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 <stdlib.h>
24
25 #include "uv.h"
26 #include "internal.h"
27 #include "handle-inl.h"
28 #include "stream-inl.h"
29 #include "req-inl.h"
30
31
32 /*
33 * Number of simultaneous pending AcceptEx calls.
34 */
35 const unsigned int uv_simultaneous_server_accepts = 32;
36
37 /* A zero-size buffer for use by uv_tcp_read */
38 static char uv_zero_[] = "";
39
uv__tcp_nodelay(uv_tcp_t * handle,SOCKET socket,int enable)40 static int uv__tcp_nodelay(uv_tcp_t* handle, SOCKET socket, int enable) {
41 if (setsockopt(socket,
42 IPPROTO_TCP,
43 TCP_NODELAY,
44 (const char*)&enable,
45 sizeof enable) == -1) {
46 return WSAGetLastError();
47 }
48 return 0;
49 }
50
51
uv__tcp_keepalive(uv_tcp_t * handle,SOCKET socket,int enable,unsigned int delay)52 static int uv__tcp_keepalive(uv_tcp_t* handle, SOCKET socket, int enable, unsigned int delay) {
53 if (setsockopt(socket,
54 SOL_SOCKET,
55 SO_KEEPALIVE,
56 (const char*)&enable,
57 sizeof enable) == -1) {
58 return WSAGetLastError();
59 }
60
61 if (!enable)
62 return 0;
63
64 if (delay < 1)
65 return UV_EINVAL;
66
67 if (setsockopt(socket,
68 IPPROTO_TCP,
69 TCP_KEEPALIVE,
70 (const char*)&delay,
71 sizeof delay) == -1) {
72 return WSAGetLastError();
73 }
74
75 return 0;
76 }
77
78
uv__tcp_set_socket(uv_loop_t * loop,uv_tcp_t * handle,SOCKET socket,int family,int imported)79 static int uv__tcp_set_socket(uv_loop_t* loop,
80 uv_tcp_t* handle,
81 SOCKET socket,
82 int family,
83 int imported) {
84 DWORD yes = 1;
85 int non_ifs_lsp;
86 int err;
87
88 if (handle->socket != INVALID_SOCKET)
89 return UV_EBUSY;
90
91 /* Set the socket to nonblocking mode */
92 if (ioctlsocket(socket, FIONBIO, &yes) == SOCKET_ERROR) {
93 return WSAGetLastError();
94 }
95
96 /* Make the socket non-inheritable */
97 if (!SetHandleInformation((HANDLE) socket, HANDLE_FLAG_INHERIT, 0))
98 return GetLastError();
99
100 /* Associate it with the I/O completion port. Use uv_handle_t pointer as
101 * completion key. */
102 if (CreateIoCompletionPort((HANDLE)socket,
103 loop->iocp,
104 (ULONG_PTR)socket,
105 0) == NULL) {
106 if (imported) {
107 handle->flags |= UV_HANDLE_EMULATE_IOCP;
108 } else {
109 return GetLastError();
110 }
111 }
112
113 if (family == AF_INET6) {
114 non_ifs_lsp = uv_tcp_non_ifs_lsp_ipv6;
115 } else {
116 non_ifs_lsp = uv_tcp_non_ifs_lsp_ipv4;
117 }
118
119 if (!(handle->flags & UV_HANDLE_EMULATE_IOCP) && !non_ifs_lsp) {
120 UCHAR sfcnm_flags =
121 FILE_SKIP_SET_EVENT_ON_HANDLE | FILE_SKIP_COMPLETION_PORT_ON_SUCCESS;
122 if (!SetFileCompletionNotificationModes((HANDLE) socket, sfcnm_flags))
123 return GetLastError();
124 handle->flags |= UV_HANDLE_SYNC_BYPASS_IOCP;
125 }
126
127 if (handle->flags & UV_HANDLE_TCP_NODELAY) {
128 err = uv__tcp_nodelay(handle, socket, 1);
129 if (err)
130 return err;
131 }
132
133 /* TODO: Use stored delay. */
134 if (handle->flags & UV_HANDLE_TCP_KEEPALIVE) {
135 err = uv__tcp_keepalive(handle, socket, 1, 60);
136 if (err)
137 return err;
138 }
139
140 handle->socket = socket;
141
142 if (family == AF_INET6) {
143 handle->flags |= UV_HANDLE_IPV6;
144 } else {
145 assert(!(handle->flags & UV_HANDLE_IPV6));
146 }
147
148 return 0;
149 }
150
151
uv_tcp_init_ex(uv_loop_t * loop,uv_tcp_t * handle,unsigned int flags)152 int uv_tcp_init_ex(uv_loop_t* loop, uv_tcp_t* handle, unsigned int flags) {
153 int domain;
154
155 /* Use the lower 8 bits for the domain */
156 domain = flags & 0xFF;
157 if (domain != AF_INET && domain != AF_INET6 && domain != AF_UNSPEC)
158 return UV_EINVAL;
159
160 if (flags & ~0xFF)
161 return UV_EINVAL;
162
163 uv__stream_init(loop, (uv_stream_t*) handle, UV_TCP);
164 handle->tcp.serv.accept_reqs = NULL;
165 handle->tcp.serv.pending_accepts = NULL;
166 handle->socket = INVALID_SOCKET;
167 handle->reqs_pending = 0;
168 handle->tcp.serv.func_acceptex = NULL;
169 handle->tcp.conn.func_connectex = NULL;
170 handle->tcp.serv.processed_accepts = 0;
171 handle->delayed_error = 0;
172
173 /* If anything fails beyond this point we need to remove the handle from
174 * the handle queue, since it was added by uv__handle_init in uv__stream_init.
175 */
176
177 if (domain != AF_UNSPEC) {
178 SOCKET sock;
179 DWORD err;
180
181 sock = socket(domain, SOCK_STREAM, 0);
182 if (sock == INVALID_SOCKET) {
183 err = WSAGetLastError();
184 uv__queue_remove(&handle->handle_queue);
185 return uv_translate_sys_error(err);
186 }
187
188 err = uv__tcp_set_socket(handle->loop, handle, sock, domain, 0);
189 if (err) {
190 closesocket(sock);
191 uv__queue_remove(&handle->handle_queue);
192 return uv_translate_sys_error(err);
193 }
194
195 }
196
197 return 0;
198 }
199
200
uv_tcp_init(uv_loop_t * loop,uv_tcp_t * handle)201 int uv_tcp_init(uv_loop_t* loop, uv_tcp_t* handle) {
202 return uv_tcp_init_ex(loop, handle, AF_UNSPEC);
203 }
204
205
uv__process_tcp_shutdown_req(uv_loop_t * loop,uv_tcp_t * stream,uv_shutdown_t * req)206 void uv__process_tcp_shutdown_req(uv_loop_t* loop, uv_tcp_t* stream, uv_shutdown_t *req) {
207 int err;
208
209 assert(req);
210 assert(stream->stream.conn.write_reqs_pending == 0);
211 assert(!(stream->flags & UV_HANDLE_SHUT));
212 assert(stream->flags & UV_HANDLE_CONNECTION);
213
214 stream->stream.conn.shutdown_req = NULL;
215 UNREGISTER_HANDLE_REQ(loop, stream);
216
217 err = 0;
218 if (stream->flags & UV_HANDLE_CLOSING)
219 /* The user destroyed the stream before we got to do the shutdown. */
220 err = UV_ECANCELED;
221 else if (shutdown(stream->socket, SD_SEND) == SOCKET_ERROR)
222 err = uv_translate_sys_error(WSAGetLastError());
223 else /* Success. */
224 stream->flags |= UV_HANDLE_SHUT;
225
226 if (req->cb)
227 req->cb(req, err);
228
229 DECREASE_PENDING_REQ_COUNT(stream);
230 }
231
232
uv__tcp_endgame(uv_loop_t * loop,uv_tcp_t * handle)233 void uv__tcp_endgame(uv_loop_t* loop, uv_tcp_t* handle) {
234 unsigned int i;
235 uv_tcp_accept_t* req;
236
237 assert(handle->flags & UV_HANDLE_CLOSING);
238 assert(handle->reqs_pending == 0);
239 assert(!(handle->flags & UV_HANDLE_CLOSED));
240 assert(handle->socket == INVALID_SOCKET);
241
242 if (!(handle->flags & UV_HANDLE_CONNECTION) && handle->tcp.serv.accept_reqs) {
243 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
244 for (i = 0; i < uv_simultaneous_server_accepts; i++) {
245 req = &handle->tcp.serv.accept_reqs[i];
246 if (req->wait_handle != INVALID_HANDLE_VALUE) {
247 UnregisterWait(req->wait_handle);
248 req->wait_handle = INVALID_HANDLE_VALUE;
249 }
250 if (req->event_handle != NULL) {
251 CloseHandle(req->event_handle);
252 req->event_handle = NULL;
253 }
254 }
255 }
256
257 uv__free(handle->tcp.serv.accept_reqs);
258 handle->tcp.serv.accept_reqs = NULL;
259 }
260
261 if (handle->flags & UV_HANDLE_CONNECTION &&
262 handle->flags & UV_HANDLE_EMULATE_IOCP) {
263 if (handle->read_req.wait_handle != INVALID_HANDLE_VALUE) {
264 UnregisterWait(handle->read_req.wait_handle);
265 handle->read_req.wait_handle = INVALID_HANDLE_VALUE;
266 }
267 if (handle->read_req.event_handle != NULL) {
268 CloseHandle(handle->read_req.event_handle);
269 handle->read_req.event_handle = NULL;
270 }
271 }
272
273 uv__handle_close(handle);
274 }
275
276
277 /* Unlike on Unix, here we don't set SO_REUSEADDR, because it doesn't just
278 * allow binding to addresses that are in use by sockets in TIME_WAIT, it
279 * effectively allows 'stealing' a port which is in use by another application.
280 *
281 * SO_EXCLUSIVEADDRUSE is also not good here because it does check all sockets,
282 * regardless of state, so we'd get an error even if the port is in use by a
283 * socket in TIME_WAIT state.
284 *
285 * See issue #1360.
286 *
287 */
uv__tcp_try_bind(uv_tcp_t * handle,const struct sockaddr * addr,unsigned int addrlen,unsigned int flags)288 static int uv__tcp_try_bind(uv_tcp_t* handle,
289 const struct sockaddr* addr,
290 unsigned int addrlen,
291 unsigned int flags) {
292 DWORD err;
293 int r;
294
295 /* There is no SO_REUSEPORT on Windows, Windows only knows SO_REUSEADDR.
296 * so we just return an error directly when UV_TCP_REUSEPORT is requested
297 * for binding the socket. */
298 if (flags & UV_TCP_REUSEPORT)
299 return ERROR_NOT_SUPPORTED;
300
301 if (handle->socket == INVALID_SOCKET) {
302 SOCKET sock;
303
304 /* Cannot set IPv6-only mode on non-IPv6 socket. */
305 if ((flags & UV_TCP_IPV6ONLY) && addr->sa_family != AF_INET6)
306 return ERROR_INVALID_PARAMETER;
307
308 sock = socket(addr->sa_family, SOCK_STREAM, 0);
309 if (sock == INVALID_SOCKET) {
310 return WSAGetLastError();
311 }
312
313 err = uv__tcp_set_socket(handle->loop, handle, sock, addr->sa_family, 0);
314 if (err) {
315 closesocket(sock);
316 return err;
317 }
318 }
319
320 #ifdef IPV6_V6ONLY
321 if (addr->sa_family == AF_INET6) {
322 int on;
323
324 on = (flags & UV_TCP_IPV6ONLY) != 0;
325
326 /* TODO: how to handle errors? This may fail if there is no ipv4 stack
327 * available, or when run on XP/2003 which have no support for dualstack
328 * sockets. For now we're silently ignoring the error. */
329 setsockopt(handle->socket,
330 IPPROTO_IPV6,
331 IPV6_V6ONLY,
332 (const char*)&on,
333 sizeof on);
334 }
335 #endif
336
337 r = bind(handle->socket, addr, addrlen);
338
339 if (r == SOCKET_ERROR) {
340 err = WSAGetLastError();
341 if (err == WSAEADDRINUSE) {
342 /* Some errors are not to be reported until connect() or listen() */
343 handle->delayed_error = err;
344 } else {
345 return err;
346 }
347 }
348
349 handle->flags |= UV_HANDLE_BOUND;
350
351 return 0;
352 }
353
354
post_completion(void * context,BOOLEAN timed_out)355 static void CALLBACK post_completion(void* context, BOOLEAN timed_out) {
356 uv_req_t* req;
357 uv_tcp_t* handle;
358
359 req = (uv_req_t*) context;
360 assert(req != NULL);
361 handle = (uv_tcp_t*)req->data;
362 assert(handle != NULL);
363 assert(!timed_out);
364
365 if (!PostQueuedCompletionStatus(handle->loop->iocp,
366 req->u.io.overlapped.InternalHigh,
367 0,
368 &req->u.io.overlapped)) {
369 uv_fatal_error(GetLastError(), "PostQueuedCompletionStatus");
370 }
371 }
372
373
post_write_completion(void * context,BOOLEAN timed_out)374 static void CALLBACK post_write_completion(void* context, BOOLEAN timed_out) {
375 uv_write_t* req;
376 uv_tcp_t* handle;
377
378 req = (uv_write_t*) context;
379 assert(req != NULL);
380 handle = (uv_tcp_t*)req->handle;
381 assert(handle != NULL);
382 assert(!timed_out);
383
384 if (!PostQueuedCompletionStatus(handle->loop->iocp,
385 req->u.io.overlapped.InternalHigh,
386 0,
387 &req->u.io.overlapped)) {
388 uv_fatal_error(GetLastError(), "PostQueuedCompletionStatus");
389 }
390 }
391
392
uv__tcp_queue_accept(uv_tcp_t * handle,uv_tcp_accept_t * req)393 static void uv__tcp_queue_accept(uv_tcp_t* handle, uv_tcp_accept_t* req) {
394 uv_loop_t* loop = handle->loop;
395 BOOL success;
396 DWORD bytes;
397 SOCKET accept_socket;
398 short family;
399
400 assert(handle->flags & UV_HANDLE_LISTENING);
401 assert(req->accept_socket == INVALID_SOCKET);
402
403 /* choose family and extension function */
404 if (handle->flags & UV_HANDLE_IPV6) {
405 family = AF_INET6;
406 } else {
407 family = AF_INET;
408 }
409
410 /* Open a socket for the accepted connection. */
411 accept_socket = socket(family, SOCK_STREAM, 0);
412 if (accept_socket == INVALID_SOCKET) {
413 SET_REQ_ERROR(req, WSAGetLastError());
414 uv__insert_pending_req(loop, (uv_req_t*)req);
415 handle->reqs_pending++;
416 return;
417 }
418
419 /* Make the socket non-inheritable */
420 if (!SetHandleInformation((HANDLE) accept_socket, HANDLE_FLAG_INHERIT, 0)) {
421 SET_REQ_ERROR(req, GetLastError());
422 uv__insert_pending_req(loop, (uv_req_t*)req);
423 handle->reqs_pending++;
424 closesocket(accept_socket);
425 return;
426 }
427
428 /* Prepare the overlapped structure. */
429 memset(&(req->u.io.overlapped), 0, sizeof(req->u.io.overlapped));
430 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
431 assert(req->event_handle != NULL);
432 req->u.io.overlapped.hEvent = (HANDLE) ((ULONG_PTR) req->event_handle | 1);
433 }
434
435 success = handle->tcp.serv.func_acceptex(handle->socket,
436 accept_socket,
437 (void*)req->accept_buffer,
438 0,
439 sizeof(struct sockaddr_storage),
440 sizeof(struct sockaddr_storage),
441 &bytes,
442 &req->u.io.overlapped);
443
444 if (UV_SUCCEEDED_WITHOUT_IOCP(success)) {
445 /* Process the req without IOCP. */
446 req->accept_socket = accept_socket;
447 handle->reqs_pending++;
448 uv__insert_pending_req(loop, (uv_req_t*)req);
449 } else if (UV_SUCCEEDED_WITH_IOCP(success)) {
450 /* The req will be processed with IOCP. */
451 req->accept_socket = accept_socket;
452 handle->reqs_pending++;
453 if (handle->flags & UV_HANDLE_EMULATE_IOCP &&
454 req->wait_handle == INVALID_HANDLE_VALUE &&
455 !RegisterWaitForSingleObject(&req->wait_handle,
456 req->event_handle, post_completion, (void*) req,
457 INFINITE, WT_EXECUTEINWAITTHREAD)) {
458 SET_REQ_ERROR(req, GetLastError());
459 uv__insert_pending_req(loop, (uv_req_t*)req);
460 }
461 } else {
462 /* Make this req pending reporting an error. */
463 SET_REQ_ERROR(req, WSAGetLastError());
464 uv__insert_pending_req(loop, (uv_req_t*)req);
465 handle->reqs_pending++;
466 /* Destroy the preallocated client socket. */
467 closesocket(accept_socket);
468 /* Destroy the event handle */
469 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
470 CloseHandle(req->event_handle);
471 req->event_handle = NULL;
472 }
473 }
474 }
475
476
uv__tcp_queue_read(uv_loop_t * loop,uv_tcp_t * handle)477 static void uv__tcp_queue_read(uv_loop_t* loop, uv_tcp_t* handle) {
478 uv_read_t* req;
479 uv_buf_t buf;
480 int result;
481 DWORD bytes, flags;
482
483 assert(handle->flags & UV_HANDLE_READING);
484 assert(!(handle->flags & UV_HANDLE_READ_PENDING));
485
486 req = &handle->read_req;
487 memset(&req->u.io.overlapped, 0, sizeof(req->u.io.overlapped));
488
489 handle->flags |= UV_HANDLE_ZERO_READ;
490 buf.base = (char*) &uv_zero_;
491 buf.len = 0;
492
493 /* Prepare the overlapped structure. */
494 memset(&(req->u.io.overlapped), 0, sizeof(req->u.io.overlapped));
495 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
496 assert(req->event_handle != NULL);
497 req->u.io.overlapped.hEvent = (HANDLE) ((ULONG_PTR) req->event_handle | 1);
498 }
499
500 flags = 0;
501 result = WSARecv(handle->socket,
502 (WSABUF*)&buf,
503 1,
504 &bytes,
505 &flags,
506 &req->u.io.overlapped,
507 NULL);
508
509 handle->flags |= UV_HANDLE_READ_PENDING;
510 handle->reqs_pending++;
511
512 if (UV_SUCCEEDED_WITHOUT_IOCP(result == 0)) {
513 /* Process the req without IOCP. */
514 req->u.io.overlapped.InternalHigh = bytes;
515 uv__insert_pending_req(loop, (uv_req_t*)req);
516 } else if (UV_SUCCEEDED_WITH_IOCP(result == 0)) {
517 /* The req will be processed with IOCP. */
518 if (handle->flags & UV_HANDLE_EMULATE_IOCP &&
519 req->wait_handle == INVALID_HANDLE_VALUE &&
520 !RegisterWaitForSingleObject(&req->wait_handle,
521 req->event_handle, post_completion, (void*) req,
522 INFINITE, WT_EXECUTEINWAITTHREAD)) {
523 SET_REQ_ERROR(req, GetLastError());
524 uv__insert_pending_req(loop, (uv_req_t*)req);
525 }
526 } else {
527 /* Make this req pending reporting an error. */
528 SET_REQ_ERROR(req, WSAGetLastError());
529 uv__insert_pending_req(loop, (uv_req_t*)req);
530 }
531 }
532
533
uv_tcp_close_reset(uv_tcp_t * handle,uv_close_cb close_cb)534 int uv_tcp_close_reset(uv_tcp_t* handle, uv_close_cb close_cb) {
535 struct linger l = { 1, 0 };
536
537 /* Disallow setting SO_LINGER to zero due to some platform inconsistencies */
538 if (uv__is_stream_shutting(handle))
539 return UV_EINVAL;
540
541 if (0 != setsockopt(handle->socket, SOL_SOCKET, SO_LINGER, (const char*)&l, sizeof(l)))
542 return uv_translate_sys_error(WSAGetLastError());
543
544 uv_close((uv_handle_t*) handle, close_cb);
545 return 0;
546 }
547
548
uv__tcp_listen(uv_tcp_t * handle,int backlog,uv_connection_cb cb)549 int uv__tcp_listen(uv_tcp_t* handle, int backlog, uv_connection_cb cb) {
550 unsigned int i, simultaneous_accepts;
551 uv_tcp_accept_t* req;
552 int err;
553
554 assert(backlog > 0);
555
556 if (handle->flags & UV_HANDLE_LISTENING) {
557 handle->stream.serv.connection_cb = cb;
558 }
559
560 if (handle->flags & UV_HANDLE_READING) {
561 return WSAEISCONN;
562 }
563
564 if (handle->delayed_error) {
565 return handle->delayed_error;
566 }
567
568 if (!(handle->flags & UV_HANDLE_BOUND)) {
569 err = uv__tcp_try_bind(handle,
570 (const struct sockaddr*) &uv_addr_ip4_any_,
571 sizeof(uv_addr_ip4_any_),
572 0);
573 if (err)
574 return err;
575 if (handle->delayed_error)
576 return handle->delayed_error;
577 }
578
579 if (!handle->tcp.serv.func_acceptex) {
580 if (!uv__get_acceptex_function(handle->socket, &handle->tcp.serv.func_acceptex)) {
581 return WSAEAFNOSUPPORT;
582 }
583 }
584
585 /* If this flag is set, we already made this listen call in xfer. */
586 if (!(handle->flags & UV_HANDLE_SHARED_TCP_SOCKET) &&
587 listen(handle->socket, backlog) == SOCKET_ERROR) {
588 return WSAGetLastError();
589 }
590
591 handle->flags |= UV_HANDLE_LISTENING;
592 handle->stream.serv.connection_cb = cb;
593 INCREASE_ACTIVE_COUNT(loop, handle);
594
595 simultaneous_accepts = handle->flags & UV_HANDLE_TCP_SINGLE_ACCEPT ? 1
596 : uv_simultaneous_server_accepts;
597
598 if (handle->tcp.serv.accept_reqs == NULL) {
599 handle->tcp.serv.accept_reqs =
600 uv__malloc(uv_simultaneous_server_accepts * sizeof(uv_tcp_accept_t));
601 if (!handle->tcp.serv.accept_reqs) {
602 uv_fatal_error(ERROR_OUTOFMEMORY, "uv__malloc");
603 }
604
605 for (i = 0; i < simultaneous_accepts; i++) {
606 req = &handle->tcp.serv.accept_reqs[i];
607 UV_REQ_INIT(req, UV_ACCEPT);
608 req->accept_socket = INVALID_SOCKET;
609 req->data = handle;
610
611 req->wait_handle = INVALID_HANDLE_VALUE;
612 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
613 req->event_handle = CreateEvent(NULL, 0, 0, NULL);
614 if (req->event_handle == NULL) {
615 uv_fatal_error(GetLastError(), "CreateEvent");
616 }
617 } else {
618 req->event_handle = NULL;
619 }
620
621 uv__tcp_queue_accept(handle, req);
622 }
623
624 /* Initialize other unused requests too, because uv_tcp_endgame doesn't
625 * know how many requests were initialized, so it will try to clean up
626 * {uv_simultaneous_server_accepts} requests. */
627 for (i = simultaneous_accepts; i < uv_simultaneous_server_accepts; i++) {
628 req = &handle->tcp.serv.accept_reqs[i];
629 UV_REQ_INIT(req, UV_ACCEPT);
630 req->accept_socket = INVALID_SOCKET;
631 req->data = handle;
632 req->wait_handle = INVALID_HANDLE_VALUE;
633 req->event_handle = NULL;
634 }
635 }
636
637 return 0;
638 }
639
640
uv__tcp_accept(uv_tcp_t * server,uv_tcp_t * client)641 int uv__tcp_accept(uv_tcp_t* server, uv_tcp_t* client) {
642 int err = 0;
643 int family;
644
645 uv_tcp_accept_t* req = server->tcp.serv.pending_accepts;
646
647 if (!req) {
648 /* No valid connections found, so we error out. */
649 return WSAEWOULDBLOCK;
650 }
651
652 if (req->accept_socket == INVALID_SOCKET) {
653 return WSAENOTCONN;
654 }
655
656 if (server->flags & UV_HANDLE_IPV6) {
657 family = AF_INET6;
658 } else {
659 family = AF_INET;
660 }
661
662 err = uv__tcp_set_socket(client->loop,
663 client,
664 req->accept_socket,
665 family,
666 0);
667 if (err) {
668 closesocket(req->accept_socket);
669 } else {
670 uv__connection_init((uv_stream_t*) client);
671 /* AcceptEx() implicitly binds the accepted socket. */
672 client->flags |= UV_HANDLE_BOUND | UV_HANDLE_READABLE | UV_HANDLE_WRITABLE;
673 }
674
675 /* Prepare the req to pick up a new connection */
676 server->tcp.serv.pending_accepts = req->next_pending;
677 req->next_pending = NULL;
678 req->accept_socket = INVALID_SOCKET;
679
680 if (!(server->flags & UV_HANDLE_CLOSING)) {
681 /* Check if we're in a middle of changing the number of pending accepts. */
682 if (!(server->flags & UV_HANDLE_TCP_ACCEPT_STATE_CHANGING)) {
683 uv__tcp_queue_accept(server, req);
684 } else {
685 /* We better be switching to a single pending accept. */
686 assert(server->flags & UV_HANDLE_TCP_SINGLE_ACCEPT);
687
688 server->tcp.serv.processed_accepts++;
689
690 if (server->tcp.serv.processed_accepts >= uv_simultaneous_server_accepts) {
691 server->tcp.serv.processed_accepts = 0;
692 /*
693 * All previously queued accept requests are now processed.
694 * We now switch to queueing just a single accept.
695 */
696 uv__tcp_queue_accept(server, &server->tcp.serv.accept_reqs[0]);
697 server->flags &= ~UV_HANDLE_TCP_ACCEPT_STATE_CHANGING;
698 server->flags |= UV_HANDLE_TCP_SINGLE_ACCEPT;
699 }
700 }
701 }
702
703 return err;
704 }
705
706
uv__tcp_read_start(uv_tcp_t * handle,uv_alloc_cb alloc_cb,uv_read_cb read_cb)707 int uv__tcp_read_start(uv_tcp_t* handle, uv_alloc_cb alloc_cb,
708 uv_read_cb read_cb) {
709 uv_loop_t* loop = handle->loop;
710
711 handle->flags |= UV_HANDLE_READING;
712 handle->read_cb = read_cb;
713 handle->alloc_cb = alloc_cb;
714 INCREASE_ACTIVE_COUNT(loop, handle);
715
716 /* If reading was stopped and then started again, there could still be a read
717 * request pending. */
718 if (!(handle->flags & UV_HANDLE_READ_PENDING)) {
719 if (handle->flags & UV_HANDLE_EMULATE_IOCP &&
720 handle->read_req.event_handle == NULL) {
721 handle->read_req.event_handle = CreateEvent(NULL, 0, 0, NULL);
722 if (handle->read_req.event_handle == NULL) {
723 uv_fatal_error(GetLastError(), "CreateEvent");
724 }
725 }
726 uv__tcp_queue_read(loop, handle);
727 }
728
729 return 0;
730 }
731
uv__is_loopback(const struct sockaddr_storage * storage)732 static int uv__is_loopback(const struct sockaddr_storage* storage) {
733 const struct sockaddr_in* in4;
734 const struct sockaddr_in6* in6;
735 int i;
736
737 if (storage->ss_family == AF_INET) {
738 in4 = (const struct sockaddr_in*) storage;
739 return in4->sin_addr.S_un.S_un_b.s_b1 == 127;
740 }
741 if (storage->ss_family == AF_INET6) {
742 in6 = (const struct sockaddr_in6*) storage;
743 for (i = 0; i < 7; ++i) {
744 if (in6->sin6_addr.u.Word[i] != 0)
745 return 0;
746 }
747 return in6->sin6_addr.u.Word[7] == htons(1);
748 }
749 return 0;
750 }
751
752 // Check if Windows version is 10.0.16299 or later
uv__is_fast_loopback_fail_supported(void)753 static int uv__is_fast_loopback_fail_supported(void) {
754 OSVERSIONINFOW os_info;
755 if (!pRtlGetVersion)
756 return 0;
757 pRtlGetVersion(&os_info);
758 if (os_info.dwMajorVersion < 10)
759 return 0;
760 if (os_info.dwMajorVersion > 10)
761 return 1;
762 if (os_info.dwMinorVersion > 0)
763 return 1;
764 return os_info.dwBuildNumber >= 16299;
765 }
766
uv__tcp_try_connect(uv_connect_t * req,uv_tcp_t * handle,const struct sockaddr * addr,unsigned int addrlen,uv_connect_cb cb)767 static int uv__tcp_try_connect(uv_connect_t* req,
768 uv_tcp_t* handle,
769 const struct sockaddr* addr,
770 unsigned int addrlen,
771 uv_connect_cb cb) {
772 uv_loop_t* loop = handle->loop;
773 TCP_INITIAL_RTO_PARAMETERS retransmit_ioctl;
774 const struct sockaddr* bind_addr;
775 struct sockaddr_storage converted;
776 BOOL success;
777 DWORD bytes;
778 int err;
779
780 err = uv__convert_to_localhost_if_unspecified(addr, &converted);
781 if (err)
782 return err;
783
784 if (handle->delayed_error != 0)
785 goto out;
786
787 if (!(handle->flags & UV_HANDLE_BOUND)) {
788 if (addrlen == sizeof(uv_addr_ip4_any_)) {
789 bind_addr = (const struct sockaddr*) &uv_addr_ip4_any_;
790 } else if (addrlen == sizeof(uv_addr_ip6_any_)) {
791 bind_addr = (const struct sockaddr*) &uv_addr_ip6_any_;
792 } else {
793 abort();
794 }
795 err = uv__tcp_try_bind(handle, bind_addr, addrlen, 0);
796 if (err)
797 return err;
798 if (handle->delayed_error != 0)
799 goto out;
800 }
801
802 if (!handle->tcp.conn.func_connectex) {
803 if (!uv__get_connectex_function(handle->socket, &handle->tcp.conn.func_connectex)) {
804 return WSAEAFNOSUPPORT;
805 }
806 }
807
808 /* This makes connect() fail instantly if the target port on the localhost
809 * is not reachable, instead of waiting for 2s. We do not care if this fails.
810 * This only works on Windows version 10.0.16299 and later.
811 */
812 if (uv__is_fast_loopback_fail_supported() && uv__is_loopback(&converted)) {
813 memset(&retransmit_ioctl, 0, sizeof(retransmit_ioctl));
814 retransmit_ioctl.Rtt = TCP_INITIAL_RTO_NO_SYN_RETRANSMISSIONS;
815 retransmit_ioctl.MaxSynRetransmissions = TCP_INITIAL_RTO_NO_SYN_RETRANSMISSIONS;
816 WSAIoctl(handle->socket,
817 SIO_TCP_INITIAL_RTO,
818 &retransmit_ioctl,
819 sizeof(retransmit_ioctl),
820 NULL,
821 0,
822 &bytes,
823 NULL,
824 NULL);
825 }
826
827 out:
828
829 UV_REQ_INIT(req, UV_CONNECT);
830 req->handle = (uv_stream_t*) handle;
831 req->cb = cb;
832 memset(&req->u.io.overlapped, 0, sizeof(req->u.io.overlapped));
833
834 if (handle->delayed_error != 0) {
835 /* Process the req without IOCP. */
836 handle->reqs_pending++;
837 REGISTER_HANDLE_REQ(loop, handle);
838 uv__insert_pending_req(loop, (uv_req_t*)req);
839 return 0;
840 }
841
842 success = handle->tcp.conn.func_connectex(handle->socket,
843 (const struct sockaddr*) &converted,
844 addrlen,
845 NULL,
846 0,
847 &bytes,
848 &req->u.io.overlapped);
849
850 if (UV_SUCCEEDED_WITHOUT_IOCP(success)) {
851 /* Process the req without IOCP. */
852 handle->reqs_pending++;
853 REGISTER_HANDLE_REQ(loop, handle);
854 uv__insert_pending_req(loop, (uv_req_t*)req);
855 } else if (UV_SUCCEEDED_WITH_IOCP(success)) {
856 /* The req will be processed with IOCP. */
857 handle->reqs_pending++;
858 REGISTER_HANDLE_REQ(loop, handle);
859 } else {
860 return WSAGetLastError();
861 }
862
863 return 0;
864 }
865
866
uv_tcp_getsockname(const uv_tcp_t * handle,struct sockaddr * name,int * namelen)867 int uv_tcp_getsockname(const uv_tcp_t* handle,
868 struct sockaddr* name,
869 int* namelen) {
870
871 return uv__getsockpeername((const uv_handle_t*) handle,
872 getsockname,
873 name,
874 namelen,
875 handle->delayed_error);
876 }
877
878
uv_tcp_getpeername(const uv_tcp_t * handle,struct sockaddr * name,int * namelen)879 int uv_tcp_getpeername(const uv_tcp_t* handle,
880 struct sockaddr* name,
881 int* namelen) {
882
883 return uv__getsockpeername((const uv_handle_t*) handle,
884 getpeername,
885 name,
886 namelen,
887 handle->delayed_error);
888 }
889
890
uv__tcp_write(uv_loop_t * loop,uv_write_t * req,uv_tcp_t * handle,const uv_buf_t bufs[],unsigned int nbufs,uv_write_cb cb)891 int uv__tcp_write(uv_loop_t* loop,
892 uv_write_t* req,
893 uv_tcp_t* handle,
894 const uv_buf_t bufs[],
895 unsigned int nbufs,
896 uv_write_cb cb) {
897 int result;
898 DWORD bytes;
899
900 UV_REQ_INIT(req, UV_WRITE);
901 req->handle = (uv_stream_t*) handle;
902 req->cb = cb;
903
904 /* Prepare the overlapped structure. */
905 memset(&(req->u.io.overlapped), 0, sizeof(req->u.io.overlapped));
906 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
907 req->event_handle = CreateEvent(NULL, 0, 0, NULL);
908 if (req->event_handle == NULL) {
909 uv_fatal_error(GetLastError(), "CreateEvent");
910 }
911 req->u.io.overlapped.hEvent = (HANDLE) ((ULONG_PTR) req->event_handle | 1);
912 req->wait_handle = INVALID_HANDLE_VALUE;
913 }
914
915 result = WSASend(handle->socket,
916 (WSABUF*) bufs,
917 nbufs,
918 &bytes,
919 0,
920 &req->u.io.overlapped,
921 NULL);
922
923 if (UV_SUCCEEDED_WITHOUT_IOCP(result == 0)) {
924 /* Request completed immediately. */
925 req->u.io.queued_bytes = 0;
926 handle->reqs_pending++;
927 handle->stream.conn.write_reqs_pending++;
928 REGISTER_HANDLE_REQ(loop, handle);
929 uv__insert_pending_req(loop, (uv_req_t*) req);
930 } else if (UV_SUCCEEDED_WITH_IOCP(result == 0)) {
931 /* Request queued by the kernel. */
932 req->u.io.queued_bytes = uv__count_bufs(bufs, nbufs);
933 handle->reqs_pending++;
934 handle->stream.conn.write_reqs_pending++;
935 REGISTER_HANDLE_REQ(loop, handle);
936 handle->write_queue_size += req->u.io.queued_bytes;
937 if (handle->flags & UV_HANDLE_EMULATE_IOCP &&
938 !RegisterWaitForSingleObject(&req->wait_handle,
939 req->event_handle, post_write_completion, (void*) req,
940 INFINITE, WT_EXECUTEINWAITTHREAD | WT_EXECUTEONLYONCE)) {
941 SET_REQ_ERROR(req, GetLastError());
942 uv__insert_pending_req(loop, (uv_req_t*)req);
943 }
944 } else {
945 /* Send failed due to an error, report it later */
946 req->u.io.queued_bytes = 0;
947 handle->reqs_pending++;
948 handle->stream.conn.write_reqs_pending++;
949 REGISTER_HANDLE_REQ(loop, handle);
950 SET_REQ_ERROR(req, WSAGetLastError());
951 uv__insert_pending_req(loop, (uv_req_t*) req);
952 }
953
954 return 0;
955 }
956
957
uv__tcp_try_write(uv_tcp_t * handle,const uv_buf_t bufs[],unsigned int nbufs)958 int uv__tcp_try_write(uv_tcp_t* handle,
959 const uv_buf_t bufs[],
960 unsigned int nbufs) {
961 int result;
962 DWORD bytes;
963
964 if (handle->stream.conn.write_reqs_pending > 0)
965 return UV_EAGAIN;
966
967 result = WSASend(handle->socket,
968 (WSABUF*) bufs,
969 nbufs,
970 &bytes,
971 0,
972 NULL,
973 NULL);
974
975 if (result == SOCKET_ERROR)
976 return uv_translate_sys_error(WSAGetLastError());
977 else
978 return bytes;
979 }
980
981
uv__process_tcp_read_req(uv_loop_t * loop,uv_tcp_t * handle,uv_req_t * req)982 void uv__process_tcp_read_req(uv_loop_t* loop, uv_tcp_t* handle,
983 uv_req_t* req) {
984 DWORD bytes, flags, err;
985 uv_buf_t buf;
986 int count;
987
988 assert(handle->type == UV_TCP);
989
990 handle->flags &= ~UV_HANDLE_READ_PENDING;
991
992 if (!REQ_SUCCESS(req)) {
993 /* An error occurred doing the read. */
994 if ((handle->flags & UV_HANDLE_READING) ||
995 !(handle->flags & UV_HANDLE_ZERO_READ)) {
996 handle->flags &= ~UV_HANDLE_READING;
997 DECREASE_ACTIVE_COUNT(loop, handle);
998 buf = (handle->flags & UV_HANDLE_ZERO_READ) ?
999 uv_buf_init(NULL, 0) : handle->tcp.conn.read_buffer;
1000
1001 err = GET_REQ_SOCK_ERROR(req);
1002
1003 if (err == WSAECONNABORTED) {
1004 /* Turn WSAECONNABORTED into UV_ECONNRESET to be consistent with Unix.
1005 */
1006 err = WSAECONNRESET;
1007 }
1008 handle->flags &= ~(UV_HANDLE_READABLE | UV_HANDLE_WRITABLE);
1009
1010 handle->read_cb((uv_stream_t*)handle,
1011 uv_translate_sys_error(err),
1012 &buf);
1013 }
1014 } else {
1015 if (!(handle->flags & UV_HANDLE_ZERO_READ)) {
1016 /* The read was done with a non-zero buffer length. */
1017 if (req->u.io.overlapped.InternalHigh > 0) {
1018 /* Successful read */
1019 handle->read_cb((uv_stream_t*)handle,
1020 req->u.io.overlapped.InternalHigh,
1021 &handle->tcp.conn.read_buffer);
1022 /* Read again only if bytes == buf.len */
1023 if (req->u.io.overlapped.InternalHigh < handle->tcp.conn.read_buffer.len) {
1024 goto done;
1025 }
1026 } else {
1027 /* Connection closed */
1028 if (handle->flags & UV_HANDLE_READING) {
1029 handle->flags &= ~UV_HANDLE_READING;
1030 DECREASE_ACTIVE_COUNT(loop, handle);
1031 }
1032
1033 buf.base = 0;
1034 buf.len = 0;
1035 handle->read_cb((uv_stream_t*)handle, UV_EOF, &handle->tcp.conn.read_buffer);
1036 goto done;
1037 }
1038 }
1039
1040 /* Do nonblocking reads until the buffer is empty */
1041 count = 32;
1042 while ((handle->flags & UV_HANDLE_READING) && (count-- > 0)) {
1043 buf = uv_buf_init(NULL, 0);
1044 handle->alloc_cb((uv_handle_t*) handle, 65536, &buf);
1045 if (buf.base == NULL || buf.len == 0) {
1046 handle->read_cb((uv_stream_t*) handle, UV_ENOBUFS, &buf);
1047 break;
1048 }
1049 assert(buf.base != NULL);
1050
1051 flags = 0;
1052 if (WSARecv(handle->socket,
1053 (WSABUF*)&buf,
1054 1,
1055 &bytes,
1056 &flags,
1057 NULL,
1058 NULL) != SOCKET_ERROR) {
1059 if (bytes > 0) {
1060 /* Successful read */
1061 handle->read_cb((uv_stream_t*)handle, bytes, &buf);
1062 /* Read again only if bytes == buf.len */
1063 if (bytes < buf.len) {
1064 break;
1065 }
1066 } else {
1067 /* Connection closed */
1068 handle->flags &= ~UV_HANDLE_READING;
1069 DECREASE_ACTIVE_COUNT(loop, handle);
1070
1071 handle->read_cb((uv_stream_t*)handle, UV_EOF, &buf);
1072 break;
1073 }
1074 } else {
1075 err = WSAGetLastError();
1076 if (err == WSAEWOULDBLOCK) {
1077 /* Read buffer was completely empty, report a 0-byte read. */
1078 handle->read_cb((uv_stream_t*)handle, 0, &buf);
1079 } else {
1080 /* Ouch! serious error. */
1081 handle->flags &= ~UV_HANDLE_READING;
1082 DECREASE_ACTIVE_COUNT(loop, handle);
1083
1084 if (err == WSAECONNABORTED) {
1085 /* Turn WSAECONNABORTED into UV_ECONNRESET to be consistent with
1086 * Unix. */
1087 err = WSAECONNRESET;
1088 }
1089 handle->flags &= ~(UV_HANDLE_READABLE | UV_HANDLE_WRITABLE);
1090
1091 handle->read_cb((uv_stream_t*)handle,
1092 uv_translate_sys_error(err),
1093 &buf);
1094 }
1095 break;
1096 }
1097 }
1098
1099 done:
1100 /* Post another read if still reading and not closing. */
1101 if ((handle->flags & UV_HANDLE_READING) &&
1102 !(handle->flags & UV_HANDLE_READ_PENDING)) {
1103 uv__tcp_queue_read(loop, handle);
1104 }
1105 }
1106
1107 DECREASE_PENDING_REQ_COUNT(handle);
1108 }
1109
1110
uv__process_tcp_write_req(uv_loop_t * loop,uv_tcp_t * handle,uv_write_t * req)1111 void uv__process_tcp_write_req(uv_loop_t* loop, uv_tcp_t* handle,
1112 uv_write_t* req) {
1113 int err;
1114
1115 assert(handle->type == UV_TCP);
1116
1117 assert(handle->write_queue_size >= req->u.io.queued_bytes);
1118 handle->write_queue_size -= req->u.io.queued_bytes;
1119
1120 UNREGISTER_HANDLE_REQ(loop, handle);
1121
1122 if (handle->flags & UV_HANDLE_EMULATE_IOCP) {
1123 if (req->wait_handle != INVALID_HANDLE_VALUE) {
1124 UnregisterWait(req->wait_handle);
1125 req->wait_handle = INVALID_HANDLE_VALUE;
1126 }
1127 if (req->event_handle != NULL) {
1128 CloseHandle(req->event_handle);
1129 req->event_handle = NULL;
1130 }
1131 }
1132
1133 if (req->cb) {
1134 err = uv_translate_sys_error(GET_REQ_SOCK_ERROR(req));
1135 if (err == UV_ECONNABORTED) {
1136 /* use UV_ECANCELED for consistency with Unix */
1137 err = UV_ECANCELED;
1138 }
1139 req->cb(req, err);
1140 }
1141
1142 handle->stream.conn.write_reqs_pending--;
1143 if (handle->stream.conn.write_reqs_pending == 0) {
1144 if (handle->flags & UV_HANDLE_CLOSING) {
1145 closesocket(handle->socket);
1146 handle->socket = INVALID_SOCKET;
1147 }
1148 if (uv__is_stream_shutting(handle))
1149 uv__process_tcp_shutdown_req(loop,
1150 handle,
1151 handle->stream.conn.shutdown_req);
1152 }
1153
1154 DECREASE_PENDING_REQ_COUNT(handle);
1155 }
1156
1157
uv__process_tcp_accept_req(uv_loop_t * loop,uv_tcp_t * handle,uv_req_t * raw_req)1158 void uv__process_tcp_accept_req(uv_loop_t* loop, uv_tcp_t* handle,
1159 uv_req_t* raw_req) {
1160 uv_tcp_accept_t* req = (uv_tcp_accept_t*) raw_req;
1161 int err;
1162
1163 assert(handle->type == UV_TCP);
1164
1165 /* If handle->accepted_socket is not a valid socket, then uv_queue_accept
1166 * must have failed. This is a serious error. We stop accepting connections
1167 * and report this error to the connection callback. */
1168 if (req->accept_socket == INVALID_SOCKET) {
1169 if (handle->flags & UV_HANDLE_LISTENING) {
1170 handle->flags &= ~UV_HANDLE_LISTENING;
1171 DECREASE_ACTIVE_COUNT(loop, handle);
1172 if (handle->stream.serv.connection_cb) {
1173 err = GET_REQ_SOCK_ERROR(req);
1174 handle->stream.serv.connection_cb((uv_stream_t*)handle,
1175 uv_translate_sys_error(err));
1176 }
1177 }
1178 } else if (REQ_SUCCESS(req) &&
1179 setsockopt(req->accept_socket,
1180 SOL_SOCKET,
1181 SO_UPDATE_ACCEPT_CONTEXT,
1182 (char*)&handle->socket,
1183 sizeof(handle->socket)) == 0) {
1184 req->next_pending = handle->tcp.serv.pending_accepts;
1185 handle->tcp.serv.pending_accepts = req;
1186
1187 /* Accept and SO_UPDATE_ACCEPT_CONTEXT were successful. */
1188 if (handle->stream.serv.connection_cb) {
1189 handle->stream.serv.connection_cb((uv_stream_t*)handle, 0);
1190 }
1191 } else {
1192 /* Error related to accepted socket is ignored because the server socket
1193 * may still be healthy. If the server socket is broken uv_queue_accept
1194 * will detect it. */
1195 closesocket(req->accept_socket);
1196 req->accept_socket = INVALID_SOCKET;
1197 if (handle->flags & UV_HANDLE_LISTENING) {
1198 uv__tcp_queue_accept(handle, req);
1199 }
1200 }
1201
1202 DECREASE_PENDING_REQ_COUNT(handle);
1203 }
1204
1205
uv__process_tcp_connect_req(uv_loop_t * loop,uv_tcp_t * handle,uv_connect_t * req)1206 void uv__process_tcp_connect_req(uv_loop_t* loop, uv_tcp_t* handle,
1207 uv_connect_t* req) {
1208 int err;
1209
1210 assert(handle->type == UV_TCP);
1211
1212 UNREGISTER_HANDLE_REQ(loop, handle);
1213
1214 err = 0;
1215 if (handle->delayed_error) {
1216 /* To smooth over the differences between unixes errors that
1217 * were reported synchronously on the first connect can be delayed
1218 * until the next tick--which is now.
1219 */
1220 err = handle->delayed_error;
1221 handle->delayed_error = 0;
1222 } else if (REQ_SUCCESS(req)) {
1223 if (handle->flags & UV_HANDLE_CLOSING) {
1224 /* use UV_ECANCELED for consistency with Unix */
1225 err = ERROR_OPERATION_ABORTED;
1226 } else if (setsockopt(handle->socket,
1227 SOL_SOCKET,
1228 SO_UPDATE_CONNECT_CONTEXT,
1229 NULL,
1230 0) == 0) {
1231 uv__connection_init((uv_stream_t*)handle);
1232 handle->flags |= UV_HANDLE_READABLE | UV_HANDLE_WRITABLE;
1233 } else {
1234 err = WSAGetLastError();
1235 }
1236 } else {
1237 err = GET_REQ_SOCK_ERROR(req);
1238 }
1239 req->cb(req, uv_translate_sys_error(err));
1240
1241 DECREASE_PENDING_REQ_COUNT(handle);
1242 }
1243
1244
uv__tcp_xfer_export(uv_tcp_t * handle,int target_pid,uv__ipc_socket_xfer_type_t * xfer_type,uv__ipc_socket_xfer_info_t * xfer_info)1245 int uv__tcp_xfer_export(uv_tcp_t* handle,
1246 int target_pid,
1247 uv__ipc_socket_xfer_type_t* xfer_type,
1248 uv__ipc_socket_xfer_info_t* xfer_info) {
1249 if (handle->flags & UV_HANDLE_CONNECTION) {
1250 *xfer_type = UV__IPC_SOCKET_XFER_TCP_CONNECTION;
1251 } else {
1252 *xfer_type = UV__IPC_SOCKET_XFER_TCP_SERVER;
1253 /* We're about to share the socket with another process. Because this is a
1254 * listening socket, we assume that the other process will be accepting
1255 * connections on it. Thus, before sharing the socket with another process,
1256 * we call listen here in the parent process. */
1257 if (!(handle->flags & UV_HANDLE_LISTENING)) {
1258 if (!(handle->flags & UV_HANDLE_BOUND)) {
1259 return ERROR_NOT_SUPPORTED;
1260 }
1261 if (handle->delayed_error == 0 &&
1262 listen(handle->socket, SOMAXCONN) == SOCKET_ERROR) {
1263 handle->delayed_error = WSAGetLastError();
1264 }
1265 }
1266 }
1267
1268 if (WSADuplicateSocketW(handle->socket, target_pid, &xfer_info->socket_info))
1269 return WSAGetLastError();
1270 xfer_info->delayed_error = handle->delayed_error;
1271
1272 /* Mark the local copy of the handle as 'shared' so we behave in a way that's
1273 * friendly to the process(es) that we share the socket with. */
1274 handle->flags |= UV_HANDLE_SHARED_TCP_SOCKET;
1275
1276 return 0;
1277 }
1278
1279
uv__tcp_xfer_import(uv_tcp_t * tcp,uv__ipc_socket_xfer_type_t xfer_type,uv__ipc_socket_xfer_info_t * xfer_info)1280 int uv__tcp_xfer_import(uv_tcp_t* tcp,
1281 uv__ipc_socket_xfer_type_t xfer_type,
1282 uv__ipc_socket_xfer_info_t* xfer_info) {
1283 int err;
1284 SOCKET socket;
1285
1286 assert(xfer_type == UV__IPC_SOCKET_XFER_TCP_SERVER ||
1287 xfer_type == UV__IPC_SOCKET_XFER_TCP_CONNECTION);
1288
1289 socket = WSASocketW(FROM_PROTOCOL_INFO,
1290 FROM_PROTOCOL_INFO,
1291 FROM_PROTOCOL_INFO,
1292 &xfer_info->socket_info,
1293 0,
1294 WSA_FLAG_OVERLAPPED);
1295
1296 if (socket == INVALID_SOCKET) {
1297 return WSAGetLastError();
1298 }
1299
1300 err = uv__tcp_set_socket(
1301 tcp->loop, tcp, socket, xfer_info->socket_info.iAddressFamily, 1);
1302 if (err) {
1303 closesocket(socket);
1304 return err;
1305 }
1306
1307 tcp->delayed_error = xfer_info->delayed_error;
1308 tcp->flags |= UV_HANDLE_BOUND | UV_HANDLE_SHARED_TCP_SOCKET;
1309
1310 if (xfer_type == UV__IPC_SOCKET_XFER_TCP_CONNECTION) {
1311 uv__connection_init((uv_stream_t*)tcp);
1312 tcp->flags |= UV_HANDLE_READABLE | UV_HANDLE_WRITABLE;
1313 }
1314
1315 return 0;
1316 }
1317
1318
uv_tcp_nodelay(uv_tcp_t * handle,int enable)1319 int uv_tcp_nodelay(uv_tcp_t* handle, int enable) {
1320 int err;
1321
1322 if (handle->socket != INVALID_SOCKET) {
1323 err = uv__tcp_nodelay(handle, handle->socket, enable);
1324 if (err)
1325 return uv_translate_sys_error(err);
1326 }
1327
1328 if (enable) {
1329 handle->flags |= UV_HANDLE_TCP_NODELAY;
1330 } else {
1331 handle->flags &= ~UV_HANDLE_TCP_NODELAY;
1332 }
1333
1334 return 0;
1335 }
1336
1337
uv_tcp_keepalive(uv_tcp_t * handle,int enable,unsigned int delay)1338 int uv_tcp_keepalive(uv_tcp_t* handle, int enable, unsigned int delay) {
1339 int err;
1340
1341 if (handle->socket != INVALID_SOCKET) {
1342 err = uv__tcp_keepalive(handle, handle->socket, enable, delay);
1343 if (err)
1344 return uv_translate_sys_error(err);
1345 }
1346
1347 if (enable) {
1348 handle->flags |= UV_HANDLE_TCP_KEEPALIVE;
1349 } else {
1350 handle->flags &= ~UV_HANDLE_TCP_KEEPALIVE;
1351 }
1352
1353 /* TODO: Store delay if handle->socket isn't created yet. */
1354
1355 return 0;
1356 }
1357
1358
uv_tcp_simultaneous_accepts(uv_tcp_t * handle,int enable)1359 int uv_tcp_simultaneous_accepts(uv_tcp_t* handle, int enable) {
1360 if (handle->flags & UV_HANDLE_CONNECTION) {
1361 return UV_EINVAL;
1362 }
1363
1364 /* Check if we're already in the desired mode. */
1365 if ((enable && !(handle->flags & UV_HANDLE_TCP_SINGLE_ACCEPT)) ||
1366 (!enable && handle->flags & UV_HANDLE_TCP_SINGLE_ACCEPT)) {
1367 return 0;
1368 }
1369
1370 /* Don't allow switching from single pending accept to many. */
1371 if (enable) {
1372 return UV_ENOTSUP;
1373 }
1374
1375 /* Check if we're in a middle of changing the number of pending accepts. */
1376 if (handle->flags & UV_HANDLE_TCP_ACCEPT_STATE_CHANGING) {
1377 return 0;
1378 }
1379
1380 handle->flags |= UV_HANDLE_TCP_SINGLE_ACCEPT;
1381
1382 /* Flip the changing flag if we have already queued multiple accepts. */
1383 if (handle->flags & UV_HANDLE_LISTENING) {
1384 handle->flags |= UV_HANDLE_TCP_ACCEPT_STATE_CHANGING;
1385 }
1386
1387 return 0;
1388 }
1389
1390
uv__tcp_try_cancel_reqs(uv_tcp_t * tcp)1391 static void uv__tcp_try_cancel_reqs(uv_tcp_t* tcp) {
1392 SOCKET socket;
1393 int non_ifs_lsp;
1394 int reading;
1395 int writing;
1396
1397 socket = tcp->socket;
1398 reading = tcp->flags & UV_HANDLE_READ_PENDING;
1399 writing = tcp->stream.conn.write_reqs_pending > 0;
1400 if (!reading && !writing)
1401 return;
1402
1403 /* TODO: in libuv v2, keep explicit track of write_reqs, so we can cancel
1404 * them each explicitly with CancelIoEx (like unix). */
1405 if (reading)
1406 CancelIoEx((HANDLE) socket, &tcp->read_req.u.io.overlapped);
1407 if (writing)
1408 CancelIo((HANDLE) socket);
1409
1410 /* Check if we have any non-IFS LSPs stacked on top of TCP */
1411 non_ifs_lsp = (tcp->flags & UV_HANDLE_IPV6) ? uv_tcp_non_ifs_lsp_ipv6 :
1412 uv_tcp_non_ifs_lsp_ipv4;
1413
1414 /* If there are non-ifs LSPs then try to obtain a base handle for the socket.
1415 */
1416 if (non_ifs_lsp) {
1417 DWORD bytes;
1418 if (WSAIoctl(socket,
1419 SIO_BASE_HANDLE,
1420 NULL,
1421 0,
1422 &socket,
1423 sizeof socket,
1424 &bytes,
1425 NULL,
1426 NULL) != 0) {
1427 /* Failed. We can't do CancelIo. */
1428 return;
1429 }
1430 }
1431
1432 assert(socket != 0 && socket != INVALID_SOCKET);
1433
1434 if (socket != tcp->socket) {
1435 if (reading)
1436 CancelIoEx((HANDLE) socket, &tcp->read_req.u.io.overlapped);
1437 if (writing)
1438 CancelIo((HANDLE) socket);
1439 }
1440 }
1441
1442
uv__tcp_close(uv_loop_t * loop,uv_tcp_t * tcp)1443 void uv__tcp_close(uv_loop_t* loop, uv_tcp_t* tcp) {
1444 if (tcp->flags & UV_HANDLE_CONNECTION) {
1445 if (tcp->flags & UV_HANDLE_READING) {
1446 uv_read_stop((uv_stream_t*) tcp);
1447 }
1448 uv__tcp_try_cancel_reqs(tcp);
1449 } else {
1450 if (tcp->tcp.serv.accept_reqs != NULL) {
1451 /* First close the incoming sockets to cancel the accept operations before
1452 * we free their resources. */
1453 unsigned int i;
1454 for (i = 0; i < uv_simultaneous_server_accepts; i++) {
1455 uv_tcp_accept_t* req = &tcp->tcp.serv.accept_reqs[i];
1456 if (req->accept_socket != INVALID_SOCKET) {
1457 closesocket(req->accept_socket);
1458 req->accept_socket = INVALID_SOCKET;
1459 }
1460 }
1461 }
1462 assert(!(tcp->flags & UV_HANDLE_READING));
1463 }
1464
1465 if (tcp->flags & UV_HANDLE_LISTENING) {
1466 tcp->flags &= ~UV_HANDLE_LISTENING;
1467 DECREASE_ACTIVE_COUNT(loop, tcp);
1468 }
1469
1470 tcp->flags &= ~(UV_HANDLE_READABLE | UV_HANDLE_WRITABLE);
1471 uv__handle_closing(tcp);
1472
1473 /* If any overlapped req failed to cancel, calling `closesocket` now would
1474 * cause Win32 to send an RST packet. Try to avoid that for writes, if
1475 * possibly applicable, by waiting to process the completion notifications
1476 * first (which typically should be cancellations). There's not much we can
1477 * do about canceled reads, which also will generate an RST packet. */
1478 if (!(tcp->flags & UV_HANDLE_CONNECTION) ||
1479 tcp->stream.conn.write_reqs_pending == 0) {
1480 closesocket(tcp->socket);
1481 tcp->socket = INVALID_SOCKET;
1482 }
1483
1484 if (tcp->reqs_pending == 0)
1485 uv__want_endgame(loop, (uv_handle_t*) tcp);
1486 }
1487
1488
uv_tcp_open(uv_tcp_t * handle,uv_os_sock_t sock)1489 int uv_tcp_open(uv_tcp_t* handle, uv_os_sock_t sock) {
1490 WSAPROTOCOL_INFOW protocol_info;
1491 int opt_len;
1492 int err;
1493 struct sockaddr_storage saddr;
1494 int saddr_len;
1495
1496 /* Detect the address family of the socket. */
1497 opt_len = (int) sizeof protocol_info;
1498 if (getsockopt(sock,
1499 SOL_SOCKET,
1500 SO_PROTOCOL_INFOW,
1501 (char*) &protocol_info,
1502 &opt_len) == SOCKET_ERROR) {
1503 return uv_translate_sys_error(GetLastError());
1504 }
1505
1506 err = uv__tcp_set_socket(handle->loop,
1507 handle,
1508 sock,
1509 protocol_info.iAddressFamily,
1510 1);
1511 if (err) {
1512 return uv_translate_sys_error(err);
1513 }
1514
1515 /* Support already active socket. */
1516 saddr_len = sizeof(saddr);
1517 if (!uv_tcp_getsockname(handle, (struct sockaddr*) &saddr, &saddr_len)) {
1518 /* Socket is already bound. */
1519 handle->flags |= UV_HANDLE_BOUND;
1520 saddr_len = sizeof(saddr);
1521 if (!uv_tcp_getpeername(handle, (struct sockaddr*) &saddr, &saddr_len)) {
1522 /* Socket is already connected. */
1523 uv__connection_init((uv_stream_t*) handle);
1524 handle->flags |= UV_HANDLE_READABLE | UV_HANDLE_WRITABLE;
1525 }
1526 }
1527
1528 return 0;
1529 }
1530
1531
1532 /* This function is an egress point, i.e. it returns libuv errors rather than
1533 * system errors.
1534 */
uv__tcp_bind(uv_tcp_t * handle,const struct sockaddr * addr,unsigned int addrlen,unsigned int flags)1535 int uv__tcp_bind(uv_tcp_t* handle,
1536 const struct sockaddr* addr,
1537 unsigned int addrlen,
1538 unsigned int flags) {
1539 int err;
1540
1541 err = uv__tcp_try_bind(handle, addr, addrlen, flags);
1542 if (err)
1543 return uv_translate_sys_error(err);
1544
1545 return 0;
1546 }
1547
1548
1549 /* This function is an egress point, i.e. it returns libuv errors rather than
1550 * system errors.
1551 */
uv__tcp_connect(uv_connect_t * req,uv_tcp_t * handle,const struct sockaddr * addr,unsigned int addrlen,uv_connect_cb cb)1552 int uv__tcp_connect(uv_connect_t* req,
1553 uv_tcp_t* handle,
1554 const struct sockaddr* addr,
1555 unsigned int addrlen,
1556 uv_connect_cb cb) {
1557 int err;
1558
1559 err = uv__tcp_try_connect(req, handle, addr, addrlen, cb);
1560 if (err)
1561 return uv_translate_sys_error(err);
1562
1563 return 0;
1564 }
1565
1566
uv_socketpair(int type,int protocol,uv_os_sock_t fds[2],int flags0,int flags1)1567 int uv_socketpair(int type, int protocol, uv_os_sock_t fds[2], int flags0, int flags1) {
1568 SOCKET server = INVALID_SOCKET;
1569 SOCKET client0 = INVALID_SOCKET;
1570 SOCKET client1 = INVALID_SOCKET;
1571 SOCKADDR_IN name;
1572 LPFN_ACCEPTEX func_acceptex;
1573 WSAOVERLAPPED overlap;
1574 char accept_buffer[sizeof(struct sockaddr_storage) * 2 + 32];
1575 int namelen;
1576 int err;
1577 DWORD bytes;
1578 DWORD flags;
1579 DWORD client0_flags = WSA_FLAG_NO_HANDLE_INHERIT;
1580 DWORD client1_flags = WSA_FLAG_NO_HANDLE_INHERIT;
1581
1582 if (flags0 & UV_NONBLOCK_PIPE)
1583 client0_flags |= WSA_FLAG_OVERLAPPED;
1584 if (flags1 & UV_NONBLOCK_PIPE)
1585 client1_flags |= WSA_FLAG_OVERLAPPED;
1586
1587 server = WSASocketW(AF_INET, type, protocol, NULL, 0,
1588 WSA_FLAG_OVERLAPPED | WSA_FLAG_NO_HANDLE_INHERIT);
1589 if (server == INVALID_SOCKET)
1590 goto wsaerror;
1591 if (!SetHandleInformation((HANDLE) server, HANDLE_FLAG_INHERIT, 0))
1592 goto error;
1593 name.sin_family = AF_INET;
1594 name.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1595 name.sin_port = 0;
1596 if (bind(server, (SOCKADDR*) &name, sizeof(name)) != 0)
1597 goto wsaerror;
1598 if (listen(server, 1) != 0)
1599 goto wsaerror;
1600 namelen = sizeof(name);
1601 if (getsockname(server, (SOCKADDR*) &name, &namelen) != 0)
1602 goto wsaerror;
1603 client0 = WSASocketW(AF_INET, type, protocol, NULL, 0, client0_flags);
1604 if (client0 == INVALID_SOCKET)
1605 goto wsaerror;
1606 if (!SetHandleInformation((HANDLE) client0, HANDLE_FLAG_INHERIT, 0))
1607 goto error;
1608 if (connect(client0, (SOCKADDR*) &name, sizeof(name)) != 0)
1609 goto wsaerror;
1610 client1 = WSASocketW(AF_INET, type, protocol, NULL, 0, client1_flags);
1611 if (client1 == INVALID_SOCKET)
1612 goto wsaerror;
1613 if (!SetHandleInformation((HANDLE) client1, HANDLE_FLAG_INHERIT, 0))
1614 goto error;
1615 if (!uv__get_acceptex_function(server, &func_acceptex)) {
1616 err = WSAEAFNOSUPPORT;
1617 goto cleanup;
1618 }
1619 memset(&overlap, 0, sizeof(overlap));
1620 if (!func_acceptex(server,
1621 client1,
1622 accept_buffer,
1623 0,
1624 sizeof(struct sockaddr_storage),
1625 sizeof(struct sockaddr_storage),
1626 &bytes,
1627 &overlap)) {
1628 err = WSAGetLastError();
1629 if (err == ERROR_IO_PENDING) {
1630 /* Result should complete immediately, since we already called connect,
1631 * but empirically, we sometimes have to poll the kernel a couple times
1632 * until it notices that. */
1633 while (!WSAGetOverlappedResult(client1, &overlap, &bytes, FALSE, &flags)) {
1634 err = WSAGetLastError();
1635 if (err != WSA_IO_INCOMPLETE)
1636 goto cleanup;
1637 SwitchToThread();
1638 }
1639 }
1640 else {
1641 goto cleanup;
1642 }
1643 }
1644 if (setsockopt(client1, SOL_SOCKET, SO_UPDATE_ACCEPT_CONTEXT,
1645 (char*) &server, sizeof(server)) != 0) {
1646 goto wsaerror;
1647 }
1648
1649 closesocket(server);
1650
1651 fds[0] = client0;
1652 fds[1] = client1;
1653
1654 return 0;
1655
1656 wsaerror:
1657 err = WSAGetLastError();
1658 goto cleanup;
1659
1660 error:
1661 err = GetLastError();
1662 goto cleanup;
1663
1664 cleanup:
1665 if (server != INVALID_SOCKET)
1666 closesocket(server);
1667 if (client0 != INVALID_SOCKET)
1668 closesocket(client0);
1669 if (client1 != INVALID_SOCKET)
1670 closesocket(client1);
1671
1672 assert(err);
1673 return uv_translate_sys_error(err);
1674 }
1675
