOD, fde->fd, &event) != 0) {
		epoll_panic(epoll_ev, "EPOLL_CTL_MOD failed", false);
		return;
	}

	/* only if we want to read we want to tell the event handler about errors */
	if (fde->flags & TEVENT_FD_READ) {
		fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;
	}
}

static void epoll_change_event(struct epoll_event_context *epoll_ev, struct tevent_fd *fde)
{
	bool got_error = (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR);
	bool want_read = (fde->flags & TEVENT_FD_READ);
	bool want_write= (fde->flags & TEVENT_FD_WRITE);

	if (epoll_ev->epoll_fd == -1) return;

	fde->additional_flags &= ~EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR;

	/* there's already an event */
	if (fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_HAS_EVENT) {
		if (want_read || (want_write && !got_error)) {
			epoll_mod_event(epoll_ev, fde);
			return;
		}
		/* 
		 * if we want to match the select behavior, we need to remove the epoll_event
		 * when the caller isn't interested in events.
		 *
		 * this is because epoll reports EPOLLERR and EPOLLHUP, even without asking for them
		 */
		epoll_del_event(epoll_ev, fde);
		return;
	}

	/* there's no epoll_event attached to the fde */
	if (want_read || (want_write && !got_error)) {
		epoll_add_event(epoll_ev, fde);
		return;
	}
}

/*
  event loop handling using epoll
*/
static int epoll_event_loop(struct epoll_event_context *epoll_ev, struct timeval *tvalp)
{
	int ret, i;
#define MAXEVENTS 1
	struct epoll_event events[MAXEVENTS];
	int timeout = -1;

	if (epoll_ev->epoll_fd == -1) return -1;

	if (tvalp) {
		/* it's better to trigger timed events a bit later than too early */
		timeout = ((tvalp->tv_usec+999) / 1000) + (tvalp->tv_sec*1000);
	}

	if (epoll_ev->ev->signal_events &&
	    tevent_common_check_signal(epoll_ev->ev)) {
		return 0;
	}

	tevent_trace_point_callback(epoll_ev->ev, TEVENT_TRACE_BEFORE_WAIT);
	ret = epoll_wait(epoll_ev->epoll_fd, events, MAXEVENTS, timeout);
	tevent_trace_point_callback(epoll_ev->ev, TEVENT_TRACE_AFTER_WAIT);

	if (ret == -1 && errno == EINTR && epoll_ev->ev->signal_events) {
		if (tevent_common_check_signal(epoll_ev->ev)) {
			return 0;
		}
	}

	if (ret == -1 && errno != EINTR) {
		epoll_panic(epoll_ev, "epoll_wait() failed", true);
		return -1;
	}

	if (ret == 0 && tvalp) {
		/* we don't care about a possible delay here */
		tevent_common_loop_timer_delay(epoll_ev->ev);
		return 0;
	}

	for (i=0;i<ret;i++) {
		struct tevent_fd *fde = talloc_get_type(events[i].data.ptr, 
						       struct tevent_fd);
		uint16_t flags = 0;

		if (fde == NULL) {
			epoll_panic(epoll_ev, "epoll_wait() gave bad data", true);
			return -1;
		}
		if (events[i].events & (EPOLLHUP|EPOLLERR)) {
			fde->additional_flags |= EPOLL_ADDITIONAL_FD_FLAG_GOT_ERROR;
			/*
			 * if we only wait for TEVENT_FD_WRITE, we should not tell the
			 * event handler about it, and remove the epoll_event,
			 * as we only report errors when waiting for read events,
			 * to match the select() behavior
			 */
			if (!(fde->additional_flags & EPOLL_ADDITIONAL_FD_FLAG_REPORT_ERROR)) {
				epoll_del_event(epoll_ev, fde);
				continue;
			}
			flags |= TEVENT_FD_READ;
		}
		if (events[i].events & EPOLLIN) flags |= TEVENT_FD_READ;
		if (events[i].events & EPOLLOUT) flags |= TEVENT_FD_WRITE;
		if (flags) {
			fde->handler(epoll_ev->ev, fde, flags, fde->private_data);
			break;
		}
	}

	return 0;
}

/*
  create a epoll_event_context structure.
*/
static int epoll_event_context_init(struct tevent_context *ev)
{
	int ret;
	struct epoll_event_context *epoll_ev;

	/*
	 * We might be called during tevent_re_initialise()
	 * which means we need to free our old additional_data.
	 */
	TALLOC_FREE(ev->additional_data);

	epoll_ev = talloc_zero(ev, struct epoll_event_context);
	if (!epoll_ev) return -1;
	epoll_ev->ev = ev;
	epoll_ev->epoll_fd = -1;

	ret = epoll_init_ctx(epoll_ev);
	if (ret != 0) {
		talloc_free(epoll_ev);
		return ret;
	}

	ev->additional_data = epoll_ev;
	return 0;
}

/*
  destroy an fd_event
*/
static int epoll_event_fd_destructor(struct tevent_fd *fde)
{
	struct tevent_context *ev = fde->event_ctx;
	struct epoll_event_context *epoll_ev = NULL;

	if (ev) {
		epoll_ev = talloc_get_type(ev->additional_data,
					   struct epoll_event_context);

		epoll_check_reopen(epoll_ev);

		epoll_del_event(epoll_ev, fde);
	}

	return tevent_common_fd_destructor(fde);
}

/*
  add a fd based event
  return NULL on failure (memory allocation error)
*/
static struct tevent_fd *epoll_event_add_fd(struct tevent_context *ev, TALLOC_CTX *mem_ctx,
					    int fd, uint16_t flags,
					    tevent_fd_handler_t handler,
					    void *private_data,
					    const char *handler_name,
					    const char *location)
{
	struct epoll_event_context *epoll_ev = talloc_get_type(ev->additional_data,
							   struct epoll_event_context);
	struct tevent_fd *fde;

	epoll_check_reopen(epoll_ev);

	fde = tevent_common_add_fd(ev, mem_ctx, fd, flags,
				   handler, private_data,
				   handler_name, location);
	if (!fde) return NULL;

	talloc_set_destructor(fde, epoll_event_fd_destructor);

	epoll_add_event(epoll_ev, fde);

	return fde;
}

/*
  set the fd event flags
*/
static void epoll_event_set_fd_flags(struct tevent_fd *fde, uint16_t flags)
{
	struct tevent_context *ev;
	struct epoll_event_context *epoll_ev;

	if (fde->flags == flags) return;

	ev = fde->event_ctx;
	epoll_ev = talloc_get_type(ev->additional_data, struct epoll_event_context);

	fde->flags = flags;

	epoll_check_reopen(epoll_ev);

	epoll_change_event(epoll_ev, fde);
}

/*
  do a single event loop using the events defined in ev 
*/
static int epoll_event_loop_once(struct tevent_context *ev, const char *location)
{
	struct epoll_event_context *epoll_ev = talloc_get_type(ev->additional_data,
		 					   struct epoll_event_context);
	struct timeval tval;

	if (ev->signal_events &&
	    tevent_common_check_signal(ev)) {
		return 0;
	}

	if (ev->immediate_events &&
	    tevent_common_loop_immediate(ev)) {
		return 0;
	}

	tval = tevent_common_loop_timer_delay(ev);
	if (tevent_timeval_is_zero(&tval)) {
		return 0;
	}

	epoll_check_reopen(epoll_ev);

	return epoll_event_loop(epoll_ev, &tval);
}

static const struct tevent_ops epoll_event_ops = {
	.context_init		= epoll_event_context_init,
	.add_fd			= epoll_event_add_fd,
	.set_fd_close_fn	= tevent_common_fd_set_close_fn,
	.get_fd_flags		= tevent_common_fd_get_flags,
	.set_fd_flags		= epoll_event_set_fd_flags,
	.add_timer		= tevent_common_add_timer,
	.schedule_immediate	= tevent_common_schedule_immediate,
	.add_signal		= tevent_common_add_signal,
	.loop_once		= epoll_event_loop_once,
	.loop_wait		= tevent_common_loop_wait,
};

_PRIVATE_ bool tevent_epoll_init(void)
{
	return tevent_register_backend("epoll", &epoll_event_ops);
}