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
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
|
/*
Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
This file is licensed to you under your choice of the GNU Lesser
General Public License, version 3 or any later version (LGPLv3 or
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
#include <sys/poll.h>
#include <pthread.h>
#include <unistd.h>
#include <fcntl.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>
#include "logging.h"
#include "event.h"
#include "mem-pool.h"
#include "common-utils.h"
#ifndef _CONFIG_H
#define _CONFIG_H
#include "config.h"
#endif
#ifdef HAVE_SYS_EPOLL_H
#include <sys/epoll.h>
struct event_slot_epoll {
int fd;
int events;
int gen;
int ref;
int do_close;
int in_handler;
void *data;
event_handler_t handler;
gf_lock_t lock;
};
struct event_thread_data {
struct event_pool *event_pool;
int event_index;
};
static struct event_slot_epoll *
__event_newtable (struct event_pool *event_pool, int table_idx)
{
struct event_slot_epoll *table = NULL;
int i = -1;
table = GF_CALLOC (sizeof (*table), EVENT_EPOLL_SLOTS,
gf_common_mt_ereg);
if (!table)
return NULL;
for (i = 0; i < EVENT_EPOLL_SLOTS; i++) {
table[i].fd = -1;
LOCK_INIT (&table[i].lock);
}
event_pool->ereg[table_idx] = table;
event_pool->slots_used[table_idx] = 0;
return table;
}
static int
__event_slot_alloc (struct event_pool *event_pool, int fd)
{
int i = 0;
int table_idx = -1;
int gen = -1;
struct event_slot_epoll *table = NULL;
for (i = 0; i < EVENT_EPOLL_TABLES; i++) {
switch (event_pool->slots_used[i]) {
case EVENT_EPOLL_SLOTS:
continue;
case 0:
if (!event_pool->ereg[i]) {
table = __event_newtable (event_pool, i);
if (!table)
return -1;
}
break;
default:
table = event_pool->ereg[i];
break;
}
if (table)
/* break out of the loop */
break;
}
if (!table)
return -1;
table_idx = i;
for (i = 0; i < EVENT_EPOLL_SLOTS; i++) {
if (table[i].fd == -1) {
/* wipe everything except bump the generation */
gen = table[i].gen;
memset (&table[i], 0, sizeof (table[i]));
table[i].gen = gen + 1;
LOCK_INIT (&table[i].lock);
table[i].fd = fd;
event_pool->slots_used[table_idx]++;
break;
}
}
return table_idx * EVENT_EPOLL_SLOTS + i;
}
static int
event_slot_alloc (struct event_pool *event_pool, int fd)
{
int idx = -1;
pthread_mutex_lock (&event_pool->mutex);
{
idx = __event_slot_alloc (event_pool, fd);
}
pthread_mutex_unlock (&event_pool->mutex);
return idx;
}
static void
__event_slot_dealloc (struct event_pool *event_pool, int idx)
{
int table_idx = 0;
int offset = 0;
struct event_slot_epoll *table = NULL;
struct event_slot_epoll *slot = NULL;
table_idx = idx / EVENT_EPOLL_SLOTS;
offset = idx % EVENT_EPOLL_SLOTS;
table = event_pool->ereg[table_idx];
if (!table)
return;
slot = &table[offset];
slot->gen++;
slot->fd = -1;
event_pool->slots_used[table_idx]--;
return;
}
static void
event_slot_dealloc (struct event_pool *event_pool, int idx)
{
pthread_mutex_lock (&event_pool->mutex);
{
__event_slot_dealloc (event_pool, idx);
}
pthread_mutex_unlock (&event_pool->mutex);
return;
}
static struct event_slot_epoll *
event_slot_get (struct event_pool *event_pool, int idx)
{
struct event_slot_epoll *slot = NULL;
struct event_slot_epoll *table = NULL;
int table_idx = 0;
int offset = 0;
table_idx = idx / EVENT_EPOLL_SLOTS;
offset = idx % EVENT_EPOLL_SLOTS;
table = event_pool->ereg[table_idx];
if (!table)
return NULL;
slot = &table[offset];
LOCK (&slot->lock);
{
slot->ref++;
}
UNLOCK (&slot->lock);
return slot;
}
static void
event_slot_unref (struct event_pool *event_pool, struct event_slot_epoll *slot,
int idx)
{
int ref = -1;
int fd = -1;
int do_close = 0;
LOCK (&slot->lock);
{
ref = --slot->ref;
fd = slot->fd;
do_close = slot->do_close;
}
UNLOCK (&slot->lock);
if (ref)
/* slot still alive */
goto done;
event_slot_dealloc (event_pool, idx);
if (do_close)
close (fd);
done:
return;
}
static struct event_pool *
event_pool_new_epoll (int count, int eventthreadcount)
{
struct event_pool *event_pool = NULL;
int epfd = -1;
event_pool = GF_CALLOC (1, sizeof (*event_pool),
gf_common_mt_event_pool);
if (!event_pool)
goto out;
epfd = epoll_create (count);
if (epfd == -1) {
gf_log ("epoll", GF_LOG_ERROR, "epoll fd creation failed (%s)",
strerror (errno));
GF_FREE (event_pool->reg);
GF_FREE (event_pool);
event_pool = NULL;
goto out;
}
event_pool->fd = epfd;
event_pool->count = count;
event_pool->eventthreadcount = eventthreadcount;
pthread_mutex_init (&event_pool->mutex, NULL);
out:
return event_pool;
}
static void
__slot_update_events (struct event_slot_epoll *slot, int poll_in, int poll_out)
{
switch (poll_in) {
case 1:
slot->events |= EPOLLIN;
break;
case 0:
slot->events &= ~EPOLLIN;
break;
case -1:
/* do nothing */
break;
default:
gf_log ("epoll", GF_LOG_ERROR,
"invalid poll_in value %d", poll_in);
break;
}
switch (poll_out) {
case 1:
slot->events |= EPOLLOUT;
break;
case 0:
slot->events &= ~EPOLLOUT;
break;
case -1:
/* do nothing */
break;
default:
gf_log ("epoll", GF_LOG_ERROR,
"invalid poll_out value %d", poll_out);
break;
}
}
int
event_register_epoll (struct event_pool *event_pool, int fd,
event_handler_t handler,
void *data, int poll_in, int poll_out)
{
int idx = -1;
int ret = -1;
struct epoll_event epoll_event = {0, };
struct event_data *ev_data = (void *)&epoll_event.data;
struct event_slot_epoll *slot = NULL;
GF_VALIDATE_OR_GOTO ("event", event_pool, out);
idx = event_slot_alloc (event_pool, fd);
if (idx == -1) {
gf_log ("epoll", GF_LOG_ERROR,
"could not find slot for fd=%d", fd);
return -1;
}
slot = event_slot_get (event_pool, idx);
assert (slot->fd == fd);
LOCK (&slot->lock);
{
/* make epoll 'singleshot', which
means we need to re-add the fd with
epoll_ctl(EPOLL_CTL_MOD) after delivery of every
single event. This assures us that while a poller
thread has picked up and is processing an event,
another poller will not try to pick this at the same
time as well.
*/
slot->events = EPOLLPRI | EPOLLONESHOT;
slot->handler = handler;
slot->data = data;
__slot_update_events (slot, poll_in, poll_out);
epoll_event.events = slot->events;
ev_data->idx = idx;
ev_data->gen = slot->gen;
ret = epoll_ctl (event_pool->fd, EPOLL_CTL_ADD, fd,
&epoll_event);
/* check ret after UNLOCK() to avoid deadlock in
event_slot_unref()
*/
}
UNLOCK (&slot->lock);
if (ret == -1) {
gf_log ("epoll", GF_LOG_ERROR,
"failed to add fd(=%d) to epoll fd(=%d) (%s)",
fd, event_pool->fd, strerror (errno));
event_slot_unref (event_pool, slot, idx);
idx = -1;
}
/* keep slot->ref (do not event_slot_unref) if successful */
out:
return idx;
}
static int
event_unregister_epoll_common (struct event_pool *event_pool, int fd,
int idx, int do_close)
{
int ret = -1;
struct event_slot_epoll *slot = NULL;
GF_VALIDATE_OR_GOTO ("event", event_pool, out);
slot = event_slot_get (event_pool, idx);
assert (slot->fd == fd);
LOCK (&slot->lock);
{
ret = epoll_ctl (event_pool->fd, EPOLL_CTL_DEL, fd, NULL);
if (ret == -1) {
gf_log ("epoll", GF_LOG_ERROR,
"fail to del fd(=%d) from epoll fd(=%d) (%s)",
fd, event_pool->fd, strerror (errno));
goto unlock;
}
slot->do_close = do_close;
slot->gen++; /* detect unregister in dispatch_handler() */
}
unlock:
UNLOCK (&slot->lock);
event_slot_unref (event_pool, slot, idx); /* one for event_register() */
event_slot_unref (event_pool, slot, idx); /* one for event_slot_get() */
out:
return ret;
}
static int
event_unregister_epoll (struct event_pool *event_pool, int fd, int idx_hint)
{
int ret = -1;
ret = event_unregister_epoll_common (event_pool, fd, idx_hint, 0);
return ret;
}
static int
event_unregister_close_epoll (struct event_pool *event_pool, int fd,
int idx_hint)
{
int ret = -1;
ret = event_unregister_epoll_common (event_pool, fd, idx_hint, 1);
return ret;
}
static int
event_select_on_epoll (struct event_pool *event_pool, int fd, int idx,
int poll_in, int poll_out)
{
int ret = -1;
struct event_slot_epoll *slot = NULL;
struct epoll_event epoll_event = {0, };
struct event_data *ev_data = (void *)&epoll_event.data;
GF_VALIDATE_OR_GOTO ("event", event_pool, out);
slot = event_slot_get (event_pool, idx);
assert (slot->fd == fd);
LOCK (&slot->lock);
{
__slot_update_events (slot, poll_in, poll_out);
epoll_event.events = slot->events;
ev_data->idx = idx;
ev_data->gen = slot->gen;
if (slot->in_handler)
/* in_handler indicates at least one thread
executing event_dispatch_epoll_handler()
which will perform epoll_ctl(EPOLL_CTL_MOD)
anyways (because of EPOLLET)
This not only saves a system call, but also
avoids possibility of another epoll thread
parallely picking up the next event while the
ongoing handler is still in progress (and
resulting in unnecessary contention on
rpc_transport_t->mutex).
*/
goto unlock;
ret = epoll_ctl (event_pool->fd, EPOLL_CTL_MOD, fd,
&epoll_event);
if (ret == -1) {
gf_log ("epoll", GF_LOG_ERROR,
"failed to modify fd(=%d) events to %d",
fd, epoll_event.events);
}
}
unlock:
UNLOCK (&slot->lock);
event_slot_unref (event_pool, slot, idx);
out:
return idx;
}
static int
event_dispatch_epoll_handler (struct event_pool *event_pool,
struct epoll_event *event)
{
struct event_data *ev_data = NULL;
struct event_slot_epoll *slot = NULL;
event_handler_t handler = NULL;
void *data = NULL;
int idx = -1;
int gen = -1;
int ret = -1;
int fd = -1;
ev_data = (void *)&event->data;
handler = NULL;
data = NULL;
idx = ev_data->idx;
gen = ev_data->gen;
slot = event_slot_get (event_pool, idx);
LOCK (&slot->lock);
{
fd = slot->fd;
if (fd == -1) {
gf_log ("epoll", GF_LOG_ERROR,
"stale fd found on idx=%d, gen=%d, events=%d, "
"slot->gen=%d",
idx, gen, event->events, slot->gen);
/* fd got unregistered in another thread */
goto pre_unlock;
}
if (gen != slot->gen) {
gf_log ("epoll", GF_LOG_ERROR,
"generation mismatch on idx=%d, gen=%d, "
"slot->gen=%d, slot->fd=%d",
idx, gen, slot->gen, slot->fd);
/* slot was re-used and therefore is another fd! */
goto pre_unlock;
}
handler = slot->handler;
data = slot->data;
slot->in_handler++;
}
pre_unlock:
UNLOCK (&slot->lock);
if (!handler)
goto out;
ret = handler (fd, idx, data,
(event->events & (EPOLLIN|EPOLLPRI)),
(event->events & (EPOLLOUT)),
(event->events & (EPOLLERR|EPOLLHUP)));
LOCK (&slot->lock);
{
slot->in_handler--;
if (gen != slot->gen) {
/* event_unregister() happened while we were
in handler()
*/
gf_log ("epoll", GF_LOG_DEBUG,
"generation bumped on idx=%d from "
"gen=%d to slot->gen=%d, fd=%d, "
"slot->fd=%d",
idx, gen, slot->gen, fd, slot->fd);
goto post_unlock;
}
/* This call also picks up the changes made by another
thread calling event_select_on_epoll() while this
thread was busy in handler()
*/
if (slot->in_handler == 0) {
event->events = slot->events;
ret = epoll_ctl (event_pool->fd, EPOLL_CTL_MOD,
fd, event);
}
}
post_unlock:
UNLOCK (&slot->lock);
out:
event_slot_unref (event_pool, slot, idx);
return ret;
}
static void *
event_dispatch_epoll_worker (void *data)
{
struct epoll_event event;
int ret = -1;
struct event_thread_data *ev_data = data;
struct event_pool *event_pool;
int myindex = -1;
int timetodie = 0;
GF_VALIDATE_OR_GOTO ("event", ev_data, out);
event_pool = ev_data->event_pool;
myindex = ev_data->event_index;
GF_VALIDATE_OR_GOTO ("event", event_pool, out);
gf_log ("epoll", GF_LOG_INFO, "Started thread with index %d", myindex);
for (;;) {
if (event_pool->eventthreadcount < myindex) {
/* ...time to die, thread count was decreased below
* this threads index */
/* Start with extra safety at this point, reducing
* lock conention in normal case when threads are not
* reconfigured always */
pthread_mutex_lock (&event_pool->mutex);
{
if (event_pool->eventthreadcount <
myindex) {
/* if found true in critical section,
* die */
event_pool->pollers[myindex - 1] = 0;
timetodie = 1;
}
}
pthread_mutex_unlock (&event_pool->mutex);
if (timetodie) {
gf_log ("epoll", GF_LOG_INFO,
"Exited thread with index %d", myindex);
goto out;
}
}
ret = epoll_wait (event_pool->fd, &event, 1, -1);
if (ret == 0)
/* timeout */
continue;
if (ret == -1 && errno == EINTR)
/* sys call */
continue;
ret = event_dispatch_epoll_handler (event_pool, &event);
}
out:
if (ev_data)
GF_FREE (ev_data);
return NULL;
}
/* Attempts to start the # of configured pollers, ensuring at least the first
* is started in a joinable state */
static int
event_dispatch_epoll (struct event_pool *event_pool)
{
int i = 0;
pthread_t t_id;
int pollercount = 0;
int ret = -1;
struct event_thread_data *ev_data = NULL;
/* Start the configured number of pollers */
pthread_mutex_lock (&event_pool->mutex);
{
pollercount = event_pool->eventthreadcount;
/* Set to MAX if greater */
if (pollercount > EVENT_MAX_THREADS)
pollercount = EVENT_MAX_THREADS;
/* Default pollers to 1 in case this is incorrectly set */
if (pollercount <= 0)
pollercount = 1;
for (i = 0; i < pollercount; i++) {
ev_data = GF_CALLOC (1, sizeof (*ev_data),
gf_common_mt_event_pool);
if (!ev_data) {
gf_log ("epoll", GF_LOG_WARNING,
"Allocation failure for index %d", i);
if (i == 0) {
/* Need to suceed creating 0'th
* thread, to joinable and wait */
break;
} else {
/* Inability to create other threads
* are a lesser evil, and ignored */
continue;
}
}
ev_data->event_pool = event_pool;
ev_data->event_index = i + 1;
ret = pthread_create (&t_id, NULL,
event_dispatch_epoll_worker,
ev_data);
if (!ret) {
event_pool->pollers[i] = t_id;
/* mark all threads other than one in index 0
* as detachable. Errors can be ignored, they
* spend their time as zombies if not detched
* and the thread counts are decreased */
if (i != 0)
pthread_detach (event_pool->pollers[i]);
} else {
gf_log ("epoll", GF_LOG_WARNING,
"Failed to start thread for index %d",
i);
if (i == 0) {
GF_FREE (ev_data);
break;
} else {
GF_FREE (ev_data);
continue;
}
}
}
}
pthread_mutex_unlock (&event_pool->mutex);
/* Just wait for the first thread, that is created in a joinable state
* and will never die, ensuring this function never returns */
if (event_pool->pollers[0] != 0)
pthread_join (event_pool->pollers[0], NULL);
return ret;
}
int
event_reconfigure_threads_epoll (struct event_pool *event_pool, int value)
{
int i;
int ret;
pthread_t t_id;
int oldthreadcount;
struct event_thread_data *ev_data = NULL;
/* Set to MAX if greater */
if (value > EVENT_MAX_THREADS)
value = EVENT_MAX_THREADS;
/* Default pollers to 1 in case this is set incorrectly */
if (value <= 0)
value = 1;
pthread_mutex_lock (&event_pool->mutex);
{
oldthreadcount = event_pool->eventthreadcount;
if (oldthreadcount < value) {
/* create more poll threads */
for (i = oldthreadcount; i < value; i++) {
/* Start a thread if the index at this location
* is a 0, so that the older thread is confirmed
* as dead */
if (event_pool->pollers[i] == 0) {
ev_data = GF_CALLOC (1,
sizeof (*ev_data),
gf_common_mt_event_pool);
if (!ev_data) {
gf_log ("epoll", GF_LOG_WARNING,
"Allocation failure for"
" index %d", i);
continue;
}
ev_data->event_pool = event_pool;
ev_data->event_index = i + 1;
ret = pthread_create (&t_id, NULL,
event_dispatch_epoll_worker,
ev_data);
if (ret) {
gf_log ("epoll", GF_LOG_WARNING,
"Failed to start thread for"
" index %d", i);
GF_FREE (ev_data);
} else {
pthread_detach (t_id);
event_pool->pollers[i] = t_id;
}
}
}
}
/* if value decreases, threads will terminate, themselves */
event_pool->eventthreadcount = value;
}
pthread_mutex_unlock (&event_pool->mutex);
return 0;
}
struct event_ops event_ops_epoll = {
.new = event_pool_new_epoll,
.event_register = event_register_epoll,
.event_select_on = event_select_on_epoll,
.event_unregister = event_unregister_epoll,
.event_unregister_close = event_unregister_close_epoll,
.event_dispatch = event_dispatch_epoll,
.event_reconfigure_threads = event_reconfigure_threads_epoll
};
#endif
|