Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
Kitware
GitHub Repository: Kitware/CMake
Path: blob/master/Utilities/cmlibuv/src/win/core.c
3153 views
1
/* Copyright Joyent, Inc. and other Node contributors. All rights reserved.
2
*
3
* Permission is hereby granted, free of charge, to any person obtaining a copy
4
* of this software and associated documentation files (the "Software"), to
5
* deal in the Software without restriction, including without limitation the
6
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
7
* sell copies of the Software, and to permit persons to whom the Software is
8
* furnished to do so, subject to the following conditions:
9
*
10
* The above copyright notice and this permission notice shall be included in
11
* all copies or substantial portions of the Software.
12
*
13
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
18
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
19
* IN THE SOFTWARE.
20
*/
21
22
#include <assert.h>
23
#include <errno.h>
24
#include <limits.h>
25
#include <stdio.h>
26
#include <stdlib.h>
27
#include <string.h>
28
#if defined(_MSC_VER) || defined(__MINGW64_VERSION_MAJOR)
29
#include <crtdbg.h>
30
#endif
31
32
#include "uv.h"
33
#include "internal.h"
34
#include "queue.h"
35
#include "handle-inl.h"
36
#include "heap-inl.h"
37
#include "req-inl.h"
38
39
/* uv_once initialization guards */
40
static uv_once_t uv_init_guard_ = UV_ONCE_INIT;
41
42
43
#if defined(_DEBUG) && (defined(_MSC_VER) || defined(__MINGW64_VERSION_MAJOR))
44
/* Our crt debug report handler allows us to temporarily disable asserts
45
* just for the current thread.
46
*/
47
48
UV_THREAD_LOCAL int uv__crt_assert_enabled = TRUE;
49
50
static int uv__crt_dbg_report_handler(int report_type, char *message, int *ret_val) {
51
if (uv__crt_assert_enabled || report_type != _CRT_ASSERT)
52
return FALSE;
53
54
if (ret_val) {
55
/* Set ret_val to 0 to continue with normal execution.
56
* Set ret_val to 1 to trigger a breakpoint.
57
*/
58
59
if(IsDebuggerPresent())
60
*ret_val = 1;
61
else
62
*ret_val = 0;
63
}
64
65
/* Don't call _CrtDbgReport. */
66
return TRUE;
67
}
68
#else
69
UV_THREAD_LOCAL int uv__crt_assert_enabled = FALSE;
70
#endif
71
72
73
#if !defined(__MINGW32__) || __MSVCRT_VERSION__ >= 0x800
74
static void uv__crt_invalid_parameter_handler(const wchar_t* expression,
75
const wchar_t* function, const wchar_t * file, unsigned int line,
76
uintptr_t reserved) {
77
/* No-op. */
78
}
79
#endif
80
81
static uv_loop_t** uv__loops;
82
static int uv__loops_size;
83
static int uv__loops_capacity;
84
#define UV__LOOPS_CHUNK_SIZE 8
85
static uv_mutex_t uv__loops_lock;
86
87
88
static void uv__loops_init(void) {
89
uv_mutex_init(&uv__loops_lock);
90
}
91
92
93
static int uv__loops_add(uv_loop_t* loop) {
94
uv_loop_t** new_loops;
95
int new_capacity, i;
96
97
uv_mutex_lock(&uv__loops_lock);
98
99
if (uv__loops_size == uv__loops_capacity) {
100
new_capacity = uv__loops_capacity + UV__LOOPS_CHUNK_SIZE;
101
new_loops = uv__realloc(uv__loops, sizeof(uv_loop_t*) * new_capacity);
102
if (!new_loops)
103
goto failed_loops_realloc;
104
uv__loops = new_loops;
105
for (i = uv__loops_capacity; i < new_capacity; ++i)
106
uv__loops[i] = NULL;
107
uv__loops_capacity = new_capacity;
108
}
109
uv__loops[uv__loops_size] = loop;
110
++uv__loops_size;
111
112
uv_mutex_unlock(&uv__loops_lock);
113
return 0;
114
115
failed_loops_realloc:
116
uv_mutex_unlock(&uv__loops_lock);
117
return ERROR_OUTOFMEMORY;
118
}
119
120
121
static void uv__loops_remove(uv_loop_t* loop) {
122
int loop_index;
123
int smaller_capacity;
124
uv_loop_t** new_loops;
125
126
uv_mutex_lock(&uv__loops_lock);
127
128
for (loop_index = 0; loop_index < uv__loops_size; ++loop_index) {
129
if (uv__loops[loop_index] == loop)
130
break;
131
}
132
/* If loop was not found, ignore */
133
if (loop_index == uv__loops_size)
134
goto loop_removed;
135
136
uv__loops[loop_index] = uv__loops[uv__loops_size - 1];
137
uv__loops[uv__loops_size - 1] = NULL;
138
--uv__loops_size;
139
140
if (uv__loops_size == 0) {
141
uv__loops_capacity = 0;
142
uv__free(uv__loops);
143
uv__loops = NULL;
144
goto loop_removed;
145
}
146
147
/* If we didn't grow to big skip downsizing */
148
if (uv__loops_capacity < 4 * UV__LOOPS_CHUNK_SIZE)
149
goto loop_removed;
150
151
/* Downsize only if more than half of buffer is free */
152
smaller_capacity = uv__loops_capacity / 2;
153
if (uv__loops_size >= smaller_capacity)
154
goto loop_removed;
155
new_loops = uv__realloc(uv__loops, sizeof(uv_loop_t*) * smaller_capacity);
156
if (!new_loops)
157
goto loop_removed;
158
uv__loops = new_loops;
159
uv__loops_capacity = smaller_capacity;
160
161
loop_removed:
162
uv_mutex_unlock(&uv__loops_lock);
163
}
164
165
void uv__wake_all_loops(void) {
166
int i;
167
uv_loop_t* loop;
168
169
uv_mutex_lock(&uv__loops_lock);
170
for (i = 0; i < uv__loops_size; ++i) {
171
loop = uv__loops[i];
172
assert(loop);
173
if (loop->iocp != INVALID_HANDLE_VALUE)
174
PostQueuedCompletionStatus(loop->iocp, 0, 0, NULL);
175
}
176
uv_mutex_unlock(&uv__loops_lock);
177
}
178
179
static void uv__init(void) {
180
/* Tell Windows that we will handle critical errors. */
181
SetErrorMode(SEM_FAILCRITICALERRORS | SEM_NOGPFAULTERRORBOX |
182
SEM_NOOPENFILEERRORBOX);
183
184
/* Tell the CRT to not exit the application when an invalid parameter is
185
* passed. The main issue is that invalid FDs will trigger this behavior.
186
*/
187
#if !defined(__MINGW32__) || __MSVCRT_VERSION__ >= 0x800
188
_set_invalid_parameter_handler(uv__crt_invalid_parameter_handler);
189
#endif
190
191
/* We also need to setup our debug report handler because some CRT
192
* functions (eg _get_osfhandle) raise an assert when called with invalid
193
* FDs even though they return the proper error code in the release build.
194
*/
195
#if defined(_DEBUG) && (defined(_MSC_VER) || defined(__MINGW64_VERSION_MAJOR))
196
_CrtSetReportHook(uv__crt_dbg_report_handler);
197
#endif
198
199
/* Initialize tracking of all uv loops */
200
uv__loops_init();
201
202
/* Fetch winapi function pointers. This must be done first because other
203
* initialization code might need these function pointers to be loaded.
204
*/
205
uv__winapi_init();
206
207
/* Initialize winsock */
208
uv__winsock_init();
209
210
/* Initialize FS */
211
uv__fs_init();
212
213
/* Initialize signal stuff */
214
uv__signals_init();
215
216
/* Initialize console */
217
uv__console_init();
218
219
/* Initialize utilities */
220
uv__util_init();
221
222
/* Initialize system wakeup detection */
223
uv__init_detect_system_wakeup();
224
}
225
226
227
int uv_loop_init(uv_loop_t* loop) {
228
uv__loop_internal_fields_t* lfields;
229
struct heap* timer_heap;
230
int err;
231
232
/* Initialize libuv itself first */
233
uv__once_init();
234
235
/* Create an I/O completion port */
236
loop->iocp = CreateIoCompletionPort(INVALID_HANDLE_VALUE, NULL, 0, 1);
237
if (loop->iocp == NULL)
238
return uv_translate_sys_error(GetLastError());
239
240
lfields = (uv__loop_internal_fields_t*) uv__calloc(1, sizeof(*lfields));
241
if (lfields == NULL)
242
return UV_ENOMEM;
243
loop->internal_fields = lfields;
244
245
err = uv_mutex_init(&lfields->loop_metrics.lock);
246
if (err)
247
goto fail_metrics_mutex_init;
248
249
/* To prevent uninitialized memory access, loop->time must be initialized
250
* to zero before calling uv_update_time for the first time.
251
*/
252
loop->time = 0;
253
uv_update_time(loop);
254
255
QUEUE_INIT(&loop->wq);
256
QUEUE_INIT(&loop->handle_queue);
257
loop->active_reqs.count = 0;
258
loop->active_handles = 0;
259
260
loop->pending_reqs_tail = NULL;
261
262
loop->endgame_handles = NULL;
263
264
loop->timer_heap = timer_heap = uv__malloc(sizeof(*timer_heap));
265
if (timer_heap == NULL) {
266
err = UV_ENOMEM;
267
goto fail_timers_alloc;
268
}
269
270
heap_init(timer_heap);
271
272
loop->check_handles = NULL;
273
loop->prepare_handles = NULL;
274
loop->idle_handles = NULL;
275
276
loop->next_prepare_handle = NULL;
277
loop->next_check_handle = NULL;
278
loop->next_idle_handle = NULL;
279
280
memset(&loop->poll_peer_sockets, 0, sizeof loop->poll_peer_sockets);
281
282
loop->active_tcp_streams = 0;
283
loop->active_udp_streams = 0;
284
285
loop->timer_counter = 0;
286
loop->stop_flag = 0;
287
288
err = uv_mutex_init(&loop->wq_mutex);
289
if (err)
290
goto fail_mutex_init;
291
292
err = uv_async_init(loop, &loop->wq_async, uv__work_done);
293
if (err)
294
goto fail_async_init;
295
296
uv__handle_unref(&loop->wq_async);
297
loop->wq_async.flags |= UV_HANDLE_INTERNAL;
298
299
err = uv__loops_add(loop);
300
if (err)
301
goto fail_async_init;
302
303
return 0;
304
305
fail_async_init:
306
uv_mutex_destroy(&loop->wq_mutex);
307
308
fail_mutex_init:
309
uv__free(timer_heap);
310
loop->timer_heap = NULL;
311
312
fail_timers_alloc:
313
uv_mutex_destroy(&lfields->loop_metrics.lock);
314
315
fail_metrics_mutex_init:
316
uv__free(lfields);
317
loop->internal_fields = NULL;
318
CloseHandle(loop->iocp);
319
loop->iocp = INVALID_HANDLE_VALUE;
320
321
return err;
322
}
323
324
325
void uv_update_time(uv_loop_t* loop) {
326
uint64_t new_time = uv__hrtime(1000);
327
assert(new_time >= loop->time);
328
loop->time = new_time;
329
}
330
331
332
void uv__once_init(void) {
333
uv_once(&uv_init_guard_, uv__init);
334
}
335
336
337
void uv__loop_close(uv_loop_t* loop) {
338
uv__loop_internal_fields_t* lfields;
339
size_t i;
340
341
uv__loops_remove(loop);
342
343
/* Close the async handle without needing an extra loop iteration.
344
* We might have a pending message, but we're just going to destroy the IOCP
345
* soon, so we can just discard it now without the usual risk of a getting
346
* another notification from GetQueuedCompletionStatusEx after calling the
347
* close_cb (which we also skip defining). We'll assert later that queue was
348
* actually empty and all reqs handled. */
349
loop->wq_async.async_sent = 0;
350
loop->wq_async.close_cb = NULL;
351
uv__handle_closing(&loop->wq_async);
352
uv__handle_close(&loop->wq_async);
353
354
for (i = 0; i < ARRAY_SIZE(loop->poll_peer_sockets); i++) {
355
SOCKET sock = loop->poll_peer_sockets[i];
356
if (sock != 0 && sock != INVALID_SOCKET)
357
closesocket(sock);
358
}
359
360
uv_mutex_lock(&loop->wq_mutex);
361
assert(QUEUE_EMPTY(&loop->wq) && "thread pool work queue not empty!");
362
assert(!uv__has_active_reqs(loop));
363
uv_mutex_unlock(&loop->wq_mutex);
364
uv_mutex_destroy(&loop->wq_mutex);
365
366
uv__free(loop->timer_heap);
367
loop->timer_heap = NULL;
368
369
lfields = uv__get_internal_fields(loop);
370
uv_mutex_destroy(&lfields->loop_metrics.lock);
371
uv__free(lfields);
372
loop->internal_fields = NULL;
373
374
CloseHandle(loop->iocp);
375
}
376
377
378
int uv__loop_configure(uv_loop_t* loop, uv_loop_option option, va_list ap) {
379
uv__loop_internal_fields_t* lfields;
380
381
lfields = uv__get_internal_fields(loop);
382
if (option == UV_METRICS_IDLE_TIME) {
383
lfields->flags |= UV_METRICS_IDLE_TIME;
384
return 0;
385
}
386
387
return UV_ENOSYS;
388
}
389
390
391
int uv_backend_fd(const uv_loop_t* loop) {
392
return -1;
393
}
394
395
396
int uv_loop_fork(uv_loop_t* loop) {
397
return UV_ENOSYS;
398
}
399
400
401
static int uv__loop_alive(const uv_loop_t* loop) {
402
return uv__has_active_handles(loop) ||
403
uv__has_active_reqs(loop) ||
404
loop->pending_reqs_tail != NULL ||
405
loop->endgame_handles != NULL;
406
}
407
408
409
int uv_loop_alive(const uv_loop_t* loop) {
410
return uv__loop_alive(loop);
411
}
412
413
414
int uv_backend_timeout(const uv_loop_t* loop) {
415
if (loop->stop_flag == 0 &&
416
/* uv__loop_alive(loop) && */
417
(uv__has_active_handles(loop) || uv__has_active_reqs(loop)) &&
418
loop->pending_reqs_tail == NULL &&
419
loop->idle_handles == NULL &&
420
loop->endgame_handles == NULL)
421
return uv__next_timeout(loop);
422
return 0;
423
}
424
425
426
static void uv__poll_wine(uv_loop_t* loop, DWORD timeout) {
427
DWORD bytes;
428
ULONG_PTR key;
429
OVERLAPPED* overlapped;
430
uv_req_t* req;
431
int repeat;
432
uint64_t timeout_time;
433
uint64_t user_timeout;
434
int reset_timeout;
435
436
timeout_time = loop->time + timeout;
437
438
if (uv__get_internal_fields(loop)->flags & UV_METRICS_IDLE_TIME) {
439
reset_timeout = 1;
440
user_timeout = timeout;
441
timeout = 0;
442
} else {
443
reset_timeout = 0;
444
}
445
446
for (repeat = 0; ; repeat++) {
447
/* Only need to set the provider_entry_time if timeout != 0. The function
448
* will return early if the loop isn't configured with UV_METRICS_IDLE_TIME.
449
*/
450
if (timeout != 0)
451
uv__metrics_set_provider_entry_time(loop);
452
453
GetQueuedCompletionStatus(loop->iocp,
454
&bytes,
455
&key,
456
&overlapped,
457
timeout);
458
459
if (reset_timeout != 0) {
460
timeout = user_timeout;
461
reset_timeout = 0;
462
}
463
464
/* Placed here because on success the loop will break whether there is an
465
* empty package or not, or if GetQueuedCompletionStatus returned early then
466
* the timeout will be updated and the loop will run again. In either case
467
* the idle time will need to be updated.
468
*/
469
uv__metrics_update_idle_time(loop);
470
471
if (overlapped) {
472
/* Package was dequeued */
473
req = uv__overlapped_to_req(overlapped);
474
uv__insert_pending_req(loop, req);
475
476
/* Some time might have passed waiting for I/O,
477
* so update the loop time here.
478
*/
479
uv_update_time(loop);
480
} else if (GetLastError() != WAIT_TIMEOUT) {
481
/* Serious error */
482
uv_fatal_error(GetLastError(), "GetQueuedCompletionStatus");
483
} else if (timeout > 0) {
484
/* GetQueuedCompletionStatus can occasionally return a little early.
485
* Make sure that the desired timeout target time is reached.
486
*/
487
uv_update_time(loop);
488
if (timeout_time > loop->time) {
489
timeout = (DWORD)(timeout_time - loop->time);
490
/* The first call to GetQueuedCompletionStatus should return very
491
* close to the target time and the second should reach it, but
492
* this is not stated in the documentation. To make sure a busy
493
* loop cannot happen, the timeout is increased exponentially
494
* starting on the third round.
495
*/
496
timeout += repeat ? (1 << (repeat - 1)) : 0;
497
continue;
498
}
499
}
500
break;
501
}
502
}
503
504
505
static void uv__poll(uv_loop_t* loop, DWORD timeout) {
506
BOOL success;
507
uv_req_t* req;
508
OVERLAPPED_ENTRY overlappeds[128];
509
ULONG count;
510
ULONG i;
511
int repeat;
512
uint64_t timeout_time;
513
uint64_t user_timeout;
514
int reset_timeout;
515
516
timeout_time = loop->time + timeout;
517
518
if (uv__get_internal_fields(loop)->flags & UV_METRICS_IDLE_TIME) {
519
reset_timeout = 1;
520
user_timeout = timeout;
521
timeout = 0;
522
} else {
523
reset_timeout = 0;
524
}
525
526
for (repeat = 0; ; repeat++) {
527
/* Only need to set the provider_entry_time if timeout != 0. The function
528
* will return early if the loop isn't configured with UV_METRICS_IDLE_TIME.
529
*/
530
if (timeout != 0)
531
uv__metrics_set_provider_entry_time(loop);
532
533
success = pGetQueuedCompletionStatusEx(loop->iocp,
534
overlappeds,
535
ARRAY_SIZE(overlappeds),
536
&count,
537
timeout,
538
FALSE);
539
540
if (reset_timeout != 0) {
541
timeout = user_timeout;
542
reset_timeout = 0;
543
}
544
545
/* Placed here because on success the loop will break whether there is an
546
* empty package or not, or if GetQueuedCompletionStatus returned early then
547
* the timeout will be updated and the loop will run again. In either case
548
* the idle time will need to be updated.
549
*/
550
uv__metrics_update_idle_time(loop);
551
552
if (success) {
553
for (i = 0; i < count; i++) {
554
/* Package was dequeued, but see if it is not a empty package
555
* meant only to wake us up.
556
*/
557
if (overlappeds[i].lpOverlapped) {
558
req = uv__overlapped_to_req(overlappeds[i].lpOverlapped);
559
uv__insert_pending_req(loop, req);
560
}
561
}
562
563
/* Some time might have passed waiting for I/O,
564
* so update the loop time here.
565
*/
566
uv_update_time(loop);
567
} else if (GetLastError() != WAIT_TIMEOUT) {
568
/* Serious error */
569
uv_fatal_error(GetLastError(), "GetQueuedCompletionStatusEx");
570
} else if (timeout > 0) {
571
/* GetQueuedCompletionStatus can occasionally return a little early.
572
* Make sure that the desired timeout target time is reached.
573
*/
574
uv_update_time(loop);
575
if (timeout_time > loop->time) {
576
timeout = (DWORD)(timeout_time - loop->time);
577
/* The first call to GetQueuedCompletionStatus should return very
578
* close to the target time and the second should reach it, but
579
* this is not stated in the documentation. To make sure a busy
580
* loop cannot happen, the timeout is increased exponentially
581
* starting on the third round.
582
*/
583
timeout += repeat ? (1 << (repeat - 1)) : 0;
584
continue;
585
}
586
}
587
break;
588
}
589
}
590
591
592
int uv_run(uv_loop_t *loop, uv_run_mode mode) {
593
DWORD timeout;
594
int r;
595
int can_sleep;
596
597
r = uv__loop_alive(loop);
598
if (!r)
599
uv_update_time(loop);
600
601
while (r != 0 && loop->stop_flag == 0) {
602
uv_update_time(loop);
603
uv__run_timers(loop);
604
605
can_sleep = loop->pending_reqs_tail == NULL && loop->idle_handles == NULL;
606
607
uv__process_reqs(loop);
608
uv__idle_invoke(loop);
609
uv__prepare_invoke(loop);
610
611
timeout = 0;
612
if ((mode == UV_RUN_ONCE && can_sleep) || mode == UV_RUN_DEFAULT)
613
timeout = uv_backend_timeout(loop);
614
615
if (pGetQueuedCompletionStatusEx)
616
uv__poll(loop, timeout);
617
else
618
uv__poll_wine(loop, timeout);
619
620
/* Process immediate callbacks (e.g. write_cb) a small fixed number of
621
* times to avoid loop starvation.*/
622
for (r = 0; r < 8 && loop->pending_reqs_tail != NULL; r++)
623
uv__process_reqs(loop);
624
625
/* Run one final update on the provider_idle_time in case uv__poll*
626
* returned because the timeout expired, but no events were received. This
627
* call will be ignored if the provider_entry_time was either never set (if
628
* the timeout == 0) or was already updated b/c an event was received.
629
*/
630
uv__metrics_update_idle_time(loop);
631
632
uv__check_invoke(loop);
633
uv__process_endgames(loop);
634
635
if (mode == UV_RUN_ONCE) {
636
/* UV_RUN_ONCE implies forward progress: at least one callback must have
637
* been invoked when it returns. uv__io_poll() can return without doing
638
* I/O (meaning: no callbacks) when its timeout expires - which means we
639
* have pending timers that satisfy the forward progress constraint.
640
*
641
* UV_RUN_NOWAIT makes no guarantees about progress so it's omitted from
642
* the check.
643
*/
644
uv_update_time(loop);
645
uv__run_timers(loop);
646
}
647
648
r = uv__loop_alive(loop);
649
if (mode == UV_RUN_ONCE || mode == UV_RUN_NOWAIT)
650
break;
651
}
652
653
/* The if statement lets the compiler compile it to a conditional store.
654
* Avoids dirtying a cache line.
655
*/
656
if (loop->stop_flag != 0)
657
loop->stop_flag = 0;
658
659
return r;
660
}
661
662
663
int uv_fileno(const uv_handle_t* handle, uv_os_fd_t* fd) {
664
uv_os_fd_t fd_out;
665
666
switch (handle->type) {
667
case UV_TCP:
668
fd_out = (uv_os_fd_t)((uv_tcp_t*) handle)->socket;
669
break;
670
671
case UV_NAMED_PIPE:
672
fd_out = ((uv_pipe_t*) handle)->handle;
673
break;
674
675
case UV_TTY:
676
fd_out = ((uv_tty_t*) handle)->handle;
677
break;
678
679
case UV_UDP:
680
fd_out = (uv_os_fd_t)((uv_udp_t*) handle)->socket;
681
break;
682
683
case UV_POLL:
684
fd_out = (uv_os_fd_t)((uv_poll_t*) handle)->socket;
685
break;
686
687
default:
688
return UV_EINVAL;
689
}
690
691
if (uv_is_closing(handle) || fd_out == INVALID_HANDLE_VALUE)
692
return UV_EBADF;
693
694
*fd = fd_out;
695
return 0;
696
}
697
698
699
int uv__socket_sockopt(uv_handle_t* handle, int optname, int* value) {
700
int r;
701
int len;
702
SOCKET socket;
703
704
if (handle == NULL || value == NULL)
705
return UV_EINVAL;
706
707
if (handle->type == UV_TCP)
708
socket = ((uv_tcp_t*) handle)->socket;
709
else if (handle->type == UV_UDP)
710
socket = ((uv_udp_t*) handle)->socket;
711
else
712
return UV_ENOTSUP;
713
714
len = sizeof(*value);
715
716
if (*value == 0)
717
r = getsockopt(socket, SOL_SOCKET, optname, (char*) value, &len);
718
else
719
r = setsockopt(socket, SOL_SOCKET, optname, (const char*) value, len);
720
721
if (r == SOCKET_ERROR)
722
return uv_translate_sys_error(WSAGetLastError());
723
724
return 0;
725
}
726
727
int uv_cpumask_size(void) {
728
return (int)(sizeof(DWORD_PTR) * 8);
729
}
730
731
int uv__getsockpeername(const uv_handle_t* handle,
732
uv__peersockfunc func,
733
struct sockaddr* name,
734
int* namelen,
735
int delayed_error) {
736
737
int result;
738
uv_os_fd_t fd;
739
740
result = uv_fileno(handle, &fd);
741
if (result != 0)
742
return result;
743
744
if (delayed_error)
745
return uv_translate_sys_error(delayed_error);
746
747
result = func((SOCKET) fd, name, namelen);
748
if (result != 0)
749
return uv_translate_sys_error(WSAGetLastError());
750
751
return 0;
752
}
753
754