Debian Codegraph @

src/backend/libpq/pqcomm.c

raw
1 /*-------------------------------------------------------------------------
2 *
3 * pqcomm.c
4 * Communication functions between the Frontend and the Backend
5 *
6 * These routines handle the low-level details of communication between
7 * frontend and backend. They just shove data across the communication
8 * channel, and are ignorant of the semantics of the data.
9 *
10 * To emit an outgoing message, use the routines in pqformat.c to construct
11 * the message in a buffer and then emit it in one call to pq_putmessage.
12 * There are no functions to send raw bytes or partial messages; this
13 * ensures that the channel will not be clogged by an incomplete message if
14 * execution is aborted by ereport(ERROR) partway through the message.
15 *
16 * At one time, libpq was shared between frontend and backend, but now
17 * the backend's "backend/libpq" is quite separate from "interfaces/libpq".
18 * All that remains is similarities of names to trap the unwary...
19 *
20 * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
21 * Portions Copyright (c) 1994, Regents of the University of California
22 *
23 * src/backend/libpq/pqcomm.c
24 *
25 *-------------------------------------------------------------------------
26 */
27
28 /*------------------------
29 * INTERFACE ROUTINES
30 *
31 * setup/teardown:
32 * ListenServerPort - Open postmaster's server port
33 * AcceptConnection - Accept new connection with client
34 * TouchSocketFiles - Protect socket files against /tmp cleaners
35 * pq_init - initialize libpq at backend startup
36 * socket_comm_reset - reset libpq during error recovery
37 * socket_close - shutdown libpq at backend exit
38 *
39 * low-level I/O:
40 * pq_getbytes - get a known number of bytes from connection
41 * pq_getmessage - get a message with length word from connection
42 * pq_getbyte - get next byte from connection
43 * pq_peekbyte - peek at next byte from connection
44 * pq_flush - flush pending output
45 * pq_flush_if_writable - flush pending output if writable without blocking
46 * pq_getbyte_if_available - get a byte if available without blocking
47 *
48 * message-level I/O
49 * pq_putmessage - send a normal message (suppressed in COPY OUT mode)
50 * pq_putmessage_noblock - buffer a normal message (suppressed in COPY OUT)
51 *
52 *------------------------
53 */
54 #include "postgres.h"
55
56 #ifdef HAVE_POLL_H
57 #include <poll.h>
58 #endif
59 #include <signal.h>
60 #include <fcntl.h>
61 #include <grp.h>
62 #include <unistd.h>
63 #include <sys/file.h>
64 #include <sys/socket.h>
65 #include <sys/stat.h>
66 #include <sys/time.h>
67 #include <netdb.h>
68 #include <netinet/in.h>
69 #include <netinet/tcp.h>
70 #include <utime.h>
71 #ifdef WIN32
72 #include <mstcpip.h>
73 #endif
74
75 #include "common/ip.h"
76 #include "libpq/libpq.h"
77 #include "miscadmin.h"
78 #include "port/pg_bswap.h"
79 #include "postmaster/postmaster.h"
80 #include "storage/ipc.h"
81 #include "utils/guc_hooks.h"
82 #include "utils/memutils.h"
83
84 /*
85 * Cope with the various platform-specific ways to spell TCP keepalive socket
86 * options. This doesn't cover Windows, which as usual does its own thing.
87 */
88 #if defined(TCP_KEEPIDLE)
89 /* TCP_KEEPIDLE is the name of this option on Linux and *BSD */
90 #define PG_TCP_KEEPALIVE_IDLE TCP_KEEPIDLE
91 #define PG_TCP_KEEPALIVE_IDLE_STR "TCP_KEEPIDLE"
92 #elif defined(TCP_KEEPALIVE_THRESHOLD)
93 /* TCP_KEEPALIVE_THRESHOLD is the name of this option on Solaris >= 11 */
94 #define PG_TCP_KEEPALIVE_IDLE TCP_KEEPALIVE_THRESHOLD
95 #define PG_TCP_KEEPALIVE_IDLE_STR "TCP_KEEPALIVE_THRESHOLD"
96 #elif defined(TCP_KEEPALIVE) && defined(__darwin__)
97 /* TCP_KEEPALIVE is the name of this option on macOS */
98 /* Caution: Solaris has this symbol but it means something different */
99 #define PG_TCP_KEEPALIVE_IDLE TCP_KEEPALIVE
100 #define PG_TCP_KEEPALIVE_IDLE_STR "TCP_KEEPALIVE"
101 #endif
102
103 /*
104 * Configuration options
105 */
106 int Unix_socket_permissions;
107 char *Unix_socket_group;
108
109 /* Where the Unix socket files are (list of palloc'd strings) */
110 static List *sock_paths = NIL;
111
112 /*
113 * Buffers for low-level I/O.
114 *
115 * The receive buffer is fixed size. Send buffer is usually 8k, but can be
116 * enlarged by pq_putmessage_noblock() if the message doesn't fit otherwise.
117 */
118
119 #define PQ_SEND_BUFFER_SIZE 8192
120 #define PQ_RECV_BUFFER_SIZE 8192
121
122 static char *PqSendBuffer;
123 static int PqSendBufferSize; /* Size send buffer */
124 static size_t PqSendPointer; /* Next index to store a byte in PqSendBuffer */
125 static size_t PqSendStart; /* Next index to send a byte in PqSendBuffer */
126
127 static char PqRecvBuffer[PQ_RECV_BUFFER_SIZE];
128 static int PqRecvPointer; /* Next index to read a byte from PqRecvBuffer */
129 static int PqRecvLength; /* End of data available in PqRecvBuffer */
130
131 /*
132 * Message status
133 */
134 static bool PqCommBusy; /* busy sending data to the client */
135 static bool PqCommReadingMsg; /* in the middle of reading a message */
136
137
138 /* Internal functions */
139 static void socket_comm_reset(void);
140 static void socket_close(int code, Datum arg);
141 static void socket_set_nonblocking(bool nonblocking);
142 static int socket_flush(void);
143 static int socket_flush_if_writable(void);
144 static bool socket_is_send_pending(void);
145 static int socket_putmessage(char msgtype, const char *s, size_t len);
146 static void socket_putmessage_noblock(char msgtype, const char *s, size_t len);
147 static inline int internal_putbytes(const void *b, size_t len);
148 static inline int internal_flush(void);
149 static pg_noinline int internal_flush_buffer(const char *buf, size_t *start,
150 size_t *end);
151
152 static int Lock_AF_UNIX(const char *unixSocketDir, const char *unixSocketPath);
153 static int Setup_AF_UNIX(const char *sock_path);
154
155 static const PQcommMethods PqCommSocketMethods = {
156 .comm_reset = socket_comm_reset,
157 .flush = socket_flush,
158 .flush_if_writable = socket_flush_if_writable,
159 .is_send_pending = socket_is_send_pending,
160 .putmessage = socket_putmessage,
161 .putmessage_noblock = socket_putmessage_noblock
162 };
163
164 const PQcommMethods *PqCommMethods = &PqCommSocketMethods;
165
166 WaitEventSet *FeBeWaitSet;
167
168
169 /* --------------------------------
170 * pq_init - initialize libpq at backend startup
171 * --------------------------------
172 */
173 Port *
174 pq_init(ClientSocket *client_sock)
175 {
176 Port *port;
177 int socket_pos PG_USED_FOR_ASSERTS_ONLY;
178 int latch_pos PG_USED_FOR_ASSERTS_ONLY;
179
180 /* allocate the Port struct and copy the ClientSocket contents to it */
181 port = palloc0(sizeof(Port));
182 port->sock = client_sock->sock;
183 memcpy(&port->raddr.addr, &client_sock->raddr.addr, client_sock->raddr.salen);
184 port->raddr.salen = client_sock->raddr.salen;
185
186 /* fill in the server (local) address */
187 port->laddr.salen = sizeof(port->laddr.addr);
188 if (getsockname(port->sock,
189 (struct sockaddr *) &port->laddr.addr,
190 &port->laddr.salen) < 0)
191 {
192 ereport(FATAL,
193 (errmsg("%s() failed: %m", "getsockname")));
194 }
195
196 /* select NODELAY and KEEPALIVE options if it's a TCP connection */
197 if (port->laddr.addr.ss_family != AF_UNIX)
198 {
199 int on;
200 #ifdef WIN32
201 int oldopt;
202 int optlen;
203 int newopt;
204 #endif
205
206 #ifdef TCP_NODELAY
207 on = 1;
208 if (setsockopt(port->sock, IPPROTO_TCP, TCP_NODELAY,
209 (char *) &on, sizeof(on)) < 0)
210 {
211 ereport(FATAL,
212 (errmsg("%s(%s) failed: %m", "setsockopt", "TCP_NODELAY")));
213 }
214 #endif
215 on = 1;
216 if (setsockopt(port->sock, SOL_SOCKET, SO_KEEPALIVE,
217 (char *) &on, sizeof(on)) < 0)
218 {
219 ereport(FATAL,
220 (errmsg("%s(%s) failed: %m", "setsockopt", "SO_KEEPALIVE")));
221 }
222
223 #ifdef WIN32
224
225 /*
226 * This is a Win32 socket optimization. The OS send buffer should be
227 * large enough to send the whole Postgres send buffer in one go, or
228 * performance suffers. The Postgres send buffer can be enlarged if a
229 * very large message needs to be sent, but we won't attempt to
230 * enlarge the OS buffer if that happens, so somewhat arbitrarily
231 * ensure that the OS buffer is at least PQ_SEND_BUFFER_SIZE * 4.
232 * (That's 32kB with the current default).
233 *
234 * The default OS buffer size used to be 8kB in earlier Windows
235 * versions, but was raised to 64kB in Windows 2012. So it shouldn't
236 * be necessary to change it in later versions anymore. Changing it
237 * unnecessarily can even reduce performance, because setting
238 * SO_SNDBUF in the application disables the "dynamic send buffering"
239 * feature that was introduced in Windows 7. So before fiddling with
240 * SO_SNDBUF, check if the current buffer size is already large enough
241 * and only increase it if necessary.
242 *
243 * See https://support.microsoft.com/kb/823764/EN-US/ and
244 * https://msdn.microsoft.com/en-us/library/bb736549%28v=vs.85%29.aspx
245 */
246 optlen = sizeof(oldopt);
247 if (getsockopt(port->sock, SOL_SOCKET, SO_SNDBUF, (char *) &oldopt,
248 &optlen) < 0)
249 {
250 ereport(FATAL,
251 (errmsg("%s(%s) failed: %m", "getsockopt", "SO_SNDBUF")));
252 }
253 newopt = PQ_SEND_BUFFER_SIZE * 4;
254 if (oldopt < newopt)
255 {
256 if (setsockopt(port->sock, SOL_SOCKET, SO_SNDBUF, (char *) &newopt,
257 sizeof(newopt)) < 0)
258 {
259 ereport(FATAL,
260 (errmsg("%s(%s) failed: %m", "setsockopt", "SO_SNDBUF")));
261 }
262 }
263 #endif
264
265 /*
266 * Also apply the current keepalive parameters. If we fail to set a
267 * parameter, don't error out, because these aren't universally
268 * supported. (Note: you might think we need to reset the GUC
269 * variables to 0 in such a case, but it's not necessary because the
270 * show hooks for these variables report the truth anyway.)
271 */
272 (void) pq_setkeepalivesidle(tcp_keepalives_idle, port);
273 (void) pq_setkeepalivesinterval(tcp_keepalives_interval, port);
274 (void) pq_setkeepalivescount(tcp_keepalives_count, port);
275 (void) pq_settcpusertimeout(tcp_user_timeout, port);
276 }
277
278 /* initialize state variables */
279 PqSendBufferSize = PQ_SEND_BUFFER_SIZE;
280 PqSendBuffer = MemoryContextAlloc(TopMemoryContext, PqSendBufferSize);
281 PqSendPointer = PqSendStart = PqRecvPointer = PqRecvLength = 0;
282 PqCommBusy = false;
283 PqCommReadingMsg = false;
284
285 /* set up process-exit hook to close the socket */
286 on_proc_exit(socket_close, 0);
287
288 /*
289 * In backends (as soon as forked) we operate the underlying socket in
290 * nonblocking mode and use latches to implement blocking semantics if
291 * needed. That allows us to provide safely interruptible reads and
292 * writes.
293 */
294 #ifndef WIN32
295 if (!pg_set_noblock(port->sock))
296 ereport(FATAL,
297 (errmsg("could not set socket to nonblocking mode: %m")));
298 #endif
299
300 #ifndef WIN32
301
302 /* Don't give the socket to any subprograms we execute. */
303 if (fcntl(port->sock, F_SETFD, FD_CLOEXEC) < 0)
304 elog(FATAL, "fcntl(F_SETFD) failed on socket: %m");
305 #endif
306
307 FeBeWaitSet = CreateWaitEventSet(NULL, FeBeWaitSetNEvents);
308 socket_pos = AddWaitEventToSet(FeBeWaitSet, WL_SOCKET_WRITEABLE,
309 port->sock, NULL, NULL);
310 latch_pos = AddWaitEventToSet(FeBeWaitSet, WL_LATCH_SET, PGINVALID_SOCKET,
311 MyLatch, NULL);
312 AddWaitEventToSet(FeBeWaitSet, WL_POSTMASTER_DEATH, PGINVALID_SOCKET,
313 NULL, NULL);
314
315 /*
316 * The event positions match the order we added them, but let's sanity
317 * check them to be sure.
318 */
319 Assert(socket_pos == FeBeWaitSetSocketPos);
320 Assert(latch_pos == FeBeWaitSetLatchPos);
321
322 return port;
323 }
324
325 /* --------------------------------
326 * socket_comm_reset - reset libpq during error recovery
327 *
328 * This is called from error recovery at the outer idle loop. It's
329 * just to get us out of trouble if we somehow manage to elog() from
330 * inside a pqcomm.c routine (which ideally will never happen, but...)
331 * --------------------------------
332 */
333 static void
334 socket_comm_reset(void)
335 {
336 /* Do not throw away pending data, but do reset the busy flag */
337 PqCommBusy = false;
338 }
339
340 /* --------------------------------
341 * socket_close - shutdown libpq at backend exit
342 *
343 * This is the one pg_on_exit_callback in place during BackendInitialize().
344 * That function's unusual signal handling constrains that this callback be
345 * safe to run at any instant.
346 * --------------------------------
347 */
348 static void
349 socket_close(int code, Datum arg)
350 {
351 /* Nothing to do in a standalone backend, where MyProcPort is NULL. */
352 if (MyProcPort != NULL)
353 {
354 #ifdef ENABLE_GSS
355 /*
356 * Shutdown GSSAPI layer. This section does nothing when interrupting
357 * BackendInitialize(), because pg_GSS_recvauth() makes first use of
358 * "ctx" and "cred".
359 *
360 * Note that we don't bother to free MyProcPort->gss, since we're
361 * about to exit anyway.
362 */
363 if (MyProcPort->gss)
364 {
365 OM_uint32 min_s;
366
367 if (MyProcPort->gss->ctx != GSS_C_NO_CONTEXT)
368 gss_delete_sec_context(&min_s, &MyProcPort->gss->ctx, NULL);
369
370 if (MyProcPort->gss->cred != GSS_C_NO_CREDENTIAL)
371 gss_release_cred(&min_s, &MyProcPort->gss->cred);
372 }
373 #endif /* ENABLE_GSS */
374
375 /*
376 * Cleanly shut down SSL layer. Nowhere else does a postmaster child
377 * call this, so this is safe when interrupting BackendInitialize().
378 */
379 secure_close(MyProcPort);
380
381 /*
382 * Formerly we did an explicit close() here, but it seems better to
383 * leave the socket open until the process dies. This allows clients
384 * to perform a "synchronous close" if they care --- wait till the
385 * transport layer reports connection closure, and you can be sure the
386 * backend has exited.
387 *
388 * We do set sock to PGINVALID_SOCKET to prevent any further I/O,
389 * though.
390 */
391 MyProcPort->sock = PGINVALID_SOCKET;
392 }
393 }
394
395
396
397 /* --------------------------------
398 * Postmaster functions to handle sockets.
399 * --------------------------------
400 */
401
402 /*
403 * ListenServerPort -- open a "listening" port to accept connections.
404 *
405 * family should be AF_UNIX or AF_UNSPEC; portNumber is the port number.
406 * For AF_UNIX ports, hostName should be NULL and unixSocketDir must be
407 * specified. For TCP ports, hostName is either NULL for all interfaces or
408 * the interface to listen on, and unixSocketDir is ignored (can be NULL).
409 *
410 * Successfully opened sockets are appended to the ListenSockets[] array. On
411 * entry, *NumListenSockets holds the number of elements currently in the
412 * array, and it is updated to reflect the opened sockets. MaxListen is the
413 * allocated size of the array.
414 *
415 * RETURNS: STATUS_OK or STATUS_ERROR
416 */
417 int
418 ListenServerPort(int family, const char *hostName, unsigned short portNumber,
419 const char *unixSocketDir,
420 pgsocket ListenSockets[], int *NumListenSockets, int MaxListen)
421 {
422 pgsocket fd;
423 int err;
424 int maxconn;
425 int ret;
426 char portNumberStr[32];
427 const char *familyDesc;
428 char familyDescBuf[64];
429 const char *addrDesc;
430 char addrBuf[NI_MAXHOST];
431 char *service;
432 struct addrinfo *addrs = NULL,
433 *addr;
434 struct addrinfo hint;
435 int added = 0;
436 char unixSocketPath[MAXPGPATH];
437 #if !defined(WIN32) || defined(IPV6_V6ONLY)
438 int one = 1;
439 #endif
440
441 /* Initialize hint structure */
442 MemSet(&hint, 0, sizeof(hint));
443 hint.ai_family = family;
444 hint.ai_flags = AI_PASSIVE;
445 hint.ai_socktype = SOCK_STREAM;
446
447 if (family == AF_UNIX)
448 {
449 /*
450 * Create unixSocketPath from portNumber and unixSocketDir and lock
451 * that file path
452 */
453 UNIXSOCK_PATH(unixSocketPath, portNumber, unixSocketDir);
454 if (strlen(unixSocketPath) >= UNIXSOCK_PATH_BUFLEN)
455 {
456 ereport(LOG,
457 (errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
458 unixSocketPath,
459 (int) (UNIXSOCK_PATH_BUFLEN - 1))));
460 return STATUS_ERROR;
461 }
462 if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
463 return STATUS_ERROR;
464 service = unixSocketPath;
465 }
466 else
467 {
468 snprintf(portNumberStr, sizeof(portNumberStr), "%d", portNumber);
469 service = portNumberStr;
470 }
471
472 ret = pg_getaddrinfo_all(hostName, service, &hint, &addrs);
473 if (ret || !addrs)
474 {
475 if (hostName)
476 ereport(LOG,
477 (errmsg("could not translate host name \"%s\", service \"%s\" to address: %s",
478 hostName, service, gai_strerror(ret))));
479 else
480 ereport(LOG,
481 (errmsg("could not translate service \"%s\" to address: %s",
482 service, gai_strerror(ret))));
483 if (addrs)
484 pg_freeaddrinfo_all(hint.ai_family, addrs);
485 return STATUS_ERROR;
486 }
487
488 for (addr = addrs; addr; addr = addr->ai_next)
489 {
490 if (family != AF_UNIX && addr->ai_family == AF_UNIX)
491 {
492 /*
493 * Only set up a unix domain socket when they really asked for it.
494 * The service/port is different in that case.
495 */
496 continue;
497 }
498
499 /* See if there is still room to add 1 more socket. */
500 if (*NumListenSockets == MaxListen)
501 {
502 ereport(LOG,
503 (errmsg("could not bind to all requested addresses: MAXLISTEN (%d) exceeded",
504 MaxListen)));
505 break;
506 }
507
508 /* set up address family name for log messages */
509 switch (addr->ai_family)
510 {
511 case AF_INET:
512 familyDesc = _("IPv4");
513 break;
514 case AF_INET6:
515 familyDesc = _("IPv6");
516 break;
517 case AF_UNIX:
518 familyDesc = _("Unix");
519 break;
520 default:
521 snprintf(familyDescBuf, sizeof(familyDescBuf),
522 _("unrecognized address family %d"),
523 addr->ai_family);
524 familyDesc = familyDescBuf;
525 break;
526 }
527
528 /* set up text form of address for log messages */
529 if (addr->ai_family == AF_UNIX)
530 addrDesc = unixSocketPath;
531 else
532 {
533 pg_getnameinfo_all((const struct sockaddr_storage *) addr->ai_addr,
534 addr->ai_addrlen,
535 addrBuf, sizeof(addrBuf),
536 NULL, 0,
537 NI_NUMERICHOST);
538 addrDesc = addrBuf;
539 }
540
541 if ((fd = socket(addr->ai_family, SOCK_STREAM, 0)) == PGINVALID_SOCKET)
542 {
543 ereport(LOG,
544 (errcode_for_socket_access(),
545 /* translator: first %s is IPv4, IPv6, or Unix */
546 errmsg("could not create %s socket for address \"%s\": %m",
547 familyDesc, addrDesc)));
548 continue;
549 }
550
551 #ifndef WIN32
552 /* Don't give the listen socket to any subprograms we execute. */
553 if (fcntl(fd, F_SETFD, FD_CLOEXEC) < 0)
554 elog(FATAL, "fcntl(F_SETFD) failed on socket: %m");
555
556 /*
557 * Without the SO_REUSEADDR flag, a new postmaster can't be started
558 * right away after a stop or crash, giving "address already in use"
559 * error on TCP ports.
560 *
561 * On win32, however, this behavior only happens if the
562 * SO_EXCLUSIVEADDRUSE is set. With SO_REUSEADDR, win32 allows
563 * multiple servers to listen on the same address, resulting in
564 * unpredictable behavior. With no flags at all, win32 behaves as Unix
565 * with SO_REUSEADDR.
566 */
567 if (addr->ai_family != AF_UNIX)
568 {
569 if ((setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
570 (char *) &one, sizeof(one))) == -1)
571 {
572 ereport(LOG,
573 (errcode_for_socket_access(),
574 /* translator: third %s is IPv4 or IPv6 */
575 errmsg("%s(%s) failed for %s address \"%s\": %m",
576 "setsockopt", "SO_REUSEADDR",
577 familyDesc, addrDesc)));
578 closesocket(fd);
579 continue;
580 }
581 }
582 #endif
583
584 #ifdef IPV6_V6ONLY
585 if (addr->ai_family == AF_INET6)
586 {
587 if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY,
588 (char *) &one, sizeof(one)) == -1)
589 {
590 ereport(LOG,
591 (errcode_for_socket_access(),
592 /* translator: third %s is IPv6 */
593 errmsg("%s(%s) failed for %s address \"%s\": %m",
594 "setsockopt", "IPV6_V6ONLY",
595 familyDesc, addrDesc)));
596 closesocket(fd);
597 continue;
598 }
599 }
600 #endif
601
602 /*
603 * Note: This might fail on some OS's, like Linux older than
604 * 2.4.21-pre3, that don't have the IPV6_V6ONLY socket option, and map
605 * ipv4 addresses to ipv6. It will show ::ffff:ipv4 for all ipv4
606 * connections.
607 */
608 err = bind(fd, addr->ai_addr, addr->ai_addrlen);
609 if (err < 0)
610 {
611 int saved_errno = errno;
612
613 ereport(LOG,
614 (errcode_for_socket_access(),
615 /* translator: first %s is IPv4, IPv6, or Unix */
616 errmsg("could not bind %s address \"%s\": %m",
617 familyDesc, addrDesc),
618 saved_errno == EADDRINUSE ?
619 (addr->ai_family == AF_UNIX ?
620 errhint("Is another postmaster already running on port %d?",
621 (int) portNumber) :
622 errhint("Is another postmaster already running on port %d?"
623 " If not, wait a few seconds and retry.",
624 (int) portNumber)) : 0));
625 closesocket(fd);
626 continue;
627 }
628
629 if (addr->ai_family == AF_UNIX)
630 {
631 if (Setup_AF_UNIX(service) != STATUS_OK)
632 {
633 closesocket(fd);
634 break;
635 }
636 }
637
638 /*
639 * Select appropriate accept-queue length limit. It seems reasonable
640 * to use a value similar to the maximum number of child processes
641 * that the postmaster will permit.
642 */
643 maxconn = MaxConnections * 2;
644
645 err = listen(fd, maxconn);
646 if (err < 0)
647 {
648 ereport(LOG,
649 (errcode_for_socket_access(),
650 /* translator: first %s is IPv4, IPv6, or Unix */
651 errmsg("could not listen on %s address \"%s\": %m",
652 familyDesc, addrDesc)));
653 closesocket(fd);
654 continue;
655 }
656
657 if (addr->ai_family == AF_UNIX)
658 ereport(LOG,
659 (errmsg("listening on Unix socket \"%s\"",
660 addrDesc)));
661 else
662 ereport(LOG,
663 /* translator: first %s is IPv4 or IPv6 */
664 (errmsg("listening on %s address \"%s\", port %d",
665 familyDesc, addrDesc, (int) portNumber)));
666
667 ListenSockets[*NumListenSockets] = fd;
668 (*NumListenSockets)++;
669 added++;
670 }
671
672 pg_freeaddrinfo_all(hint.ai_family, addrs);
673
674 if (!added)
675 return STATUS_ERROR;
676
677 return STATUS_OK;
678 }
679
680
681 /*
682 * Lock_AF_UNIX -- configure unix socket file path
683 */
684 static int
685 Lock_AF_UNIX(const char *unixSocketDir, const char *unixSocketPath)
686 {
687 /* no lock file for abstract sockets */
688 if (unixSocketPath[0] == '@')
689 return STATUS_OK;
690
691 /*
692 * Grab an interlock file associated with the socket file.
693 *
694 * Note: there are two reasons for using a socket lock file, rather than
695 * trying to interlock directly on the socket itself. First, it's a lot
696 * more portable, and second, it lets us remove any pre-existing socket
697 * file without race conditions.
698 */
699 CreateSocketLockFile(unixSocketPath, true, unixSocketDir);
700
701 /*
702 * Once we have the interlock, we can safely delete any pre-existing
703 * socket file to avoid failure at bind() time.
704 */
705 (void) unlink(unixSocketPath);
706
707 /*
708 * Remember socket file pathnames for later maintenance.
709 */
710 sock_paths = lappend(sock_paths, pstrdup(unixSocketPath));
711
712 return STATUS_OK;
713 }
714
715
716 /*
717 * Setup_AF_UNIX -- configure unix socket permissions
718 */
719 static int
720 Setup_AF_UNIX(const char *sock_path)
721 {
722 /* no file system permissions for abstract sockets */
723 if (sock_path[0] == '@')
724 return STATUS_OK;
725
726 /*
727 * Fix socket ownership/permission if requested. Note we must do this
728 * before we listen() to avoid a window where unwanted connections could
729 * get accepted.
730 */
731 Assert(Unix_socket_group);
732 if (Unix_socket_group[0] != '\0')
733 {
734 #ifdef WIN32
735 elog(WARNING, "configuration item \"unix_socket_group\" is not supported on this platform");
736 #else
737 char *endptr;
738 unsigned long val;
739 gid_t gid;
740
741 val = strtoul(Unix_socket_group, &endptr, 10);
742 if (*endptr == '\0')
743 { /* numeric group id */
744 gid = val;
745 }
746 else
747 { /* convert group name to id */
748 struct group *gr;
749
750 gr = getgrnam(Unix_socket_group);
751 if (!gr)
752 {
753 ereport(LOG,
754 (errmsg("group \"%s\" does not exist",
755 Unix_socket_group)));
756 return STATUS_ERROR;
757 }
758 gid = gr->gr_gid;
759 }
760 if (chown(sock_path, -1, gid) == -1)
761 {
762 ereport(LOG,
763 (errcode_for_file_access(),
764 errmsg("could not set group of file \"%s\": %m",
765 sock_path)));
766 return STATUS_ERROR;
767 }
768 #endif
769 }
770
771 if (chmod(sock_path, Unix_socket_permissions) == -1)
772 {
773 ereport(LOG,
774 (errcode_for_file_access(),
775 errmsg("could not set permissions of file \"%s\": %m",
776 sock_path)));
777 return STATUS_ERROR;
778 }
779 return STATUS_OK;
780 }
781
782
783 /*
784 * AcceptConnection -- accept a new connection with client using
785 * server port. Fills *client_sock with the FD and endpoint info
786 * of the new connection.
787 *
788 * ASSUME: that this doesn't need to be non-blocking because
789 * the Postmaster waits for the socket to be ready to accept().
790 *
791 * RETURNS: STATUS_OK or STATUS_ERROR
792 */
793 int
794 AcceptConnection(pgsocket server_fd, ClientSocket *client_sock)
795 {
796 /* accept connection and fill in the client (remote) address */
797 client_sock->raddr.salen = sizeof(client_sock->raddr.addr);
798 if ((client_sock->sock = accept(server_fd,
799 (struct sockaddr *) &client_sock->raddr.addr,
800 &client_sock->raddr.salen)) == PGINVALID_SOCKET)
801 {
802 ereport(LOG,
803 (errcode_for_socket_access(),
804 errmsg("could not accept new connection: %m")));
805
806 /*
807 * If accept() fails then postmaster.c will still see the server
808 * socket as read-ready, and will immediately try again. To avoid
809 * uselessly sucking lots of CPU, delay a bit before trying again.
810 * (The most likely reason for failure is being out of kernel file
811 * table slots; we can do little except hope some will get freed up.)
812 */
813 pg_usleep(100000L); /* wait 0.1 sec */
814 return STATUS_ERROR;
815 }
816
817 return STATUS_OK;
818 }
819
820 /*
821 * TouchSocketFiles -- mark socket files as recently accessed
822 *
823 * This routine should be called every so often to ensure that the socket
824 * files have a recent mod date (ordinary operations on sockets usually won't
825 * change the mod date). That saves them from being removed by
826 * overenthusiastic /tmp-directory-cleaner daemons. (Another reason we should
827 * never have put the socket file in /tmp...)
828 */
829 void
830 TouchSocketFiles(void)
831 {
832 ListCell *l;
833
834 /* Loop through all created sockets... */
835 foreach(l, sock_paths)
836 {
837 char *sock_path = (char *) lfirst(l);
838
839 /* Ignore errors; there's no point in complaining */
840 (void) utime(sock_path, NULL);
841 }
842 }
843
844 /*
845 * RemoveSocketFiles -- unlink socket files at postmaster shutdown
846 */
847 void
848 RemoveSocketFiles(void)
849 {
850 ListCell *l;
851
852 /* Loop through all created sockets... */
853 foreach(l, sock_paths)
854 {
855 char *sock_path = (char *) lfirst(l);
856
857 /* Ignore any error. */
858 (void) unlink(sock_path);
859 }
860 /* Since we're about to exit, no need to reclaim storage */
861 sock_paths = NIL;
862 }
863
864
865 /* --------------------------------
866 * Low-level I/O routines begin here.
867 *
868 * These routines communicate with a frontend client across a connection
869 * already established by the preceding routines.
870 * --------------------------------
871 */
872
873 /* --------------------------------
874 * socket_set_nonblocking - set socket blocking/non-blocking
875 *
876 * Sets the socket non-blocking if nonblocking is true, or sets it
877 * blocking otherwise.
878 * --------------------------------
879 */
880 static void
881 socket_set_nonblocking(bool nonblocking)
882 {
883 if (MyProcPort == NULL)
884 ereport(ERROR,
885 (errcode(ERRCODE_CONNECTION_DOES_NOT_EXIST),
886 errmsg("there is no client connection")));
887
888 MyProcPort->noblock = nonblocking;
889 }
890
891 /* --------------------------------
892 * pq_recvbuf - load some bytes into the input buffer
893 *
894 * returns 0 if OK, EOF if trouble
895 * --------------------------------
896 */
897 static int
898 pq_recvbuf(void)
899 {
900 if (PqRecvPointer > 0)
901 {
902 if (PqRecvLength > PqRecvPointer)
903 {
904 /* still some unread data, left-justify it in the buffer */
905 memmove(PqRecvBuffer, PqRecvBuffer + PqRecvPointer,
906 PqRecvLength - PqRecvPointer);
907 PqRecvLength -= PqRecvPointer;
908 PqRecvPointer = 0;
909 }
910 else
911 PqRecvLength = PqRecvPointer = 0;
912 }
913
914 /* Ensure that we're in blocking mode */
915 socket_set_nonblocking(false);
916
917 /* Can fill buffer from PqRecvLength and upwards */
918 for (;;)
919 {
920 int r;
921
922 errno = 0;
923
924 r = secure_read(MyProcPort, PqRecvBuffer + PqRecvLength,
925 PQ_RECV_BUFFER_SIZE - PqRecvLength);
926
927 if (r < 0)
928 {
929 if (errno == EINTR)
930 continue; /* Ok if interrupted */
931
932 /*
933 * Careful: an ereport() that tries to write to the client would
934 * cause recursion to here, leading to stack overflow and core
935 * dump! This message must go *only* to the postmaster log.
936 *
937 * If errno is zero, assume it's EOF and let the caller complain.
938 */
939 if (errno != 0)
940 ereport(COMMERROR,
941 (errcode_for_socket_access(),
942 errmsg("could not receive data from client: %m")));
943 return EOF;
944 }
945 if (r == 0)
946 {
947 /*
948 * EOF detected. We used to write a log message here, but it's
949 * better to expect the ultimate caller to do that.
950 */
951 return EOF;
952 }
953 /* r contains number of bytes read, so just incr length */
954 PqRecvLength += r;
955 return 0;
956 }
957 }
958
959 /* --------------------------------
960 * pq_getbyte - get a single byte from connection, or return EOF
961 * --------------------------------
962 */
963 int
964 pq_getbyte(void)
965 {
966 Assert(PqCommReadingMsg);
967
968 while (PqRecvPointer >= PqRecvLength)
969 {
970 if (pq_recvbuf()) /* If nothing in buffer, then recv some */
971 return EOF; /* Failed to recv data */
972 }
973 return (unsigned char) PqRecvBuffer[PqRecvPointer++];
974 }
975
976 /* --------------------------------
977 * pq_peekbyte - peek at next byte from connection
978 *
979 * Same as pq_getbyte() except we don't advance the pointer.
980 * --------------------------------
981 */
982 int
983 pq_peekbyte(void)
984 {
985 Assert(PqCommReadingMsg);
986
987 while (PqRecvPointer >= PqRecvLength)
988 {
989 if (pq_recvbuf()) /* If nothing in buffer, then recv some */
990 return EOF; /* Failed to recv data */
991 }
992 return (unsigned char) PqRecvBuffer[PqRecvPointer];
993 }
994
995 /* --------------------------------
996 * pq_getbyte_if_available - get a single byte from connection,
997 * if available
998 *
999 * The received byte is stored in *c. Returns 1 if a byte was read,
1000 * 0 if no data was available, or EOF if trouble.
1001 * --------------------------------
1002 */
1003 int
1004 pq_getbyte_if_available(unsigned char *c)
1005 {
1006 int r;
1007
1008 Assert(PqCommReadingMsg);
1009
1010 if (PqRecvPointer < PqRecvLength)
1011 {
1012 *c = PqRecvBuffer[PqRecvPointer++];
1013 return 1;
1014 }
1015
1016 /* Put the socket into non-blocking mode */
1017 socket_set_nonblocking(true);
1018
1019 errno = 0;
1020
1021 r = secure_read(MyProcPort, c, 1);
1022 if (r < 0)
1023 {
1024 /*
1025 * Ok if no data available without blocking or interrupted (though
1026 * EINTR really shouldn't happen with a non-blocking socket). Report
1027 * other errors.
1028 */
1029 if (errno == EAGAIN || errno == EWOULDBLOCK || errno == EINTR)
1030 r = 0;
1031 else
1032 {
1033 /*
1034 * Careful: an ereport() that tries to write to the client would
1035 * cause recursion to here, leading to stack overflow and core
1036 * dump! This message must go *only* to the postmaster log.
1037 *
1038 * If errno is zero, assume it's EOF and let the caller complain.
1039 */
1040 if (errno != 0)
1041 ereport(COMMERROR,
1042 (errcode_for_socket_access(),
1043 errmsg("could not receive data from client: %m")));
1044 r = EOF;
1045 }
1046 }
1047 else if (r == 0)
1048 {
1049 /* EOF detected */
1050 r = EOF;
1051 }
1052
1053 return r;
1054 }
1055
1056 /* --------------------------------
1057 * pq_getbytes - get a known number of bytes from connection
1058 *
1059 * returns 0 if OK, EOF if trouble
1060 * --------------------------------
1061 */
1062 int
1063 pq_getbytes(void *b, size_t len)
1064 {
1065 char *s = b;
1066 size_t amount;
1067
1068 Assert(PqCommReadingMsg);
1069
1070 while (len > 0)
1071 {
1072 while (PqRecvPointer >= PqRecvLength)
1073 {
1074 if (pq_recvbuf()) /* If nothing in buffer, then recv some */
1075 return EOF; /* Failed to recv data */
1076 }
1077 amount = PqRecvLength - PqRecvPointer;
1078 if (amount > len)
1079 amount = len;
1080 memcpy(s, PqRecvBuffer + PqRecvPointer, amount);
1081 PqRecvPointer += amount;
1082 s += amount;
1083 len -= amount;
1084 }
1085 return 0;
1086 }
1087
1088 /* --------------------------------
1089 * pq_discardbytes - throw away a known number of bytes
1090 *
1091 * same as pq_getbytes except we do not copy the data to anyplace.
1092 * this is used for resynchronizing after read errors.
1093 *
1094 * returns 0 if OK, EOF if trouble
1095 * --------------------------------
1096 */
1097 static int
1098 pq_discardbytes(size_t len)
1099 {
1100 size_t amount;
1101
1102 Assert(PqCommReadingMsg);
1103
1104 while (len > 0)
1105 {
1106 while (PqRecvPointer >= PqRecvLength)
1107 {
1108 if (pq_recvbuf()) /* If nothing in buffer, then recv some */
1109 return EOF; /* Failed to recv data */
1110 }
1111 amount = PqRecvLength - PqRecvPointer;
1112 if (amount > len)
1113 amount = len;
1114 PqRecvPointer += amount;
1115 len -= amount;
1116 }
1117 return 0;
1118 }
1119
1120 /* --------------------------------
1121 * pq_buffer_remaining_data - return number of bytes in receive buffer
1122 *
1123 * This will *not* attempt to read more data. And reading up to that number of
1124 * bytes should not cause reading any more data either.
1125 * --------------------------------
1126 */
1127 ssize_t
1128 pq_buffer_remaining_data(void)
1129 {
1130 Assert(PqRecvLength >= PqRecvPointer);
1131 return (PqRecvLength - PqRecvPointer);
1132 }
1133
1134
1135 /* --------------------------------
1136 * pq_startmsgread - begin reading a message from the client.
1137 *
1138 * This must be called before any of the pq_get* functions.
1139 * --------------------------------
1140 */
1141 void
1142 pq_startmsgread(void)
1143 {
1144 /*
1145 * There shouldn't be a read active already, but let's check just to be
1146 * sure.
1147 */
1148 if (PqCommReadingMsg)
1149 ereport(FATAL,
1150 (errcode(ERRCODE_PROTOCOL_VIOLATION),
1151 errmsg("terminating connection because protocol synchronization was lost")));
1152
1153 PqCommReadingMsg = true;
1154 }
1155
1156
1157 /* --------------------------------
1158 * pq_endmsgread - finish reading message.
1159 *
1160 * This must be called after reading a message with pq_getbytes()
1161 * and friends, to indicate that we have read the whole message.
1162 * pq_getmessage() does this implicitly.
1163 * --------------------------------
1164 */
1165 void
1166 pq_endmsgread(void)
1167 {
1168 Assert(PqCommReadingMsg);
1169
1170 PqCommReadingMsg = false;
1171 }
1172
1173 /* --------------------------------
1174 * pq_is_reading_msg - are we currently reading a message?
1175 *
1176 * This is used in error recovery at the outer idle loop to detect if we have
1177 * lost protocol sync, and need to terminate the connection. pq_startmsgread()
1178 * will check for that too, but it's nicer to detect it earlier.
1179 * --------------------------------
1180 */
1181 bool
1182 pq_is_reading_msg(void)
1183 {
1184 return PqCommReadingMsg;
1185 }
1186
1187 /* --------------------------------
1188 * pq_getmessage - get a message with length word from connection
1189 *
1190 * The return value is placed in an expansible StringInfo, which has
1191 * already been initialized by the caller.
1192 * Only the message body is placed in the StringInfo; the length word
1193 * is removed. Also, s->cursor is initialized to zero for convenience
1194 * in scanning the message contents.
1195 *
1196 * maxlen is the upper limit on the length of the
1197 * message we are willing to accept. We abort the connection (by
1198 * returning EOF) if client tries to send more than that.
1199 *
1200 * returns 0 if OK, EOF if trouble
1201 * --------------------------------
1202 */
1203 int
1204 pq_getmessage(StringInfo s, int maxlen)
1205 {
1206 int32 len;
1207
1208 Assert(PqCommReadingMsg);
1209
1210 resetStringInfo(s);
1211
1212 /* Read message length word */
1213 if (pq_getbytes(&len, 4) == EOF)
1214 {
1215 ereport(COMMERROR,
1216 (errcode(ERRCODE_PROTOCOL_VIOLATION),
1217 errmsg("unexpected EOF within message length word")));
1218 return EOF;
1219 }
1220
1221 len = pg_ntoh32(len);
1222
1223 if (len < 4 || len > maxlen)
1224 {
1225 ereport(COMMERROR,
1226 (errcode(ERRCODE_PROTOCOL_VIOLATION),
1227 errmsg("invalid message length")));
1228 return EOF;
1229 }
1230
1231 len -= 4; /* discount length itself */
1232
1233 if (len > 0)
1234 {
1235 /*
1236 * Allocate space for message. If we run out of room (ridiculously
1237 * large message), we will elog(ERROR), but we want to discard the
1238 * message body so as not to lose communication sync.
1239 */
1240 PG_TRY();
1241 {
1242 enlargeStringInfo(s, len);
1243 }
1244 PG_CATCH();
1245 {
1246 if (pq_discardbytes(len) == EOF)
1247 ereport(COMMERROR,
1248 (errcode(ERRCODE_PROTOCOL_VIOLATION),
1249 errmsg("incomplete message from client")));
1250
1251 /* we discarded the rest of the message so we're back in sync. */
1252 PqCommReadingMsg = false;
1253 PG_RE_THROW();
1254 }
1255 PG_END_TRY();
1256
1257 /* And grab the message */
1258 if (pq_getbytes(s->data, len) == EOF)
1259 {
1260 ereport(COMMERROR,
1261 (errcode(ERRCODE_PROTOCOL_VIOLATION),
1262 errmsg("incomplete message from client")));
1263 return EOF;
1264 }
1265 s->len = len;
1266 /* Place a trailing null per StringInfo convention */
1267 s->data[len] = '\0';
1268 }
1269
1270 /* finished reading the message. */
1271 PqCommReadingMsg = false;
1272
1273 return 0;
1274 }
1275
1276
1277 static inline int
1278 internal_putbytes(const void *b, size_t len)
1279 {
1280 const char *s = b;
1281
1282 while (len > 0)
1283 {
1284 /* If buffer is full, then flush it out */
1285 if (PqSendPointer >= PqSendBufferSize)
1286 {
1287 socket_set_nonblocking(false);
1288 if (internal_flush())
1289 return EOF;
1290 }
1291
1292 /*
1293 * If the buffer is empty and data length is larger than the buffer
1294 * size, send it without buffering. Otherwise, copy as much data as
1295 * possible into the buffer.
1296 */
1297 if (len >= PqSendBufferSize && PqSendStart == PqSendPointer)
1298 {
1299 size_t start = 0;
1300
1301 socket_set_nonblocking(false);
1302 if (internal_flush_buffer(s, &start, &len))
1303 return EOF;
1304 }
1305 else
1306 {
1307 size_t amount = PqSendBufferSize - PqSendPointer;
1308
1309 if (amount > len)
1310 amount = len;
1311 memcpy(PqSendBuffer + PqSendPointer, s, amount);
1312 PqSendPointer += amount;
1313 s += amount;
1314 len -= amount;
1315 }
1316 }
1317
1318 return 0;
1319 }
1320
1321 /* --------------------------------
1322 * socket_flush - flush pending output
1323 *
1324 * returns 0 if OK, EOF if trouble
1325 * --------------------------------
1326 */
1327 static int
1328 socket_flush(void)
1329 {
1330 int res;
1331
1332 /* No-op if reentrant call */
1333 if (PqCommBusy)
1334 return 0;
1335 PqCommBusy = true;
1336 socket_set_nonblocking(false);
1337 res = internal_flush();
1338 PqCommBusy = false;
1339 return res;
1340 }
1341
1342 /* --------------------------------
1343 * internal_flush - flush pending output
1344 *
1345 * Returns 0 if OK (meaning everything was sent, or operation would block
1346 * and the socket is in non-blocking mode), or EOF if trouble.
1347 * --------------------------------
1348 */
1349 static inline int
1350 internal_flush(void)
1351 {
1352 return internal_flush_buffer(PqSendBuffer, &PqSendStart, &PqSendPointer);
1353 }
1354
1355 /* --------------------------------
1356 * internal_flush_buffer - flush the given buffer content
1357 *
1358 * Returns 0 if OK (meaning everything was sent, or operation would block
1359 * and the socket is in non-blocking mode), or EOF if trouble.
1360 * --------------------------------
1361 */
1362 static pg_noinline int
1363 internal_flush_buffer(const char *buf, size_t *start, size_t *end)
1364 {
1365 static int last_reported_send_errno = 0;
1366
1367 const char *bufptr = buf + *start;
1368 const char *bufend = buf + *end;
1369
1370 while (bufptr < bufend)
1371 {
1372 int r;
1373
1374 r = secure_write(MyProcPort, bufptr, bufend - bufptr);
1375
1376 if (r <= 0)
1377 {
1378 if (errno == EINTR)
1379 continue; /* Ok if we were interrupted */
1380
1381 /*
1382 * Ok if no data writable without blocking, and the socket is in
1383 * non-blocking mode.
1384 */
1385 if (errno == EAGAIN ||
1386 errno == EWOULDBLOCK)
1387 {
1388 return 0;
1389 }
1390
1391 /*
1392 * Careful: an ereport() that tries to write to the client would
1393 * cause recursion to here, leading to stack overflow and core
1394 * dump! This message must go *only* to the postmaster log.
1395 *
1396 * If a client disconnects while we're in the midst of output, we
1397 * might write quite a bit of data before we get to a safe query
1398 * abort point. So, suppress duplicate log messages.
1399 */
1400 if (errno != last_reported_send_errno)
1401 {
1402 last_reported_send_errno = errno;
1403 ereport(COMMERROR,
1404 (errcode_for_socket_access(),
1405 errmsg("could not send data to client: %m")));
1406 }
1407
1408 /*
1409 * We drop the buffered data anyway so that processing can
1410 * continue, even though we'll probably quit soon. We also set a
1411 * flag that'll cause the next CHECK_FOR_INTERRUPTS to terminate
1412 * the connection.
1413 */
1414 *start = *end = 0;
1415 ClientConnectionLost = 1;
1416 InterruptPending = 1;
1417 return EOF;
1418 }
1419
1420 last_reported_send_errno = 0; /* reset after any successful send */
1421 bufptr += r;
1422 *start += r;
1423 }
1424
1425 *start = *end = 0;
1426 return 0;
1427 }
1428
1429 /* --------------------------------
1430 * pq_flush_if_writable - flush pending output if writable without blocking
1431 *
1432 * Returns 0 if OK, or EOF if trouble.
1433 * --------------------------------
1434 */
1435 static int
1436 socket_flush_if_writable(void)
1437 {
1438 int res;
1439
1440 /* Quick exit if nothing to do */
1441 if (PqSendPointer == PqSendStart)
1442 return 0;
1443
1444 /* No-op if reentrant call */
1445 if (PqCommBusy)
1446 return 0;
1447
1448 /* Temporarily put the socket into non-blocking mode */
1449 socket_set_nonblocking(true);
1450
1451 PqCommBusy = true;
1452 res = internal_flush();
1453 PqCommBusy = false;
1454 return res;
1455 }
1456
1457 /* --------------------------------
1458 * socket_is_send_pending - is there any pending data in the output buffer?
1459 * --------------------------------
1460 */
1461 static bool
1462 socket_is_send_pending(void)
1463 {
1464 return (PqSendStart < PqSendPointer);
1465 }
1466
1467 /* --------------------------------
1468 * Message-level I/O routines begin here.
1469 * --------------------------------
1470 */
1471
1472
1473 /* --------------------------------
1474 * socket_putmessage - send a normal message (suppressed in COPY OUT mode)
1475 *
1476 * msgtype is a message type code to place before the message body.
1477 *
1478 * len is the length of the message body data at *s. A message length
1479 * word (equal to len+4 because it counts itself too) is inserted by this
1480 * routine.
1481 *
1482 * We suppress messages generated while pqcomm.c is busy. This
1483 * avoids any possibility of messages being inserted within other
1484 * messages. The only known trouble case arises if SIGQUIT occurs
1485 * during a pqcomm.c routine --- quickdie() will try to send a warning
1486 * message, and the most reasonable approach seems to be to drop it.
1487 *
1488 * returns 0 if OK, EOF if trouble
1489 * --------------------------------
1490 */
1491 static int
1492 socket_putmessage(char msgtype, const char *s, size_t len)
1493 {
1494 uint32 n32;
1495
1496 Assert(msgtype != 0);
1497
1498 if (PqCommBusy)
1499 return 0;
1500 PqCommBusy = true;
1501 if (internal_putbytes(&msgtype, 1))
1502 goto fail;
1503
1504 n32 = pg_hton32((uint32) (len + 4));
1505 if (internal_putbytes(&n32, 4))
1506 goto fail;
1507
1508 if (internal_putbytes(s, len))
1509 goto fail;
1510 PqCommBusy = false;
1511 return 0;
1512
1513 fail:
1514 PqCommBusy = false;
1515 return EOF;
1516 }
1517
1518 /* --------------------------------
1519 * pq_putmessage_noblock - like pq_putmessage, but never blocks
1520 *
1521 * If the output buffer is too small to hold the message, the buffer
1522 * is enlarged.
1523 */
1524 static void
1525 socket_putmessage_noblock(char msgtype, const char *s, size_t len)
1526 {
1527 int res PG_USED_FOR_ASSERTS_ONLY;
1528 int required;
1529
1530 /*
1531 * Ensure we have enough space in the output buffer for the message header
1532 * as well as the message itself.
1533 */
1534 required = PqSendPointer + 1 + 4 + len;
1535 if (required > PqSendBufferSize)
1536 {
1537 PqSendBuffer = repalloc(PqSendBuffer, required);
1538 PqSendBufferSize = required;
1539 }
1540 res = pq_putmessage(msgtype, s, len);
1541 Assert(res == 0); /* should not fail when the message fits in
1542 * buffer */
1543 }
1544
1545 /* --------------------------------
1546 * pq_putmessage_v2 - send a message in protocol version 2
1547 *
1548 * msgtype is a message type code to place before the message body.
1549 *
1550 * We no longer support protocol version 2, but we have kept this
1551 * function so that if a client tries to connect with protocol version 2,
1552 * as a courtesy we can still send the "unsupported protocol version"
1553 * error to the client in the old format.
1554 *
1555 * Like in pq_putmessage(), we suppress messages generated while
1556 * pqcomm.c is busy.
1557 *
1558 * returns 0 if OK, EOF if trouble
1559 * --------------------------------
1560 */
1561 int
1562 pq_putmessage_v2(char msgtype, const char *s, size_t len)
1563 {
1564 Assert(msgtype != 0);
1565
1566 if (PqCommBusy)
1567 return 0;
1568 PqCommBusy = true;
1569 if (internal_putbytes(&msgtype, 1))
1570 goto fail;
1571
1572 if (internal_putbytes(s, len))
1573 goto fail;
1574 PqCommBusy = false;
1575 return 0;
1576
1577 fail:
1578 PqCommBusy = false;
1579 return EOF;
1580 }
1581
1582 /*
1583 * Support for TCP Keepalive parameters
1584 */
1585
1586 /*
1587 * On Windows, we need to set both idle and interval at the same time.
1588 * We also cannot reset them to the default (setting to zero will
1589 * actually set them to zero, not default), therefore we fallback to
1590 * the out-of-the-box default instead.
1591 */
1592 #if defined(WIN32) && defined(SIO_KEEPALIVE_VALS)
1593 static int
1594 pq_setkeepaliveswin32(Port *port, int idle, int interval)
1595 {
1596 struct tcp_keepalive ka;
1597 DWORD retsize;
1598
1599 if (idle <= 0)
1600 idle = 2 * 60 * 60; /* default = 2 hours */
1601 if (interval <= 0)
1602 interval = 1; /* default = 1 second */
1603
1604 ka.onoff = 1;
1605 ka.keepalivetime = idle * 1000;
1606 ka.keepaliveinterval = interval * 1000;
1607
1608 if (WSAIoctl(port->sock,
1609 SIO_KEEPALIVE_VALS,
1610 (LPVOID) &ka,
1611 sizeof(ka),
1612 NULL,
1613 0,
1614 &retsize,
1615 NULL,
1616 NULL)
1617 != 0)
1618 {
1619 ereport(LOG,
1620 (errmsg("%s(%s) failed: error code %d",
1621 "WSAIoctl", "SIO_KEEPALIVE_VALS", WSAGetLastError())));
1622 return STATUS_ERROR;
1623 }
1624 if (port->keepalives_idle != idle)
1625 port->keepalives_idle = idle;
1626 if (port->keepalives_interval != interval)
1627 port->keepalives_interval = interval;
1628 return STATUS_OK;
1629 }
1630 #endif
1631
1632 int
1633 pq_getkeepalivesidle(Port *port)
1634 {
1635 #if defined(PG_TCP_KEEPALIVE_IDLE) || defined(SIO_KEEPALIVE_VALS)
1636 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1637 return 0;
1638
1639 if (port->keepalives_idle != 0)
1640 return port->keepalives_idle;
1641
1642 if (port->default_keepalives_idle == 0)
1643 {
1644 #ifndef WIN32
1645 socklen_t size = sizeof(port->default_keepalives_idle);
1646
1647 if (getsockopt(port->sock, IPPROTO_TCP, PG_TCP_KEEPALIVE_IDLE,
1648 (char *) &port->default_keepalives_idle,
1649 &size) < 0)
1650 {
1651 ereport(LOG,
1652 (errmsg("%s(%s) failed: %m", "getsockopt", PG_TCP_KEEPALIVE_IDLE_STR)));
1653 port->default_keepalives_idle = -1; /* don't know */
1654 }
1655 #else /* WIN32 */
1656 /* We can't get the defaults on Windows, so return "don't know" */
1657 port->default_keepalives_idle = -1;
1658 #endif /* WIN32 */
1659 }
1660
1661 return port->default_keepalives_idle;
1662 #else
1663 return 0;
1664 #endif
1665 }
1666
1667 int
1668 pq_setkeepalivesidle(int idle, Port *port)
1669 {
1670 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1671 return STATUS_OK;
1672
1673 /* check SIO_KEEPALIVE_VALS here, not just WIN32, as some toolchains lack it */
1674 #if defined(PG_TCP_KEEPALIVE_IDLE) || defined(SIO_KEEPALIVE_VALS)
1675 if (idle == port->keepalives_idle)
1676 return STATUS_OK;
1677
1678 #ifndef WIN32
1679 if (port->default_keepalives_idle <= 0)
1680 {
1681 if (pq_getkeepalivesidle(port) < 0)
1682 {
1683 if (idle == 0)
1684 return STATUS_OK; /* default is set but unknown */
1685 else
1686 return STATUS_ERROR;
1687 }
1688 }
1689
1690 if (idle == 0)
1691 idle = port->default_keepalives_idle;
1692
1693 if (setsockopt(port->sock, IPPROTO_TCP, PG_TCP_KEEPALIVE_IDLE,
1694 (char *) &idle, sizeof(idle)) < 0)
1695 {
1696 ereport(LOG,
1697 (errmsg("%s(%s) failed: %m", "setsockopt", PG_TCP_KEEPALIVE_IDLE_STR)));
1698 return STATUS_ERROR;
1699 }
1700
1701 port->keepalives_idle = idle;
1702 #else /* WIN32 */
1703 return pq_setkeepaliveswin32(port, idle, port->keepalives_interval);
1704 #endif
1705 #else
1706 if (idle != 0)
1707 {
1708 ereport(LOG,
1709 (errmsg("setting the keepalive idle time is not supported")));
1710 return STATUS_ERROR;
1711 }
1712 #endif
1713
1714 return STATUS_OK;
1715 }
1716
1717 int
1718 pq_getkeepalivesinterval(Port *port)
1719 {
1720 #if defined(TCP_KEEPINTVL) || defined(SIO_KEEPALIVE_VALS)
1721 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1722 return 0;
1723
1724 if (port->keepalives_interval != 0)
1725 return port->keepalives_interval;
1726
1727 if (port->default_keepalives_interval == 0)
1728 {
1729 #ifndef WIN32
1730 socklen_t size = sizeof(port->default_keepalives_interval);
1731
1732 if (getsockopt(port->sock, IPPROTO_TCP, TCP_KEEPINTVL,
1733 (char *) &port->default_keepalives_interval,
1734 &size) < 0)
1735 {
1736 ereport(LOG,
1737 (errmsg("%s(%s) failed: %m", "getsockopt", "TCP_KEEPINTVL")));
1738 port->default_keepalives_interval = -1; /* don't know */
1739 }
1740 #else
1741 /* We can't get the defaults on Windows, so return "don't know" */
1742 port->default_keepalives_interval = -1;
1743 #endif /* WIN32 */
1744 }
1745
1746 return port->default_keepalives_interval;
1747 #else
1748 return 0;
1749 #endif
1750 }
1751
1752 int
1753 pq_setkeepalivesinterval(int interval, Port *port)
1754 {
1755 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1756 return STATUS_OK;
1757
1758 #if defined(TCP_KEEPINTVL) || defined(SIO_KEEPALIVE_VALS)
1759 if (interval == port->keepalives_interval)
1760 return STATUS_OK;
1761
1762 #ifndef WIN32
1763 if (port->default_keepalives_interval <= 0)
1764 {
1765 if (pq_getkeepalivesinterval(port) < 0)
1766 {
1767 if (interval == 0)
1768 return STATUS_OK; /* default is set but unknown */
1769 else
1770 return STATUS_ERROR;
1771 }
1772 }
1773
1774 if (interval == 0)
1775 interval = port->default_keepalives_interval;
1776
1777 if (setsockopt(port->sock, IPPROTO_TCP, TCP_KEEPINTVL,
1778 (char *) &interval, sizeof(interval)) < 0)
1779 {
1780 ereport(LOG,
1781 (errmsg("%s(%s) failed: %m", "setsockopt", "TCP_KEEPINTVL")));
1782 return STATUS_ERROR;
1783 }
1784
1785 port->keepalives_interval = interval;
1786 #else /* WIN32 */
1787 return pq_setkeepaliveswin32(port, port->keepalives_idle, interval);
1788 #endif
1789 #else
1790 if (interval != 0)
1791 {
1792 ereport(LOG,
1793 (errmsg("%s(%s) not supported", "setsockopt", "TCP_KEEPINTVL")));
1794 return STATUS_ERROR;
1795 }
1796 #endif
1797
1798 return STATUS_OK;
1799 }
1800
1801 int
1802 pq_getkeepalivescount(Port *port)
1803 {
1804 #ifdef TCP_KEEPCNT
1805 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1806 return 0;
1807
1808 if (port->keepalives_count != 0)
1809 return port->keepalives_count;
1810
1811 if (port->default_keepalives_count == 0)
1812 {
1813 socklen_t size = sizeof(port->default_keepalives_count);
1814
1815 if (getsockopt(port->sock, IPPROTO_TCP, TCP_KEEPCNT,
1816 (char *) &port->default_keepalives_count,
1817 &size) < 0)
1818 {
1819 ereport(LOG,
1820 (errmsg("%s(%s) failed: %m", "getsockopt", "TCP_KEEPCNT")));
1821 port->default_keepalives_count = -1; /* don't know */
1822 }
1823 }
1824
1825 return port->default_keepalives_count;
1826 #else
1827 return 0;
1828 #endif
1829 }
1830
1831 int
1832 pq_setkeepalivescount(int count, Port *port)
1833 {
1834 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1835 return STATUS_OK;
1836
1837 #ifdef TCP_KEEPCNT
1838 if (count == port->keepalives_count)
1839 return STATUS_OK;
1840
1841 if (port->default_keepalives_count <= 0)
1842 {
1843 if (pq_getkeepalivescount(port) < 0)
1844 {
1845 if (count == 0)
1846 return STATUS_OK; /* default is set but unknown */
1847 else
1848 return STATUS_ERROR;
1849 }
1850 }
1851
1852 if (count == 0)
1853 count = port->default_keepalives_count;
1854
1855 if (setsockopt(port->sock, IPPROTO_TCP, TCP_KEEPCNT,
1856 (char *) &count, sizeof(count)) < 0)
1857 {
1858 ereport(LOG,
1859 (errmsg("%s(%s) failed: %m", "setsockopt", "TCP_KEEPCNT")));
1860 return STATUS_ERROR;
1861 }
1862
1863 port->keepalives_count = count;
1864 #else
1865 if (count != 0)
1866 {
1867 ereport(LOG,
1868 (errmsg("%s(%s) not supported", "setsockopt", "TCP_KEEPCNT")));
1869 return STATUS_ERROR;
1870 }
1871 #endif
1872
1873 return STATUS_OK;
1874 }
1875
1876 int
1877 pq_gettcpusertimeout(Port *port)
1878 {
1879 #ifdef TCP_USER_TIMEOUT
1880 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1881 return 0;
1882
1883 if (port->tcp_user_timeout != 0)
1884 return port->tcp_user_timeout;
1885
1886 if (port->default_tcp_user_timeout == 0)
1887 {
1888 socklen_t size = sizeof(port->default_tcp_user_timeout);
1889
1890 if (getsockopt(port->sock, IPPROTO_TCP, TCP_USER_TIMEOUT,
1891 (char *) &port->default_tcp_user_timeout,
1892 &size) < 0)
1893 {
1894 ereport(LOG,
1895 (errmsg("%s(%s) failed: %m", "getsockopt", "TCP_USER_TIMEOUT")));
1896 port->default_tcp_user_timeout = -1; /* don't know */
1897 }
1898 }
1899
1900 return port->default_tcp_user_timeout;
1901 #else
1902 return 0;
1903 #endif
1904 }
1905
1906 int
1907 pq_settcpusertimeout(int timeout, Port *port)
1908 {
1909 if (port == NULL || port->laddr.addr.ss_family == AF_UNIX)
1910 return STATUS_OK;
1911
1912 #ifdef TCP_USER_TIMEOUT
1913 if (timeout == port->tcp_user_timeout)
1914 return STATUS_OK;
1915
1916 if (port->default_tcp_user_timeout <= 0)
1917 {
1918 if (pq_gettcpusertimeout(port) < 0)
1919 {
1920 if (timeout == 0)
1921 return STATUS_OK; /* default is set but unknown */
1922 else
1923 return STATUS_ERROR;
1924 }
1925 }
1926
1927 if (timeout == 0)
1928 timeout = port->default_tcp_user_timeout;
1929
1930 if (setsockopt(port->sock, IPPROTO_TCP, TCP_USER_TIMEOUT,
1931 (char *) &timeout, sizeof(timeout)) < 0)
1932 {
1933 ereport(LOG,
1934 (errmsg("%s(%s) failed: %m", "setsockopt", "TCP_USER_TIMEOUT")));
1935 return STATUS_ERROR;
1936 }
1937
1938 port->tcp_user_timeout = timeout;
1939 #else
1940 if (timeout != 0)
1941 {
1942 ereport(LOG,
1943 (errmsg("%s(%s) not supported", "setsockopt", "TCP_USER_TIMEOUT")));
1944 return STATUS_ERROR;
1945 }
1946 #endif
1947
1948 return STATUS_OK;
1949 }
1950
1951 /*
1952 * GUC assign_hook for tcp_keepalives_idle
1953 */
1954 void
1955 assign_tcp_keepalives_idle(int newval, void *extra)
1956 {
1957 /*
1958 * The kernel API provides no way to test a value without setting it; and
1959 * once we set it we might fail to unset it. So there seems little point
1960 * in fully implementing the check-then-assign GUC API for these
1961 * variables. Instead we just do the assignment on demand.
1962 * pq_setkeepalivesidle reports any problems via ereport(LOG).
1963 *
1964 * This approach means that the GUC value might have little to do with the
1965 * actual kernel value, so we use a show_hook that retrieves the kernel
1966 * value rather than trusting GUC's copy.
1967 */
1968 (void) pq_setkeepalivesidle(newval, MyProcPort);
1969 }
1970
1971 /*
1972 * GUC show_hook for tcp_keepalives_idle
1973 */
1974 const char *
1975 show_tcp_keepalives_idle(void)
1976 {
1977 /* See comments in assign_tcp_keepalives_idle */
1978 static char nbuf[16];
1979
1980 snprintf(nbuf, sizeof(nbuf), "%d", pq_getkeepalivesidle(MyProcPort));
1981 return nbuf;
1982 }
1983
1984 /*
1985 * GUC assign_hook for tcp_keepalives_interval
1986 */
1987 void
1988 assign_tcp_keepalives_interval(int newval, void *extra)
1989 {
1990 /* See comments in assign_tcp_keepalives_idle */
1991 (void) pq_setkeepalivesinterval(newval, MyProcPort);
1992 }
1993
1994 /*
1995 * GUC show_hook for tcp_keepalives_interval
1996 */
1997 const char *
1998 show_tcp_keepalives_interval(void)
1999 {
2000 /* See comments in assign_tcp_keepalives_idle */
2001 static char nbuf[16];
2002
2003 snprintf(nbuf, sizeof(nbuf), "%d", pq_getkeepalivesinterval(MyProcPort));
2004 return nbuf;
2005 }
2006
2007 /*
2008 * GUC assign_hook for tcp_keepalives_count
2009 */
2010 void
2011 assign_tcp_keepalives_count(int newval, void *extra)
2012 {
2013 /* See comments in assign_tcp_keepalives_idle */
2014 (void) pq_setkeepalivescount(newval, MyProcPort);
2015 }
2016
2017 /*
2018 * GUC show_hook for tcp_keepalives_count
2019 */
2020 const char *
2021 show_tcp_keepalives_count(void)
2022 {
2023 /* See comments in assign_tcp_keepalives_idle */
2024 static char nbuf[16];
2025
2026 snprintf(nbuf, sizeof(nbuf), "%d", pq_getkeepalivescount(MyProcPort));
2027 return nbuf;
2028 }
2029
2030 /*
2031 * GUC assign_hook for tcp_user_timeout
2032 */
2033 void
2034 assign_tcp_user_timeout(int newval, void *extra)
2035 {
2036 /* See comments in assign_tcp_keepalives_idle */
2037 (void) pq_settcpusertimeout(newval, MyProcPort);
2038 }
2039
2040 /*
2041 * GUC show_hook for tcp_user_timeout
2042 */
2043 const char *
2044 show_tcp_user_timeout(void)
2045 {
2046 /* See comments in assign_tcp_keepalives_idle */
2047 static char nbuf[16];
2048
2049 snprintf(nbuf, sizeof(nbuf), "%d", pq_gettcpusertimeout(MyProcPort));
2050 return nbuf;
2051 }
2052
2053 /*
2054 * Check if the client is still connected.
2055 */
2056 bool
2057 pq_check_connection(void)
2058 {
2059 WaitEvent events[FeBeWaitSetNEvents];
2060 int rc;
2061
2062 /*
2063 * It's OK to modify the socket event filter without restoring, because
2064 * all FeBeWaitSet socket wait sites do the same.
2065 */
2066 ModifyWaitEvent(FeBeWaitSet, FeBeWaitSetSocketPos, WL_SOCKET_CLOSED, NULL);
2067
2068 retry:
2069 rc = WaitEventSetWait(FeBeWaitSet, 0, events, lengthof(events), 0);
2070 for (int i = 0; i < rc; ++i)
2071 {
2072 if (events[i].events & WL_SOCKET_CLOSED)
2073 return false;
2074 if (events[i].events & WL_LATCH_SET)
2075 {
2076 /*
2077 * A latch event might be preventing other events from being
2078 * reported. Reset it and poll again. No need to restore it
2079 * because no code should expect latches to survive across
2080 * CHECK_FOR_INTERRUPTS().
2081 */
2082 ResetLatch(MyLatch);
2083 goto retry;
2084 }
2085 }
2086
2087 return true;
2088 }