xref: /libuv/src/win/core.c (revision 6af08fb5)
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 
uv__crt_dbg_report_handler(int report_type,char * message,int * ret_val)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
uv__crt_invalid_parameter_handler(const wchar_t * expression,const wchar_t * function,const wchar_t * file,unsigned int line,uintptr_t reserved)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 
uv__loops_init(void)88 static void uv__loops_init(void) {
89   uv_mutex_init(&uv__loops_lock);
90 }
91 
92 
uv__loops_add(uv_loop_t * loop)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 
uv__loops_remove(uv_loop_t * loop)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 
uv__wake_all_loops(void)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 
uv__init(void)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 
uv_loop_init(uv_loop_t * loop)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   memset(&lfields->loop_metrics.metrics,
249          0,
250          sizeof(lfields->loop_metrics.metrics));
251 
252   /* To prevent uninitialized memory access, loop->time must be initialized
253    * to zero before calling uv_update_time for the first time.
254    */
255   loop->time = 0;
256   uv_update_time(loop);
257 
258   uv__queue_init(&loop->wq);
259   uv__queue_init(&loop->handle_queue);
260   loop->active_reqs.count = 0;
261   loop->active_handles = 0;
262 
263   loop->pending_reqs_tail = NULL;
264 
265   loop->endgame_handles = NULL;
266 
267   loop->timer_heap = timer_heap = uv__malloc(sizeof(*timer_heap));
268   if (timer_heap == NULL) {
269     err = UV_ENOMEM;
270     goto fail_timers_alloc;
271   }
272 
273   heap_init(timer_heap);
274 
275   loop->check_handles = NULL;
276   loop->prepare_handles = NULL;
277   loop->idle_handles = NULL;
278 
279   loop->next_prepare_handle = NULL;
280   loop->next_check_handle = NULL;
281   loop->next_idle_handle = NULL;
282 
283   memset(&loop->poll_peer_sockets, 0, sizeof loop->poll_peer_sockets);
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 
uv_update_time(uv_loop_t * loop)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 
uv__once_init(void)332 void uv__once_init(void) {
333   uv_once(&uv_init_guard_, uv__init);
334 }
335 
336 
uv__loop_close(uv_loop_t * loop)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(uv__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 
uv__loop_configure(uv_loop_t * loop,uv_loop_option option,va_list ap)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 
uv_backend_fd(const uv_loop_t * loop)391 int uv_backend_fd(const uv_loop_t* loop) {
392   return -1;
393 }
394 
395 
uv_loop_fork(uv_loop_t * loop)396 int uv_loop_fork(uv_loop_t* loop) {
397   return UV_ENOSYS;
398 }
399 
400 
uv__loop_alive(const uv_loop_t * loop)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 
uv_loop_alive(const uv_loop_t * loop)409 int uv_loop_alive(const uv_loop_t* loop) {
410   return uv__loop_alive(loop);
411 }
412 
413 
uv_backend_timeout(const uv_loop_t * loop)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 
uv__poll(uv_loop_t * loop,DWORD timeout)426 static void uv__poll(uv_loop_t* loop, DWORD timeout) {
427   uv__loop_internal_fields_t* lfields;
428   BOOL success;
429   uv_req_t* req;
430   OVERLAPPED_ENTRY overlappeds[128];
431   ULONG count;
432   ULONG i;
433   int repeat;
434   uint64_t timeout_time;
435   uint64_t user_timeout;
436   uint64_t actual_timeout;
437   int reset_timeout;
438 
439   lfields = uv__get_internal_fields(loop);
440   timeout_time = loop->time + timeout;
441 
442   if (lfields->flags & UV_METRICS_IDLE_TIME) {
443     reset_timeout = 1;
444     user_timeout = timeout;
445     timeout = 0;
446   } else {
447     reset_timeout = 0;
448   }
449 
450   for (repeat = 0; ; repeat++) {
451     actual_timeout = timeout;
452 
453     /* Only need to set the provider_entry_time if timeout != 0. The function
454      * will return early if the loop isn't configured with UV_METRICS_IDLE_TIME.
455      */
456     if (timeout != 0)
457       uv__metrics_set_provider_entry_time(loop);
458 
459     /* Store the current timeout in a location that's globally accessible so
460      * other locations like uv__work_done() can determine whether the queue
461      * of events in the callback were waiting when poll was called.
462      */
463     lfields->current_timeout = timeout;
464 
465     success = GetQueuedCompletionStatusEx(loop->iocp,
466                                           overlappeds,
467                                           ARRAY_SIZE(overlappeds),
468                                           &count,
469                                           timeout,
470                                           FALSE);
471 
472     if (reset_timeout != 0) {
473       timeout = user_timeout;
474       reset_timeout = 0;
475     }
476 
477     /* Placed here because on success the loop will break whether there is an
478      * empty package or not, or if GetQueuedCompletionStatusEx returned early
479      * then the timeout will be updated and the loop will run again. In either
480      * case the idle time will need to be updated.
481      */
482     uv__metrics_update_idle_time(loop);
483 
484     if (success) {
485       for (i = 0; i < count; i++) {
486         /* Package was dequeued, but see if it is not a empty package
487          * meant only to wake us up.
488          */
489         if (overlappeds[i].lpOverlapped) {
490           uv__metrics_inc_events(loop, 1);
491           if (actual_timeout == 0)
492             uv__metrics_inc_events_waiting(loop, 1);
493 
494           req = uv__overlapped_to_req(overlappeds[i].lpOverlapped);
495           uv__insert_pending_req(loop, req);
496         }
497       }
498 
499       /* Some time might have passed waiting for I/O,
500        * so update the loop time here.
501        */
502       uv_update_time(loop);
503     } else if (GetLastError() != WAIT_TIMEOUT) {
504       /* Serious error */
505       uv_fatal_error(GetLastError(), "GetQueuedCompletionStatusEx");
506     } else if (timeout > 0) {
507       /* GetQueuedCompletionStatus can occasionally return a little early.
508        * Make sure that the desired timeout target time is reached.
509        */
510       uv_update_time(loop);
511       if (timeout_time > loop->time) {
512         timeout = (DWORD)(timeout_time - loop->time);
513         /* The first call to GetQueuedCompletionStatus should return very
514          * close to the target time and the second should reach it, but
515          * this is not stated in the documentation. To make sure a busy
516          * loop cannot happen, the timeout is increased exponentially
517          * starting on the third round.
518          */
519         timeout += repeat ? (1 << (repeat - 1)) : 0;
520         continue;
521       }
522     }
523     break;
524   }
525 }
526 
527 
uv_run(uv_loop_t * loop,uv_run_mode mode)528 int uv_run(uv_loop_t *loop, uv_run_mode mode) {
529   DWORD timeout;
530   int r;
531   int can_sleep;
532 
533   r = uv__loop_alive(loop);
534   if (!r)
535     uv_update_time(loop);
536 
537   /* Maintain backwards compatibility by processing timers before entering the
538    * while loop for UV_RUN_DEFAULT. Otherwise timers only need to be executed
539    * once, which should be done after polling in order to maintain proper
540    * execution order of the conceptual event loop. */
541   if (mode == UV_RUN_DEFAULT && r != 0 && loop->stop_flag == 0) {
542     uv_update_time(loop);
543     uv__run_timers(loop);
544   }
545 
546   while (r != 0 && loop->stop_flag == 0) {
547     can_sleep = loop->pending_reqs_tail == NULL && loop->idle_handles == NULL;
548 
549     uv__process_reqs(loop);
550     uv__idle_invoke(loop);
551     uv__prepare_invoke(loop);
552 
553     timeout = 0;
554     if ((mode == UV_RUN_ONCE && can_sleep) || mode == UV_RUN_DEFAULT)
555       timeout = uv_backend_timeout(loop);
556 
557     uv__metrics_inc_loop_count(loop);
558 
559     uv__poll(loop, timeout);
560 
561     /* Process immediate callbacks (e.g. write_cb) a small fixed number of
562      * times to avoid loop starvation.*/
563     for (r = 0; r < 8 && loop->pending_reqs_tail != NULL; r++)
564       uv__process_reqs(loop);
565 
566     /* Run one final update on the provider_idle_time in case uv__poll*
567      * returned because the timeout expired, but no events were received. This
568      * call will be ignored if the provider_entry_time was either never set (if
569      * the timeout == 0) or was already updated b/c an event was received.
570      */
571     uv__metrics_update_idle_time(loop);
572 
573     uv__check_invoke(loop);
574     uv__process_endgames(loop);
575 
576     uv_update_time(loop);
577     uv__run_timers(loop);
578 
579     r = uv__loop_alive(loop);
580     if (mode == UV_RUN_ONCE || mode == UV_RUN_NOWAIT)
581       break;
582   }
583 
584   /* The if statement lets the compiler compile it to a conditional store.
585    * Avoids dirtying a cache line.
586    */
587   if (loop->stop_flag != 0)
588     loop->stop_flag = 0;
589 
590   return r;
591 }
592 
593 
uv_fileno(const uv_handle_t * handle,uv_os_fd_t * fd)594 int uv_fileno(const uv_handle_t* handle, uv_os_fd_t* fd) {
595   uv_os_fd_t fd_out;
596 
597   switch (handle->type) {
598   case UV_TCP:
599     fd_out = (uv_os_fd_t)((uv_tcp_t*) handle)->socket;
600     break;
601 
602   case UV_NAMED_PIPE:
603     fd_out = ((uv_pipe_t*) handle)->handle;
604     break;
605 
606   case UV_TTY:
607     fd_out = ((uv_tty_t*) handle)->handle;
608     break;
609 
610   case UV_UDP:
611     fd_out = (uv_os_fd_t)((uv_udp_t*) handle)->socket;
612     break;
613 
614   case UV_POLL:
615     fd_out = (uv_os_fd_t)((uv_poll_t*) handle)->socket;
616     break;
617 
618   default:
619     return UV_EINVAL;
620   }
621 
622   if (uv_is_closing(handle) || fd_out == INVALID_HANDLE_VALUE)
623     return UV_EBADF;
624 
625   *fd = fd_out;
626   return 0;
627 }
628 
629 
uv__socket_sockopt(uv_handle_t * handle,int optname,int * value)630 int uv__socket_sockopt(uv_handle_t* handle, int optname, int* value) {
631   int r;
632   int len;
633   SOCKET socket;
634 
635   if (handle == NULL || value == NULL)
636     return UV_EINVAL;
637 
638   if (handle->type == UV_TCP)
639     socket = ((uv_tcp_t*) handle)->socket;
640   else if (handle->type == UV_UDP)
641     socket = ((uv_udp_t*) handle)->socket;
642   else
643     return UV_ENOTSUP;
644 
645   len = sizeof(*value);
646 
647   if (*value == 0)
648     r = getsockopt(socket, SOL_SOCKET, optname, (char*) value, &len);
649   else
650     r = setsockopt(socket, SOL_SOCKET, optname, (const char*) value, len);
651 
652   if (r == SOCKET_ERROR)
653     return uv_translate_sys_error(WSAGetLastError());
654 
655   return 0;
656 }
657 
uv_cpumask_size(void)658 int uv_cpumask_size(void) {
659   return (int)(sizeof(DWORD_PTR) * 8);
660 }
661 
uv__getsockpeername(const uv_handle_t * handle,uv__peersockfunc func,struct sockaddr * name,int * namelen,int delayed_error)662 int uv__getsockpeername(const uv_handle_t* handle,
663                         uv__peersockfunc func,
664                         struct sockaddr* name,
665                         int* namelen,
666                         int delayed_error) {
667 
668   int result;
669   uv_os_fd_t fd;
670 
671   result = uv_fileno(handle, &fd);
672   if (result != 0)
673     return result;
674 
675   if (delayed_error)
676     return uv_translate_sys_error(delayed_error);
677 
678   result = func((SOCKET) fd, name, namelen);
679   if (result != 0)
680     return uv_translate_sys_error(WSAGetLastError());
681 
682   return 0;
683 }
684