Mercurial > hg > xemacs-beta

comparison src/process-unix.c @ 263:727739f917cb r20-5b30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Import from CVS: tag r20-5b30
author cvs
date Mon, 13 Aug 2007 10:24:41 +0200
parents
children c5d627a313b1
comparison
equal deleted inserted replaced
262:9d8607af9e13 263:727739f917cb
1 /* Asynchronous subprocess implemenation for UNIX
2 Copyright (C) 1985, 1986, 1987, 1988, 1992, 1993, 1994, 1995
3 Free Software Foundation, Inc.
4 Copyright (C) 1995 Sun Microsystems, Inc.
5 Copyright (C) 1995, 1996 Ben Wing.
6
7 This file is part of XEmacs.
8
9 XEmacs is free software; you can redistribute it and/or modify it
10 under the terms of the GNU General Public License as published by the
11 Free Software Foundation; either version 2, or (at your option) any
12 later version.
13
14 XEmacs is distributed in the hope that it will be useful, but WITHOUT
15 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
16 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17 for more details.
18
19 You should have received a copy of the GNU General Public License
20 along with XEmacs; see the file COPYING. If not, write to
21 the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
22 Boston, MA 02111-1307, USA. */
23
24 /* This file has been Mule-ized except for `start-process-internal',
25 `open-network-stream-internal' and `open-multicast-group-internal'. */
26
27 /* This file has been split into process.c and process-unix.c by
28 Kirill M. Katsnelson <kkm@kis.ru>, so please bash him and not
29 the original author(s) */
30
31 #include <config.h>
32
33 #if !defined (NO_SUBPROCESSES)
34
35 /* The entire file is within this conditional */
36
37 #include "lisp.h"
38
39 #include "buffer.h"
40 #include "commands.h"
41 #include "events.h"
42 #include "frame.h"
43 #include "hash.h"
44 #include "insdel.h"
45 #include "lstream.h"
46 #include "opaque.h"
47 #include "process.h"
48 #include "procimpl.h"
49 #include "sysdep.h"
50 #include "window.h"
51 #ifdef FILE_CODING
52 #include "file-coding.h"
53 #endif
54
55 #include <setjmp.h>
56 #include "sysfile.h"
57 #include "sysproc.h"
58 #include "systime.h"
59 #include "syssignal.h" /* Always include before systty.h */
60 #include "systty.h"
61 #include "syswait.h"
62
63
64 /*
65 * Implemenation-specific data. Pointed to by Lisp_Process->process_data
66 */
67
68 struct unix_process_data
69 {
70 /* Non-0 if this is really a ToolTalk channel. */
71 int connected_via_filedesc_p;
72 /* Descriptor by which we read from this process. -1 for dead process */
73 int infd;
74 /* Descriptor for the tty which this process is using.
75 -1 if we didn't record it (on some systems, there's no need). */
76 int subtty;
77 /* Name of subprocess terminal. */
78 Lisp_Object tty_name;
79 /* Non-false if communicating through a pty. */
80 char pty_flag;
81 };
82
83 #define UNIX_DATA(p) ((struct unix_process_data*)((p)->process_data))
84
85 #ifdef HAVE_PTYS
86 /* The file name of the pty opened by allocate_pty. */
87
88 static char pty_name[24];
89 #endif
90
91
92
93 /**********************************************************************/
94 /* Static helper routines */
95 /**********************************************************************/
96
97 static SIGTYPE
98 close_safely_handler (int signo)
99 {
100 EMACS_REESTABLISH_SIGNAL (signo, close_safely_handler);
101 SIGRETURN;
102 }
103
104 static void
105 close_safely (int fd)
106 {
107 stop_interrupts ();
108 signal (SIGALRM, close_safely_handler);
109 alarm (1);
110 close (fd);
111 alarm (0);
112 start_interrupts ();
113 }
114
115 static void
116 close_descriptor_pair (int in, int out)
117 {
118 if (in >= 0)
119 close (in);
120 if (out != in && out >= 0)
121 close (out);
122 }
123
124 /* Close all descriptors currently in use for communication
125 with subprocess. This is used in a newly-forked subprocess
126 to get rid of irrelevant descriptors. */
127
128 static int
129 close_process_descs_mapfun (CONST void* key, void* contents, void* arg)
130 {
131 Lisp_Object proc;
132 CVOID_TO_LISP (proc, contents);
133 event_stream_delete_stream_pair (XPROCESS(proc)->pipe_instream,
134 XPROCESS(proc)->pipe_outstream);
135 return 0;
136 }
137
138 /* #### This function is currently called from child_setup
139 in callproc.c. It should become static though - kkm */
140 void
141 close_process_descs (void)
142 {
143 maphash (close_process_descs_mapfun, usid_to_process, 0);
144 }
145
146 /* connect to an existing file descriptor. This is very similar to
147 open-network-stream except that it assumes that the connection has
148 already been initialized. It is currently used for ToolTalk
149 communication. */
150
151 /* This function used to be visible on the Lisp level, but there is no
152 real point in doing that. Here is the doc string:
153
154 "Connect to an existing file descriptor.\n\
155 Returns a subprocess-object to represent the connection.\n\
156 Input and output work as for subprocesses; `delete-process' closes it.\n\
157 Args are NAME BUFFER INFD OUTFD.\n\
158 NAME is name for process. It is modified if necessary to make it unique.\n\
159 BUFFER is the buffer (or buffer-name) to associate with the process.\n\
160 Process output goes at end of that buffer, unless you specify\n\
161 an output stream or filter function to handle the output.\n\
162 BUFFER may be also nil, meaning that this process is not associated\n\
163 with any buffer\n\
164 INFD and OUTFD specify the file descriptors to use for input and\n\
165 output, respectively."
166 */
167
168 Lisp_Object
169 connect_to_file_descriptor (Lisp_Object name, Lisp_Object buffer,
170 Lisp_Object infd, Lisp_Object outfd)
171 {
172 /* This function can GC */
173 Lisp_Object proc;
174 int inch;
175
176 CHECK_STRING (name);
177 CHECK_INT (infd);
178 CHECK_INT (outfd);
179
180 inch = XINT (infd);
181 if (get_process_from_usid (FD_TO_USID(inch)))
182 error ("There is already a process connected to fd %d", inch);
183 if (!NILP (buffer))
184 buffer = Fget_buffer_create (buffer);
185 proc = make_process_internal (name);
186
187 XPROCESS (proc)->pid = Fcons (infd, name);
188 XPROCESS (proc)->buffer = buffer;
189 init_process_io_handles (XPROCESS (proc), (void*)inch, (void*)XINT (outfd), 0);
190 UNIX_DATA (XPROCESS (proc))->connected_via_filedesc_p = 1;
191
192 event_stream_select_process (XPROCESS (proc));
193
194 return proc;
195 }
196
197 #ifdef HAVE_PTYS
198
199 /* Open an available pty, returning a file descriptor.
200 Return -1 on failure.
201 The file name of the terminal corresponding to the pty
202 is left in the variable pty_name. */
203
204 static int
205 allocate_pty (void)
206 {
207 struct stat stb;
208 int c, i;
209 int fd;
210
211 /* Some systems name their pseudoterminals so that there are gaps in
212 the usual sequence - for example, on HP9000/S700 systems, there
213 are no pseudoterminals with names ending in 'f'. So we wait for
214 three failures in a row before deciding that we've reached the
215 end of the ptys. */
216 int failed_count = 0;
217
218 #ifdef PTY_ITERATION
219 PTY_ITERATION
220 #else
221 for (c = FIRST_PTY_LETTER; c <= 'z'; c++)
222 for (i = 0; i < 16; i++)
223 #endif
224 {
225 #ifdef PTY_NAME_SPRINTF
226 PTY_NAME_SPRINTF
227 #else
228 sprintf (pty_name, "/dev/pty%c%x", c, i);
229 #endif /* no PTY_NAME_SPRINTF */
230
231 #ifdef PTY_OPEN
232 PTY_OPEN;
233 #else /* no PTY_OPEN */
234 #ifdef IRIS
235 /* Unusual IRIS code */
236 *ptyv = open ("/dev/ptc", O_RDWR | O_NDELAY | OPEN_BINARY, 0);
237 if (fd < 0)
238 return -1;
239 if (fstat (fd, &stb) < 0)
240 return -1;
241 #else /* not IRIS */
242 if (stat (pty_name, &stb) < 0)
243 {
244 failed_count++;
245 if (failed_count >= 3)
246 return -1;
247 }
248 else
249 failed_count = 0;
250 #ifdef O_NONBLOCK
251 fd = open (pty_name, O_RDWR | O_NONBLOCK | OPEN_BINARY, 0);
252 #else
253 fd = open (pty_name, O_RDWR | O_NDELAY | OPEN_BINARY, 0);
254 #endif
255 #endif /* not IRIS */
256 #endif /* no PTY_OPEN */
257
258 if (fd >= 0)
259 {
260 /* check to make certain that both sides are available
261 this avoids a nasty yet stupid bug in rlogins */
262 #ifdef PTY_TTY_NAME_SPRINTF
263 PTY_TTY_NAME_SPRINTF
264 #else
265 sprintf (pty_name, "/dev/tty%c%x", c, i);
266 #endif /* no PTY_TTY_NAME_SPRINTF */
267 #ifndef UNIPLUS
268 if (access (pty_name, 6) != 0)
269 {
270 close (fd);
271 #if !defined(IRIS) && !defined(__sgi)
272 continue;
273 #else
274 return -1;
275 #endif /* IRIS */
276 }
277 #endif /* not UNIPLUS */
278 setup_pty (fd);
279 return fd;
280 }
281 }
282 return -1;
283 }
284 #endif /* HAVE_PTYS */
285
286 static int
287 create_bidirectional_pipe (int *inchannel, int *outchannel,
288 volatile int *forkin, volatile int *forkout)
289 {
290 int sv[2];
291
292 #ifdef SKTPAIR
293 if (socketpair (AF_UNIX, SOCK_STREAM, 0, sv) < 0)
294 return -1;
295 *outchannel = *inchannel = sv[0];
296 *forkout = *forkin = sv[1];
297 #else /* not SKTPAIR */
298 int temp;
299 temp = pipe (sv);
300 if (temp < 0) return -1;
301 *inchannel = sv[0];
302 *forkout = sv[1];
303 temp = pipe (sv);
304 if (temp < 0) return -1;
305 *outchannel = sv[1];
306 *forkin = sv[0];
307 #endif /* not SKTPAIR */
308 return 0;
309 }
310
311
312 #ifdef HAVE_SOCKETS
313
314 static int
315 get_internet_address (Lisp_Object host, struct sockaddr_in *address,
316 Error_behavior errb)
317 {
318 struct hostent *host_info_ptr = NULL;
319 #ifdef TRY_AGAIN
320 int count = 0;
321 #endif
322
323 memset (address, 0, sizeof (*address));
324
325 while (1)
326 {
327 #ifdef TRY_AGAIN
328 if (count++ > 10) break;
329 #ifndef BROKEN_CYGWIN
330 h_errno = 0;
331 #endif
332 #endif
333 /* Some systems can't handle SIGIO/SIGALARM in gethostbyname. */
334 slow_down_interrupts ();
335 host_info_ptr = gethostbyname ((char *) XSTRING_DATA (host));
336 speed_up_interrupts ();
337 #ifdef TRY_AGAIN
338 if (! (host_info_ptr == 0 && h_errno == TRY_AGAIN))
339 #endif
340 break;
341 Fsleep_for (make_int (1));
342 }
343 if (host_info_ptr)
344 {
345 address->sin_family = host_info_ptr->h_addrtype;
346 memcpy (&address->sin_addr, host_info_ptr->h_addr, host_info_ptr->h_length);
347 }
348 else
349 {
350 IN_ADDR numeric_addr;
351 /* Attempt to interpret host as numeric inet address */
352 numeric_addr = inet_addr ((char *) XSTRING_DATA (host));
353 if (NUMERIC_ADDR_ERROR)
354 {
355 maybe_error (Qprocess, errb,
356 "Unknown host \"%s\"", XSTRING_DATA (host));
357 return 0;
358 }
359
360 /* There was some broken code here that called strlen() here
361 on (char *) &numeric_addr and even sometimes accessed
362 uninitialized data. */
363 address->sin_family = AF_INET;
364 * (IN_ADDR *) &address->sin_addr = numeric_addr;
365 }
366
367 return 1;
368 }
369
370 static void
371 set_socket_nonblocking_maybe (int fd, int port, CONST char* proto)
372 {
373 #ifdef PROCESS_IO_BLOCKING
374 Lisp_Object tail;
375
376 for (tail = network_stream_blocking_port_list; CONSP (tail); tail = XCDR (tail))
377 {
378 Lisp_Object tail_port = XCAR (tail);
379
380 if (STRINGP (tail_port))
381 {
382 struct servent *svc_info;
383 CHECK_STRING (tail_port);
384 svc_info = getservbyname ((char *) XSTRING_DATA (tail_port), proto);
385 if ((svc_info != 0) && (svc_info->s_port == port))
386 break;
387 else
388 continue;
389 }
390 else if ((INTP (tail_port)) && (htons ((unsigned short) XINT (tail_port)) == port))
391 break;
392 }
393
394 if (!CONSP (tail))
395 {
396 set_descriptor_non_blocking (fd);
397 }
398 #else
399 set_descriptor_non_blocking (fd);
400 #endif /* PROCESS_IO_BLOCKING */
401 }
402
403 #endif /* HAVE_SOCKETS */
404
405 /* Compute the Lisp form of the process status from
406 the numeric status that was returned by `wait'. */
407
408 static void
409 update_status_from_wait_code (struct Lisp_Process *p, int *w_fmh)
410 {
411 /* C compiler lossage when attempting to pass w directly */
412 int w = *w_fmh;
413
414 if (WIFSTOPPED (w))
415 {
416 p->status_symbol = Qstop;
417 p->exit_code = WSTOPSIG (w);
418 p->core_dumped = 0;
419 }
420 else if (WIFEXITED (w))
421 {
422 p->status_symbol = Qexit;
423 p->exit_code = WEXITSTATUS (w);
424 p->core_dumped = 0;
425 }
426 else if (WIFSIGNALED (w))
427 {
428 p->status_symbol = Qsignal;
429 p->exit_code = WTERMSIG (w);
430 p->core_dumped = WCOREDUMP (w);
431 }
432 else
433 {
434 p->status_symbol = Qrun;
435 p->exit_code = 0;
436 }
437 }
438
439 #ifdef SIGCHLD
440
441 #define MAX_EXITED_PROCESSES 1000
442 static volatile pid_t exited_processes[MAX_EXITED_PROCESSES];
443 static volatile int exited_processes_status[MAX_EXITED_PROCESSES];
444 static volatile int exited_processes_index;
445
446 static volatile int sigchld_happened;
447
448 /* On receipt of a signal that a child status has changed,
449 loop asking about children with changed statuses until
450 the system says there are no more. All we do is record
451 the processes and wait status.
452
453 This function could be called from within the SIGCHLD
454 handler, so it must be completely reentrant. When
455 not called from a SIGCHLD handler, BLOCK_SIGCHLD should
456 be non-zero so that SIGCHLD is blocked while this
457 function is running. (This is necessary so avoid
458 race conditions with the SIGCHLD_HAPPENED flag). */
459
460 static void
461 record_exited_processes (int block_sigchld)
462 {
463 if (!sigchld_happened)
464 {
465 return;
466 }
467
468 #ifdef EMACS_BLOCK_SIGNAL
469 if (block_sigchld)
470 EMACS_BLOCK_SIGNAL (SIGCHLD);
471 #endif
472
473 while (sigchld_happened)
474 {
475 int pid;
476 int w;
477
478 /* Keep trying to get a status until we get a definitive result. */
479 do
480 {
481 errno = 0;
482 #ifdef WNOHANG
483 # ifndef WUNTRACED
484 # define WUNTRACED 0
485 # endif /* not WUNTRACED */
486 # ifdef HAVE_WAITPID
487 pid = waitpid ((pid_t) -1, &w, WNOHANG | WUNTRACED);
488 # else
489 pid = wait3 (&w, WNOHANG | WUNTRACED, 0);
490 # endif
491 #else /* not WNOHANG */
492 pid = wait (&w);
493 #endif /* not WNOHANG */
494 }
495 while (pid <= 0 && errno == EINTR);
496
497 if (pid <= 0)
498 break;
499
500 if (exited_processes_index < MAX_EXITED_PROCESSES)
501 {
502 exited_processes[exited_processes_index] = pid;
503 exited_processes_status[exited_processes_index] = w;
504 exited_processes_index++;
505 }
506
507 /* On systems with WNOHANG, we just ignore the number
508 of times that SIGCHLD was signalled, and keep looping
509 until there are no more processes to wait on. If we
510 don't have WNOHANG, we have to rely on the count in
511 SIGCHLD_HAPPENED. */
512 #ifndef WNOHANG
513 sigchld_happened--;
514 #endif /* not WNOHANG */
515 }
516
517 sigchld_happened = 0;
518
519 if (block_sigchld)
520 EMACS_UNBLOCK_SIGNAL (SIGCHLD);
521 }
522
523 /* For any processes that have changed status and are recorded
524 and such, update the corresponding struct Lisp_Process.
525 We separate this from record_exited_processes() so that
526 we never have to call this function from within a signal
527 handler. We block SIGCHLD in case record_exited_processes()
528 is called from a signal handler. */
529
530 /** USG WARNING: Although it is not obvious from the documentation
531 in signal(2), on a USG system the SIGCLD handler MUST NOT call
532 signal() before executing at least one wait(), otherwise the handler
533 will be called again, resulting in an infinite loop. The relevant
534 portion of the documentation reads "SIGCLD signals will be queued
535 and the signal-catching function will be continually reentered until
536 the queue is empty". Invoking signal() causes the kernel to reexamine
537 the SIGCLD queue. Fred Fish, UniSoft Systems Inc.
538
539 (Note that now this only applies in SYS V Release 2 and before.
540 On SYS V Release 3, we use sigset() to set the signal handler for
541 the first time, and so we don't have to reestablish the signal handler
542 in the handler below. On SYS V Release 4, we don't get this weirdo
543 behavior when we use sigaction(), which we do use.) */
544
545 static SIGTYPE
546 sigchld_handler (int signo)
547 {
548 #ifdef OBNOXIOUS_SYSV_SIGCLD_BEHAVIOR
549 int old_errno = errno;
550
551 sigchld_happened++;
552 record_exited_processes (0);
553 errno = old_errno;
554 #else
555 sigchld_happened++;
556 #endif
557 #ifdef HAVE_UNIXOID_EVENT_LOOP
558 signal_fake_event ();
559 #endif
560 /* WARNING - must come after wait3() for USG systems */
561 EMACS_REESTABLISH_SIGNAL (signo, sigchld_handler);
562 SIGRETURN;
563 }
564
565 #endif /* SIGCHLD */
566
567 #ifdef SIGNALS_VIA_CHARACTERS
568 /* Get signal character to send to process if SIGNALS_VIA_CHARACTERS */
569
570 static int
571 process_signal_char (int tty_fd, int signo)
572 {
573 /* If it's not a tty, pray that these default values work */
574 if (!isatty(tty_fd)) {
575 #define CNTL(ch) (037 & (ch))
576 switch (signo)
577 {
578 case SIGINT: return CNTL('C');
579 case SIGQUIT: return CNTL('\\');
580 #ifdef SIGTSTP
581 case SIGTSTP: return CNTL('Z');
582 #endif
583 }
584 }
585
586 #ifdef HAVE_TERMIOS
587 /* TERMIOS is the latest and bestest, and seems most likely to work.
588 If the system has it, use it. */
589 {
590 struct termios t;
591 tcgetattr (tty_fd, &t);
592 switch (signo)
593 {
594 case SIGINT: return t.c_cc[VINTR];
595 case SIGQUIT: return t.c_cc[VQUIT];
596 # if defined (VSWTCH) && !defined (PREFER_VSUSP)
597 case SIGTSTP: return t.c_cc[VSWTCH];
598 # else
599 case SIGTSTP: return t.c_cc[VSUSP];
600 # endif
601 }
602 }
603
604 # elif defined (TIOCGLTC) && defined (TIOCGETC) /* not HAVE_TERMIOS */
605 {
606 /* On Berkeley descendants, the following IOCTL's retrieve the
607 current control characters. */
608 struct tchars c;
609 struct ltchars lc;
610 switch (signo)
611 {
612 case SIGINT: ioctl (tty_fd, TIOCGETC, &c); return c.t_intrc;
613 case SIGQUIT: ioctl (tty_fd, TIOCGETC, &c); return c.t_quitc;
614 # ifdef SIGTSTP
615 case SIGTSTP: ioctl (tty_fd, TIOCGLTC, &lc); return lc.t_suspc;
616 # endif /* SIGTSTP */
617 }
618 }
619
620 # elif defined (TCGETA) /* ! defined (TIOCGLTC) && defined (TIOCGETC) */
621 {
622 /* On SYSV descendants, the TCGETA ioctl retrieves the current
623 control characters. */
624 struct termio t;
625 ioctl (tty_fd, TCGETA, &t);
626 switch (signo) {
627 case SIGINT: return t.c_cc[VINTR];
628 case SIGQUIT: return t.c_cc[VQUIT];
629 # ifdef SIGTSTP
630 case SIGTSTP: return t.c_cc[VSWTCH];
631 # endif /* SIGTSTP */
632 }
633 }
634 # else /* ! defined (TCGETA) */
635 #error ERROR! Using SIGNALS_VIA_CHARACTERS, but not HAVE_TERMIOS || (TIOCGLTC && TIOCGETC) || TCGETA
636 /* If your system configuration files define SIGNALS_VIA_CHARACTERS,
637 you'd better be using one of the alternatives above! */
638 # endif /* ! defined (TCGETA) */
639 return '\0';
640 }
641 #endif /* SIGNALS_VIA_CHARACTERS */
642
643
644
645
646 /**********************************************************************/
647 /* Process implementation methods */
648 /**********************************************************************/
649
650 /*
651 * Allocate and initialize Lisp_Process->process_data
652 */
653
654 static void
655 unix_alloc_process_data (struct Lisp_Process *p)
656 {
657 p->process_data = xnew (struct unix_process_data);
658
659 UNIX_DATA(p)->connected_via_filedesc_p = 0;
660 UNIX_DATA(p)->infd = -1;
661 UNIX_DATA(p)->subtty = -1;
662 UNIX_DATA(p)->tty_name = Qnil;
663 UNIX_DATA(p)->pty_flag = 0;
664 }
665
666 /*
667 * Mark any Lisp objects in Lisp_Process->process_data
668 */
669
670 static void
671 unix_mark_process_data (struct Lisp_Process *proc,
672 void (*markobj) (Lisp_Object))
673 {
674 ((markobj) (UNIX_DATA(proc)->tty_name));
675 }
676
677 /*
678 * Initialize XEmacs process implemenation once
679 */
680
681 #ifdef SIGCHLD
682 static void
683 unix_init_process (void)
684 {
685 #ifndef CANNOT_DUMP
686 if (! noninteractive || initialized)
687 #endif
688 signal (SIGCHLD, sigchld_handler);
689 }
690 #endif /* SIGCHLD */
691
692 /*
693 * Initialize any process local data. This is called when newly
694 * created process is connected to real OS file handles. The
695 * handles are generally represented by void* type, but are
696 * of type int (file descriptors) for UNIX
697 */
698
699 static void
700 unix_init_process_io_handles (struct Lisp_Process *p, void* in, void* out, int flags)
701 {
702 UNIX_DATA(p)->infd = (int)in;
703 }
704
705 /*
706 * Fork off a subprocess. P is a pointer to newly created subprocess
707 * object. If this function signals, the caller is responsible for
708 * deleting (and finalizing) the process object.
709 *
710 * The method must return PID of the new proces, a (positive??? ####) number
711 * which fits into Lisp_Int. No return value indicates an error, the method
712 * must signal an error instead.
713 */
714
715 static int
716 unix_create_process (struct Lisp_Process *p,
717 char **new_argv, CONST char *current_dir)
718 {
719 /* This function rewritten by wing@666.com. */
720
721 int pid, inchannel, outchannel;
722 /* Use volatile to protect variables from being clobbered by longjmp. */
723 volatile int forkin, forkout;
724 volatile int pty_flag = 0;
725 char **env;
726
727 env = environ;
728
729 inchannel = outchannel = forkin = forkout = -1;
730
731 #ifdef HAVE_PTYS
732 if (!NILP (Vprocess_connection_type))
733 {
734 /* find a new pty, open the master side, return the opened
735 file handle, and store the name of the corresponding slave
736 side in global variable pty_name. */
737 outchannel = inchannel = allocate_pty ();
738 }
739
740 if (inchannel >= 0)
741 {
742 /* You're "supposed" to now open the slave in the child.
743 On some systems, we can open it here; this allows for
744 better error checking. */
745 #if !defined(USG)
746 /* On USG systems it does not work to open the pty's tty here
747 and then close and reopen it in the child. */
748 #ifdef O_NOCTTY
749 /* Don't let this terminal become our controlling terminal
750 (in case we don't have one). */
751 forkout = forkin = open (pty_name, O_RDWR | O_NOCTTY | OPEN_BINARY, 0);
752 #else
753 forkout = forkin = open (pty_name, O_RDWR | OPEN_BINARY, 0);
754 #endif
755 if (forkin < 0)
756 goto io_failure;
757 #endif /* not USG */
758 UNIX_DATA(p)->pty_flag = pty_flag = 1;
759 }
760 else
761 #endif /* HAVE_PTYS */
762 if (create_bidirectional_pipe (&inchannel, &outchannel,
763 &forkin, &forkout) < 0)
764 goto io_failure;
765
766 #if 0
767 /* Replaced by close_process_descs */
768 set_exclusive_use (inchannel);
769 set_exclusive_use (outchannel);
770 #endif
771
772 set_descriptor_non_blocking (inchannel);
773
774 /* Record this as an active process, with its channels.
775 As a result, child_setup will close Emacs's side of the pipes. */
776 init_process_io_handles (p, (void*)inchannel, (void*)outchannel, pty_flag);
777 /* Record the tty descriptor used in the subprocess. */
778 UNIX_DATA(p)->subtty = forkin;
779
780 {
781 #if !defined(__CYGWIN32__)
782 /* child_setup must clobber environ on systems with true vfork.
783 Protect it from permanent change. */
784 char **save_environ = environ;
785 #endif
786
787 #ifdef EMACS_BTL
788 /* when performance monitoring is on, turn it off before the vfork(),
789 as the child has no handler for the signal -- when back in the
790 parent process, turn it back on if it was really on when you "turned
791 it off" */
792 int logging_on = cadillac_stop_logging (); /* #### rename me */
793 #endif
794
795 pid = fork ();
796 if (pid == 0)
797 {
798 /**** Now we're in the child process ****/
799 int xforkin = forkin;
800 int xforkout = forkout;
801
802 if (!pty_flag)
803 EMACS_SEPARATE_PROCESS_GROUP ();
804 #ifdef HAVE_PTYS
805 else
806 {
807 /* Disconnect the current controlling terminal, pursuant to
808 making the pty be the controlling terminal of the process.
809 Also put us in our own process group. */
810
811 disconnect_controlling_terminal ();
812
813 /* Open the pty connection and make the pty's terminal
814 our controlling terminal.
815
816 On systems with TIOCSCTTY, we just use it to set
817 the controlling terminal. On other systems, the
818 first TTY we open becomes the controlling terminal.
819 So, we end up with four possibilities:
820
821 (1) on USG and TIOCSCTTY systems, we open the pty
822 and use TIOCSCTTY.
823 (2) on other USG systems, we just open the pty.
824 (3) on non-USG systems with TIOCSCTTY, we
825 just use TIOCSCTTY. (On non-USG systems, we
826 already opened the pty in the parent process.)
827 (4) on non-USG systems without TIOCSCTTY, we
828 close the pty and reopen it.
829
830 This would be cleaner if we didn't open the pty
831 in the parent process, but doing it that way
832 makes it possible to trap error conditions.
833 It's harder to convey an error from the child
834 process, and I don't feel like messing with
835 this now. */
836
837 /* There was some weirdo, probably wrong,
838 conditionalization on RTU and UNIPLUS here.
839 I deleted it. So sue me. */
840
841 /* SunOS has TIOCSCTTY but the close/open method
842 also works. */
843
844 # if defined (USG) || !defined (TIOCSCTTY)
845 /* Now close the pty (if we had it open) and reopen it.
846 This makes the pty the controlling terminal of the
847 subprocess. */
848 /* I wonder if close (open (pty_name, ...)) would work? */
849 if (xforkin >= 0)
850 close (xforkin);
851 xforkout = xforkin = open (pty_name, O_RDWR | OPEN_BINARY, 0);
852 if (xforkin < 0)
853 {
854 write (1, "Couldn't open the pty terminal ", 31);
855 write (1, pty_name, strlen (pty_name));
856 write (1, "\n", 1);
857 _exit (1);
858 }
859 # endif /* USG or not TIOCSCTTY */
860
861 /* Miscellaneous setup required for some systems.
862 Must be done before using tc* functions on xforkin.
863 This guarantees that isatty(xforkin) is true. */
864
865 # ifdef SETUP_SLAVE_PTY
866 SETUP_SLAVE_PTY;
867 # endif /* SETUP_SLAVE_PTY */
868
869 # ifdef TIOCSCTTY
870 /* We ignore the return value
871 because faith@cs.unc.edu says that is necessary on Linux. */
872 assert (isatty (xforkin));
873 ioctl (xforkin, TIOCSCTTY, 0);
874 # endif /* TIOCSCTTY */
875
876 /* Change the line discipline. */
877
878 # if defined (HAVE_TERMIOS) && defined (LDISC1)
879 {
880 struct termios t;
881 assert (isatty (xforkin));
882 tcgetattr (xforkin, &t);
883 t.c_lflag = LDISC1;
884 if (tcsetattr (xforkin, TCSANOW, &t) < 0)
885 perror ("create_process/tcsetattr LDISC1 failed\n");
886 }
887 # elif defined (NTTYDISC) && defined (TIOCSETD)
888 {
889 /* Use new line discipline. TIOCSETD is accepted and
890 ignored on Sys5.4 systems with ttcompat. */
891 int ldisc = NTTYDISC;
892 assert (isatty (xforkin));
893 ioctl (xforkin, TIOCSETD, &ldisc);
894 }
895 # endif /* TIOCSETD & NTTYDISC */
896
897 /* Make our process group be the foreground group
898 of our new controlling terminal. */
899
900 {
901 int piddly = EMACS_GET_PROCESS_GROUP ();
902 EMACS_SET_TTY_PROCESS_GROUP (xforkin, &piddly);
903 }
904
905 # ifdef AIX
906 /* On AIX, we've disabled SIGHUP above once we start a
907 child on a pty. Now reenable it in the child, so it
908 will die when we want it to. */
909 signal (SIGHUP, SIG_DFL);
910 # endif /* AIX */
911 }
912 #endif /* HAVE_PTYS */
913
914 signal (SIGINT, SIG_DFL);
915 signal (SIGQUIT, SIG_DFL);
916
917 if (pty_flag)
918 {
919 /* Set up the terminal characteristics of the pty. */
920 child_setup_tty (xforkout);
921 }
922
923 child_setup (xforkin, xforkout, xforkout, new_argv, current_dir);
924 }
925 #ifdef EMACS_BTL
926 else if (logging_on)
927 cadillac_start_logging (); /* #### rename me */
928 #endif
929
930 #if !defined(__CYGWIN32__)
931 environ = save_environ;
932 #endif
933 }
934
935 if (pid < 0)
936 {
937 close_descriptor_pair (forkin, forkout);
938 report_file_error ("Doing fork", Qnil);
939 }
940
941 /* #### dmoore - why is this commented out, otherwise we leave
942 subtty = forkin, but then we close forkin just below. */
943 /* UNIX_DATA(p)->subtty = -1; */
944
945 /* If the subfork execv fails, and it exits,
946 this close hangs. I don't know why.
947 So have an interrupt jar it loose. */
948 if (forkin >= 0)
949 close_safely (forkin);
950 if (forkin != forkout && forkout >= 0)
951 close (forkout);
952
953 #ifdef HAVE_PTYS
954 if (pty_flag)
955 UNIX_DATA (p)->tty_name = build_string (pty_name);
956 else
957 #endif
958 UNIX_DATA (p)->tty_name = Qnil;
959
960 /* Notice that SIGCHLD was not blocked. (This is not possible on
961 some systems.) No biggie if SIGCHLD occurs right around the
962 time that this call happens, because SIGCHLD() does not actually
963 deselect the process (that doesn't occur until the next time
964 we're waiting for an event, when status_notify() is called). */
965 return pid;
966
967 io_failure:
968 {
969 int temp = errno;
970 close_descriptor_pair (forkin, forkout);
971 close_descriptor_pair (inchannel, outchannel);
972 errno = temp;
973 report_file_error ("Opening pty or pipe", Qnil);
974 }
975
976 RETURN_NOT_REACHED (0);
977 }
978
979 /*
980 * Return nonzero if this process is a ToolTalk connection.
981 */
982
983 static int
984 unix_tooltalk_connection_p (struct Lisp_Process *p)
985 {
986 return UNIX_DATA(p)->connected_via_filedesc_p;
987 }
988
989 /*
990 * This is called to set process' virtual terminal size
991 */
992
993 static int
994 unix_set_window_size (struct Lisp_Process* p, int cols, int rows)
995 {
996 return set_window_size (UNIX_DATA(p)->infd, cols, rows);
997 }
998
999 /*
1000 * This method is called to update status fields of the process
1001 * structure. If the process has not existed, this method is
1002 * expected to do nothing.
1003 *
1004 * The method is called only for real child processes.
1005 */
1006
1007 #ifdef HAVE_WAITPID
1008 static void
1009 unix_update_status_if_terminated (struct Lisp_Process* p)
1010 {
1011 int w;
1012 #ifdef SIGCHLD
1013 EMACS_BLOCK_SIGNAL (SIGCHLD);
1014 #endif
1015 if (waitpid (XINT (p->pid), &w, WNOHANG) == XINT (p->pid))
1016 {
1017 p->tick++;
1018 update_status_from_wait_code (p, &w);
1019 }
1020 #ifdef SIGCHLD
1021 EMACS_UNBLOCK_SIGNAL (SIGCHLD);
1022 #endif
1023 }
1024 #endif
1025
1026 /*
1027 * Update status of all exited processes. Called when SIGCLD has signaled.
1028 */
1029
1030 #ifdef SIGCHLD
1031 static void
1032 unix_reap_exited_processes (void)
1033 {
1034 int i;
1035 struct Lisp_Process *p;
1036
1037 #ifndef OBNOXIOUS_SYSV_SIGCLD_BEHAVIOR
1038 record_exited_processes (1);
1039 #endif
1040
1041 if (exited_processes_index <= 0)
1042 {
1043 return;
1044 }
1045
1046 #ifdef EMACS_BLOCK_SIGNAL
1047 EMACS_BLOCK_SIGNAL (SIGCHLD);
1048 #endif
1049 for (i = 0; i < exited_processes_index; i++)
1050 {
1051 int pid = exited_processes[i];
1052 int w = exited_processes_status[i];
1053
1054 /* Find the process that signaled us, and record its status. */
1055
1056 p = 0;
1057 {
1058 Lisp_Object tail;
1059 LIST_LOOP (tail, Vprocess_list)
1060 {
1061 Lisp_Object proc = XCAR (tail);
1062 p = XPROCESS (proc);
1063 if (INTP (p->pid) && XINT (p->pid) == pid)
1064 break;
1065 p = 0;
1066 }
1067 }
1068
1069 if (p)
1070 {
1071 /* Change the status of the process that was found. */
1072 p->tick++;
1073 process_tick++;
1074 update_status_from_wait_code (p, &w);
1075
1076 /* If process has terminated, stop waiting for its output. */
1077 if (WIFSIGNALED (w) || WIFEXITED (w))
1078 {
1079 if (!NILP(p->pipe_instream))
1080 {
1081 /* We can't just call event_stream->unselect_process_cb (p)
1082 here, because that calls XtRemoveInput, which is not
1083 necessarily reentrant, so we can't call this at interrupt
1084 level.
1085 */
1086 }
1087 }
1088 }
1089 else
1090 {
1091 /* There was no asynchronous process found for that id. Check
1092 if we have a synchronous process. Only set sync process status
1093 if there is one, so we work OK with the waitpid() call in
1094 wait_for_termination(). */
1095 if (synch_process_alive != 0)
1096 { /* Set the global sync process status variables. */
1097 synch_process_alive = 0;
1098
1099 /* Report the status of the synchronous process. */
1100 if (WIFEXITED (w))
1101 synch_process_retcode = WEXITSTATUS (w);
1102 else if (WIFSIGNALED (w))
1103 synch_process_death = signal_name (WTERMSIG (w));
1104 }
1105 }
1106 }
1107
1108 exited_processes_index = 0;
1109
1110 EMACS_UNBLOCK_SIGNAL (SIGCHLD);
1111 }
1112 #endif /* SIGCHLD */
1113
1114 /*
1115 * Stuff the entire contents of LSTREAM to the process ouptut pipe
1116 */
1117
1118 static JMP_BUF send_process_frame;
1119
1120 static SIGTYPE
1121 send_process_trap (int signum)
1122 {
1123 EMACS_REESTABLISH_SIGNAL (signum, send_process_trap);
1124 EMACS_UNBLOCK_SIGNAL (signum);
1125 LONGJMP (send_process_frame, 1);
1126 }
1127
1128 static void
1129 unix_send_process (volatile Lisp_Object proc, struct lstream* lstream)
1130 {
1131 /* Use volatile to protect variables from being clobbered by longjmp. */
1132 SIGTYPE (*volatile old_sigpipe) (int) = 0;
1133 volatile struct Lisp_Process *p = XPROCESS (proc);
1134
1135 if (!SETJMP (send_process_frame))
1136 {
1137 /* use a reasonable-sized buffer (somewhere around the size of the
1138 stream buffer) so as to avoid inundating the stream with blocked
1139 data. */
1140 Bufbyte chunkbuf[512];
1141 Bytecount chunklen;
1142
1143 while (1)
1144 {
1145 int writeret;
1146
1147 chunklen = Lstream_read (lstream, chunkbuf, 512);
1148 if (chunklen <= 0)
1149 break; /* perhaps should abort() if < 0?
1150 This should never happen. */
1151 old_sigpipe =
1152 (SIGTYPE (*) (int)) signal (SIGPIPE, send_process_trap);
1153 /* Lstream_write() will never successfully write less than
1154 the amount sent in. In the worst case, it just buffers
1155 the unwritten data. */
1156 writeret = Lstream_write (XLSTREAM (DATA_OUTSTREAM(p)), chunkbuf,
1157 chunklen);
1158 signal (SIGPIPE, old_sigpipe);
1159 if (writeret < 0)
1160 /* This is a real error. Blocking errors are handled
1161 specially inside of the filedesc stream. */
1162 report_file_error ("writing to process",
1163 list1 (proc));
1164 while (Lstream_was_blocked_p (XLSTREAM (p->pipe_outstream)))
1165 {
1166 /* Buffer is full. Wait, accepting input;
1167 that may allow the program
1168 to finish doing output and read more. */
1169 Faccept_process_output (Qnil, make_int (1), Qnil);
1170 old_sigpipe =
1171 (SIGTYPE (*) (int)) signal (SIGPIPE, send_process_trap);
1172 Lstream_flush (XLSTREAM (p->pipe_outstream));
1173 signal (SIGPIPE, old_sigpipe);
1174 }
1175 }
1176 }
1177 else
1178 { /* We got here from a longjmp() from the SIGPIPE handler */
1179 signal (SIGPIPE, old_sigpipe);
1180 p->status_symbol = Qexit;
1181 p->exit_code = 256; /* #### SIGPIPE ??? */
1182 p->core_dumped = 0;
1183 p->tick++;
1184 process_tick++;
1185 deactivate_process (proc);
1186 error ("SIGPIPE raised on process %s; closed it",
1187 XSTRING_DATA (p->name));
1188 }
1189
1190 old_sigpipe = (SIGTYPE (*) (int)) signal (SIGPIPE, send_process_trap);
1191 Lstream_flush (XLSTREAM (DATA_OUTSTREAM(p)));
1192 signal (SIGPIPE, old_sigpipe);
1193 }
1194
1195 /*
1196 * Send EOF to the process. The default implementation simply
1197 * closes the output stream. The method must return 0 to call
1198 * the default implementation, or 1 if it has taken all care about
1199 * sending EOF to the process.
1200 */
1201
1202 static int
1203 unix_process_send_eof (Lisp_Object proc)
1204 {
1205 if (!UNIX_DATA (XPROCESS (proc))->pty_flag)
1206 return 0;
1207
1208 /* #### get_eof_char simply doesn't return the correct character
1209 here. Maybe it is needed to determine the right eof
1210 character in init_process_io_handles but here it simply screws
1211 things up. */
1212 #if 0
1213 Bufbyte eof_char = get_eof_char (XPROCESS (proc));
1214 send_process (proc, Qnil, &eof_char, 0, 1);
1215 #else
1216 send_process (proc, Qnil, (CONST Bufbyte *) "\004", 0, 1);
1217 #endif
1218 return 1;
1219 }
1220
1221 /*
1222 * Called before the process is deactivated. The process object
1223 * is not immediately finalized, just undergoes a transition to
1224 * inactive state.
1225 *
1226 * The return value is a unique stream ID, as returned by
1227 * event_stream_delete_stream_pair
1228 *
1229 * In the lack of this method, only event_stream_delete_stream_pair
1230 * is called on both I/O streams of the process.
1231 *
1232 * The UNIX version quards this by ignoring possible SIGPIPE.
1233 */
1234
1235 static USID
1236 unix_deactivate_process (struct Lisp_Process *p)
1237 {
1238 SIGTYPE (*old_sigpipe) (int) = 0;
1239 USID usid;
1240
1241 if (UNIX_DATA(p)->infd >= 0)
1242 flush_pending_output (UNIX_DATA(p)->infd);
1243
1244 /* closing the outstream could result in SIGPIPE, so ignore it. */
1245 old_sigpipe = (SIGTYPE (*) (int)) signal (SIGPIPE, SIG_IGN);
1246 usid = event_stream_delete_stream_pair (p->pipe_instream, p->pipe_outstream);
1247 signal (SIGPIPE, old_sigpipe);
1248
1249 UNIX_DATA(p)->infd = -1;
1250
1251 return usid;
1252 }
1253
1254 /* send a signal number SIGNO to PROCESS.
1255 CURRENT_GROUP means send to the process group that currently owns
1256 the terminal being used to communicate with PROCESS.
1257 This is used for various commands in shell mode.
1258 If NOMSG is zero, insert signal-announcements into process's buffers
1259 right away.
1260
1261 If we can, we try to signal PROCESS by sending control characters
1262 down the pty. This allows us to signal inferiors who have changed
1263 their uid, for which killpg would return an EPERM error.
1264
1265 The method signals an error if the given SIGNO is not valid
1266 */
1267
1268 static void
1269 unix_kill_child_process (Lisp_Object proc, int signo,
1270 int current_group, int nomsg)
1271 {
1272 int gid;
1273 int no_pgrp = 0;
1274 int kill_retval;
1275 struct Lisp_Process *p = XPROCESS (proc);
1276
1277 if (!UNIX_DATA(p)->pty_flag)
1278 current_group = 0;
1279
1280 /* If we are using pgrps, get a pgrp number and make it negative. */
1281 if (current_group)
1282 {
1283 #ifdef SIGNALS_VIA_CHARACTERS
1284 /* If possible, send signals to the entire pgrp
1285 by sending an input character to it. */
1286 {
1287 char sigchar = process_signal_char(UNIX_DATA(p)->subtty, signo);
1288 if (sigchar) {
1289 send_process (proc, Qnil, (Bufbyte *) &sigchar, 0, 1);
1290 return;
1291 }
1292 }
1293 #endif /* ! defined (SIGNALS_VIA_CHARACTERS) */
1294
1295 #ifdef TIOCGPGRP
1296 /* Get the pgrp using the tty itself, if we have that.
1297 Otherwise, use the pty to get the pgrp.
1298 On pfa systems, saka@pfu.fujitsu.co.JP writes:
1299 "TIOCGPGRP symbol defined in sys/ioctl.h at E50.
1300 But, TIOCGPGRP does not work on E50 ;-P works fine on E60"
1301 His patch indicates that if TIOCGPGRP returns an error, then
1302 we should just assume that p->pid is also the process group id. */
1303 {
1304 int err;
1305
1306 err = ioctl ( (UNIX_DATA(p)->subtty != -1
1307 ? UNIX_DATA(p)->subtty
1308 : UNIX_DATA(p)->infd), TIOCGPGRP, &gid);
1309
1310 #ifdef pfa
1311 if (err == -1)
1312 gid = - XINT (p->pid);
1313 #endif /* ! defined (pfa) */
1314 }
1315 if (gid == -1)
1316 no_pgrp = 1;
1317 else
1318 gid = - gid;
1319 #else /* ! defined (TIOCGPGRP ) */
1320 /* Can't select pgrps on this system, so we know that
1321 the child itself heads the pgrp. */
1322 gid = - XINT (p->pid);
1323 #endif /* ! defined (TIOCGPGRP ) */
1324 }
1325 else
1326 gid = - XINT (p->pid);
1327
1328 switch (signo)
1329 {
1330 #ifdef SIGCONT
1331 case SIGCONT:
1332 p->status_symbol = Qrun;
1333 p->exit_code = 0;
1334 p->tick++;
1335 process_tick++;
1336 if (!nomsg)
1337 status_notify ();
1338 break;
1339 #endif /* ! defined (SIGCONT) */
1340 case SIGINT:
1341 case SIGQUIT:
1342 case SIGKILL:
1343 flush_pending_output (UNIX_DATA(p)->infd);
1344 break;
1345 }
1346
1347 /* If we don't have process groups, send the signal to the immediate
1348 subprocess. That isn't really right, but it's better than any
1349 obvious alternative. */
1350 if (no_pgrp)
1351 {
1352 kill_retval = kill (XINT (p->pid), signo) ? errno : 0;
1353 }
1354 else
1355 {
1356 /* gid may be a pid, or minus a pgrp's number */
1357 #ifdef TIOCSIGSEND
1358 if (current_group)
1359 kill_retval = ioctl (UNIX_DATA(p)->infd, TIOCSIGSEND, signo);
1360 else
1361 kill_retval = kill (- XINT (p->pid), signo) ? errno : 0;
1362 #else /* ! defined (TIOCSIGSEND) */
1363 kill_retval = EMACS_KILLPG (-gid, signo) ? errno : 0;
1364 #endif /* ! defined (TIOCSIGSEND) */
1365 }
1366
1367 if (kill_retval < 0 && errno == EINVAL)
1368 error ("Signal number %d is invalid for this system", make_int (signo));
1369 }
1370
1371 /*
1372 * Kill any process in the system given its PID.
1373 *
1374 * Returns zero if a signal successfully sent, or
1375 * negative number upon failure
1376 */
1377
1378 static int
1379 unix_kill_process_by_pid (int pid, int sigcode)
1380 {
1381 return kill (pid, sigcode);
1382 }
1383
1384 /*
1385 * Return TTY name used to communicate with subprocess
1386 */
1387
1388 static Lisp_Object
1389 unix_get_tty_name (struct Lisp_Process *p)
1390 {
1391 return UNIX_DATA (p)->tty_name;
1392 }
1393
1394 /*
1395 * Canonicalize host name HOST, and return its canonical form
1396 *
1397 * The default implemenation just takes HOST for a canonical name.
1398 */
1399
1400 #ifdef HAVE_SOCKETS
1401 static Lisp_Object
1402 unix_canonicalize_host_name (Lisp_Object host)
1403 {
1404 struct sockaddr_in address;
1405
1406 if (!get_internet_address (host, &address, ERROR_ME_NOT))
1407 return host;
1408
1409 if (address.sin_family == AF_INET)
1410 return build_string (inet_ntoa (address.sin_addr));
1411 else
1412 /* #### any clue what to do here? */
1413 return host;
1414 }
1415
1416 /* open a TCP network connection to a given HOST/SERVICE. Treated
1417 exactly like a normal process when reading and writing. Only
1418 differences are in status display and process deletion. A network
1419 connection has no PID; you cannot signal it. All you can do is
1420 deactivate and close it via delete-process */
1421
1422 static void
1423 unix_open_network_stream (Lisp_Object name, Lisp_Object host, Lisp_Object service,
1424 Lisp_Object family, void** vinfd, void** voutfd)
1425 {
1426 struct sockaddr_in address;
1427 int s, inch, outch;
1428 volatile int port;
1429 volatile int retry = 0;
1430 int retval;
1431
1432 CHECK_STRING (host);
1433
1434 if (!EQ (family, Qtcpip))
1435 error ("Unsupported protocol family \"%s\"",
1436 string_data (symbol_name (XSYMBOL (family))));
1437
1438 if (INTP (service))
1439 port = htons ((unsigned short) XINT (service));
1440 else
1441 {
1442 struct servent *svc_info;
1443 CHECK_STRING (service);
1444 svc_info = getservbyname ((char *) XSTRING_DATA (service), "tcp");
1445 if (svc_info == 0)
1446 error ("Unknown service \"%s\"", XSTRING_DATA (service));
1447 port = svc_info->s_port;
1448 }
1449
1450 get_internet_address (host, &address, ERROR_ME);
1451 address.sin_port = port;
1452
1453 s = socket (address.sin_family, SOCK_STREAM, 0);
1454 if (s < 0)
1455 report_file_error ("error creating socket", list1 (name));
1456
1457 /* Turn off interrupts here -- see comments below. There used to
1458 be code which called bind_polling_period() to slow the polling
1459 period down rather than turn it off, but that seems rather
1460 bogus to me. Best thing here is to use a non-blocking connect
1461 or something, to check for QUIT. */
1462
1463 /* Comments that are not quite valid: */
1464
1465 /* Kernel bugs (on Ultrix at least) cause lossage (not just EINTR)
1466 when connect is interrupted. So let's not let it get interrupted.
1467 Note we do not turn off polling, because polling is only used
1468 when not interrupt_input, and thus not normally used on the systems
1469 which have this bug. On systems which use polling, there's no way
1470 to quit if polling is turned off. */
1471
1472 /* Slow down polling. Some kernels have a bug which causes retrying
1473 connect to fail after a connect. */
1474
1475 slow_down_interrupts ();
1476
1477 loop:
1478
1479 /* A system call interrupted with a SIGALRM or SIGIO comes back
1480 here, with can_break_system_calls reset to 0. */
1481 SETJMP (break_system_call_jump);
1482 if (QUITP)
1483 {
1484 speed_up_interrupts ();
1485 REALLY_QUIT;
1486 /* In case something really weird happens ... */
1487 slow_down_interrupts ();
1488 }
1489
1490 /* Break out of connect with a signal (it isn't otherwise possible).
1491 Thus you don't get screwed with a hung network. */
1492 can_break_system_calls = 1;
1493 retval = connect (s, (struct sockaddr *) &address, sizeof (address));
1494 can_break_system_calls = 0;
1495 if (retval == -1 && errno != EISCONN)
1496 {
1497 int xerrno = errno;
1498 if (errno == EINTR)
1499 goto loop;
1500 if (errno == EADDRINUSE && retry < 20)
1501 {
1502 /* A delay here is needed on some FreeBSD systems,
1503 and it is harmless, since this retrying takes time anyway
1504 and should be infrequent.
1505 `sleep-for' allowed for quitting this loop with interrupts
1506 slowed down so it can't be used here. Async timers should
1507 already be disabled at this point so we can use `sleep'. */
1508 sleep (1);
1509 retry++;
1510 goto loop;
1511 }
1512
1513 close (s);
1514
1515 speed_up_interrupts ();
1516
1517 errno = xerrno;
1518 report_file_error ("connection failed", list2 (host, name));
1519 }
1520
1521 speed_up_interrupts ();
1522
1523 inch = s;
1524 outch = dup (s);
1525 if (outch < 0)
1526 {
1527 close (s); /* this used to be leaked; from Kyle Jones */
1528 report_file_error ("error duplicating socket", list1 (name));
1529 }
1530
1531 set_socket_nonblocking_maybe (inch, port, "tcp");
1532
1533 *vinfd = (void*)inch;
1534 *voutfd = (void*)outch;
1535 }
1536
1537
1538 #ifdef HAVE_MULTICAST
1539
1540 /* Didier Verna <verna@inf.enst.fr> Nov. 28 1997.
1541
1542 This function is similar to open-network-stream-internal, but provides a
1543 mean to open an UDP multicast connection instead of a TCP one. Like in the
1544 TCP case, the multicast connection will be seen as a sub-process,
1545
1546 Some notes:
1547 - Normaly, we should use sendto and recvfrom with non connected
1548 sockets. The current code doesn't allow us to do this. In the future, it
1549 would be a good idea to extend the process data structure in order to deal
1550 properly with the different types network connections.
1551 - For the same reason, when leaving a multicast group, it is better to make
1552 a setsockopt - IP_DROP_MEMBERSHIP before closing the descriptors.
1553 Unfortunately, this can't be done here because delete_process doesn't know
1554 about the kind of connection we have. However, this is not such an
1555 important issue.
1556 */
1557
1558 static void
1559 unix_open_multicast_group (Lisp_Object name, Lisp_Object dest, Lisp_Object port,
1560 Lisp_Object ttl, void** vinfd, void** voutfd)
1561 {
1562 struct ip_mreq imr;
1563 struct sockaddr_in sa;
1564 struct protoent *udp;
1565 int ws, rs;
1566 int theport;
1567 unsigned char thettl;
1568 int one = 1; /* For REUSEADDR */
1569 int ret;
1570 volatile int retry = 0;
1571
1572 CHECK_STRING (dest);
1573
1574 CHECK_NATNUM (port);
1575 theport = htons ((unsigned short) XINT (port));
1576
1577 CHECK_NATNUM (ttl);
1578 thettl = (unsigned char) XINT (ttl);
1579
1580 if ((udp = getprotobyname ("udp")) == NULL)
1581 error ("No info available for UDP protocol");
1582
1583 /* Init the sockets. Yes, I need 2 sockets. I couldn't duplicate one. */
1584 if ((rs = socket (PF_INET, SOCK_DGRAM, udp->p_proto)) < 0)
1585 report_file_error ("error creating socket", list1(name));
1586 if ((ws = socket (PF_INET, SOCK_DGRAM, udp->p_proto)) < 0)
1587 {
1588 close (rs);
1589 report_file_error ("error creating socket", list1(name));
1590 }
1591
1592 /* This will be used for both sockets */
1593 bzero(&sa, sizeof(sa));
1594 sa.sin_family = AF_INET;
1595 sa.sin_port = theport;
1596 sa.sin_addr.s_addr = htonl (inet_addr ((char *) XSTRING_DATA (dest)));
1597
1598 /* Socket configuration for reading ------------------------ */
1599
1600 /* Multiple connections from the same machine. This must be done before
1601 bind. If it fails, it shouldn't be fatal. The only consequence is that
1602 people won't be able to connect twice from the same machine. */
1603 if (setsockopt (rs, SOL_SOCKET, SO_REUSEADDR, (char *) &one, sizeof (one))
1604 < 0)
1605 warn_when_safe (Qmulticast, Qwarning, "Cannot reuse socket address");
1606
1607 /* bind socket name */
1608 if (bind (rs, (struct sockaddr *)&sa, sizeof(sa)))
1609 {
1610 close (rs);
1611 close (ws);
1612 report_file_error ("error binding socket", list2(name, port));
1613 }
1614
1615 /* join multicast group */
1616 imr.imr_multiaddr.s_addr = htonl (inet_addr ((char *) XSTRING_DATA (dest)));
1617 imr.imr_interface.s_addr = htonl (INADDR_ANY);
1618 if (setsockopt (rs, IPPROTO_IP, IP_ADD_MEMBERSHIP,
1619 (char *) &imr, sizeof (struct ip_mreq)) < 0)
1620 {
1621 close (ws);
1622 close (rs);
1623 report_file_error ("error adding membership", list2(name, dest));
1624 }
1625
1626 /* Socket configuration for writing ----------------------- */
1627
1628 /* Normaly, there's no 'connect' in multicast, since we use preferentialy
1629 'sendto' and 'recvfrom'. However, in order to handle this connection in
1630 the process-like way it is done for TCP, we must be able to use 'write'
1631 instead of 'sendto'. Consequently, we 'connect' this socket. */
1632
1633 /* See open-network-stream-internal for comments on this part of the code */
1634 slow_down_interrupts ();
1635
1636 loop:
1637
1638 /* A system call interrupted with a SIGALRM or SIGIO comes back
1639 here, with can_break_system_calls reset to 0. */
1640 SETJMP (break_system_call_jump);
1641 if (QUITP)
1642 {
1643 speed_up_interrupts ();
1644 REALLY_QUIT;
1645 /* In case something really weird happens ... */
1646 slow_down_interrupts ();
1647 }
1648
1649 /* Break out of connect with a signal (it isn't otherwise possible).
1650 Thus you don't get screwed with a hung network. */
1651 can_break_system_calls = 1;
1652 ret = connect (ws, (struct sockaddr *) &sa, sizeof (sa));
1653 can_break_system_calls = 0;
1654 if (ret == -1 && errno != EISCONN)
1655 {
1656 int xerrno = errno;
1657
1658 if (errno == EINTR)
1659 goto loop;
1660 if (errno == EADDRINUSE && retry < 20)
1661 {
1662 /* A delay here is needed on some FreeBSD systems,
1663 and it is harmless, since this retrying takes time anyway
1664 and should be infrequent.
1665 `sleep-for' allowed for quitting this loop with interrupts
1666 slowed down so it can't be used here. Async timers should
1667 already be disabled at this point so we can use `sleep'. */
1668 sleep (1);
1669 retry++;
1670 goto loop;
1671 }
1672
1673 close (rs);
1674 close (ws);
1675 speed_up_interrupts ();
1676
1677 errno = xerrno;
1678 report_file_error ("error connecting socket", list2(name, port));
1679 }
1680
1681 speed_up_interrupts ();
1682
1683 /* scope */
1684 if (setsockopt (ws, IPPROTO_IP, IP_MULTICAST_TTL,
1685 (char *) &thettl, sizeof (thettl)) < 0)
1686 {
1687 close (rs);
1688 close (ws);
1689 report_file_error ("error setting ttl", list2(name, ttl));
1690 }
1691
1692 set_socket_nonblocking_maybe (rs, theport, "udp");
1693
1694 *vinfd = (void*)rs;
1695 *voutfd = (void*)ws;
1696 }
1697
1698 #endif /* HAVE_MULTICAST */
1699
1700 #endif /* HAVE_SOCKETS */
1701
1702
1703 /**********************************************************************/
1704 /* Initialization */
1705 /**********************************************************************/
1706
1707 void
1708 process_type_create_unix (void)
1709 {
1710 PROCESS_HAS_METHOD (unix, alloc_process_data);
1711 PROCESS_HAS_METHOD (unix, mark_process_data);
1712 #ifdef SIGCHLD
1713 PROCESS_HAS_METHOD (unix, init_process);
1714 PROCESS_HAS_METHOD (unix, reap_exited_processes);
1715 #endif
1716 PROCESS_HAS_METHOD (unix, init_process_io_handles);
1717 PROCESS_HAS_METHOD (unix, create_process);
1718 PROCESS_HAS_METHOD (unix, tooltalk_connection_p);
1719 PROCESS_HAS_METHOD (unix, set_window_size);
1720 #ifdef HAVE_WAITPID
1721 PROCESS_HAS_METHOD (unix, update_status_if_terminated);
1722 #endif
1723 PROCESS_HAS_METHOD (unix, send_process);
1724 PROCESS_HAS_METHOD (unix, process_send_eof);
1725 PROCESS_HAS_METHOD (unix, deactivate_process);
1726 PROCESS_HAS_METHOD (unix, kill_child_process);
1727 PROCESS_HAS_METHOD (unix, kill_process_by_pid);
1728 PROCESS_HAS_METHOD (unix, get_tty_name);
1729 #ifdef HAVE_SOCKETS
1730 PROCESS_HAS_METHOD (unix, canonicalize_host_name);
1731 PROCESS_HAS_METHOD (unix, open_network_stream);
1732 #ifdef HAVE_MULTICAST
1733 PROCESS_HAS_METHOD (unix, open_multicast_group);
1734 #endif
1735 #endif
1736 }
1737
1738 void
1739 vars_of_process_unix (void)
1740 {
1741 Fprovide (intern ("unix-processes"));
1742 }
1743
1744 #endif /* !defined (NO_SUBPROCESSES) */