1 /* File Transfer Protocol support.
2 Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001
3 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2 of the License, or
10 (at your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget; if not, write to the Free Software
19 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 In addition, as a special exception, the Free Software Foundation
22 gives permission to link the code of its release of Wget with the
23 OpenSSL project's "OpenSSL" library (or with modified versions of it
24 that use the same license as the "OpenSSL" library), and distribute
25 the linked executables. You must obey the GNU General Public License
26 in all respects for all of the code used other than "OpenSSL". If you
27 modify this file, you may extend this exception to your version of the
28 file, but you are not obligated to do so. If you do not wish to do
29 so, delete this exception statement from your version. */
43 #include <sys/types.h>
55 #include "convert.h" /* for downloaded_file */
56 #include "recur.h" /* for INFINITE_RECURSION */
62 extern LARGE_INT total_downloaded_bytes;
64 /* File where the "ls -al" listing will be saved. */
65 #define LIST_FILENAME ".listing"
67 extern char ftp_last_respline[];
69 extern FILE *output_stream;
70 extern int output_stream_regular;
74 int st; /* connection status */
75 int cmd; /* command code */
76 int csock; /* control connection socket */
77 double dltime; /* time of the download in msecs */
78 enum stype rs; /* remote system reported by ftp server */
79 char *id; /* initial directory */
80 char *target; /* target file name */
81 struct url *proxy; /* FTWK-style proxy */
85 /* Look for regexp "( *[0-9]+ *byte" (literal parenthesis) anywhere in
86 the string S, and return the number converted to wgint, if found, 0
89 ftp_expected_bytes (const char *s)
95 while (*s && *s != '(')
99 ++s; /* skip the '(' */
100 res = str_to_wgint (s, (char **) &s, 10);
103 while (*s && ISSPACE (*s))
107 if (TOLOWER (*s) != 'b')
109 if (strncasecmp (s, "byte", 4))
119 * This function sets up a passive data connection with the FTP server.
120 * It is merely a wrapper around ftp_epsv, ftp_lpsv and ftp_pasv.
123 ftp_do_pasv (int csock, ip_address *addr, int *port)
127 /* We need to determine the address family and need to call
128 getpeername, so while we're at it, store the address to ADDR.
129 ftp_pasv and ftp_lpsv can simply override it. */
130 if (!socket_ip_address (csock, addr, ENDPOINT_PEER))
133 /* If our control connection is over IPv6, then we first try EPSV and then
134 * LPSV if the former is not supported. If the control connection is over
135 * IPv4, we simply issue the good old PASV request. */
139 if (!opt.server_response)
140 logputs (LOG_VERBOSE, "==> PASV ... ");
141 err = ftp_pasv (csock, addr, port);
144 if (!opt.server_response)
145 logputs (LOG_VERBOSE, "==> EPSV ... ");
146 err = ftp_epsv (csock, addr, port);
148 /* If EPSV is not supported try LPSV */
149 if (err == FTPNOPASV)
151 if (!opt.server_response)
152 logputs (LOG_VERBOSE, "==> LPSV ... ");
153 err = ftp_lpsv (csock, addr, port);
164 * This function sets up an active data connection with the FTP server.
165 * It is merely a wrapper around ftp_eprt, ftp_lprt and ftp_port.
168 ftp_do_port (int csock, int *local_sock)
173 if (!socket_ip_address (csock, &cip, ENDPOINT_PEER))
176 /* If our control connection is over IPv6, then we first try EPRT and then
177 * LPRT if the former is not supported. If the control connection is over
178 * IPv4, we simply issue the good old PORT request. */
182 if (!opt.server_response)
183 logputs (LOG_VERBOSE, "==> PORT ... ");
184 err = ftp_port (csock, local_sock);
187 if (!opt.server_response)
188 logputs (LOG_VERBOSE, "==> EPRT ... ");
189 err = ftp_eprt (csock, local_sock);
191 /* If EPRT is not supported try LPRT */
192 if (err == FTPPORTERR)
194 if (!opt.server_response)
195 logputs (LOG_VERBOSE, "==> LPRT ... ");
196 err = ftp_lprt (csock, local_sock);
207 ftp_do_pasv (int csock, ip_address *addr, int *port)
209 if (!opt.server_response)
210 logputs (LOG_VERBOSE, "==> PASV ... ");
211 return ftp_pasv (csock, addr, port);
215 ftp_do_port (int csock, int *local_sock)
217 if (!opt.server_response)
218 logputs (LOG_VERBOSE, "==> PORT ... ");
219 return ftp_port (csock, local_sock);
223 /* Retrieves a file with denoted parameters through opening an FTP
224 connection to the server. It always closes the data connection,
225 and closes the control connection in case of error. */
227 getftp (struct url *u, wgint *len, wgint restval, ccon *con)
229 int csock, dtsock, local_sock, res;
230 uerr_t err = RETROK; /* appease the compiler */
232 char *user, *passwd, *respline;
235 int pasv_mode_open = 0;
236 wgint expected_bytes = 0L;
241 assert (con != NULL);
242 assert (con->target != NULL);
244 /* Debug-check of the sanity of the request by making sure that LIST
245 and RETR are never both requested (since we can handle only one
247 assert (!((cmd & DO_LIST) && (cmd & DO_RETR)));
248 /* Make sure that at least *something* is requested. */
249 assert ((cmd & (DO_LIST | DO_CWD | DO_RETR | DO_LOGIN)) != 0);
253 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 1);
254 user = user ? user : opt.ftp_acc;
255 passwd = passwd ? passwd : opt.ftp_pass;
256 assert (user && passwd);
262 if (!(cmd & DO_LOGIN))
264 else /* cmd & DO_LOGIN */
267 char *host = con->proxy ? con->proxy->host : u->host;
268 int port = con->proxy ? con->proxy->port : u->port;
269 char *logname = user;
273 /* If proxy is in use, log in as username@target-site. */
274 logname = xmalloc (strlen (user) + 1 + strlen (u->host) + 1);
275 sprintf (logname, "%s@%s", user, u->host);
278 /* Login to the server: */
280 /* First: Establish the control connection. */
282 csock = connect_to_host (host, port);
286 return (retryable_socket_connect_error (errno)
287 ? CONERROR : CONIMPOSSIBLE);
289 if (cmd & LEAVE_PENDING)
294 /* Second: Login with proper USER/PASS sequence. */
295 logprintf (LOG_VERBOSE, _("Logging in as %s ... "), user);
296 if (opt.server_response)
297 logputs (LOG_ALWAYS, "\n");
298 err = ftp_login (csock, logname, passwd);
303 /* FTPRERR, FTPSRVERR, WRITEFAILED, FTPLOGREFUSED, FTPLOGINC */
307 logputs (LOG_VERBOSE, "\n");
308 logputs (LOG_NOTQUIET, _("\
309 Error in server response, closing control connection.\n"));
315 logputs (LOG_VERBOSE, "\n");
316 logputs (LOG_NOTQUIET, _("Error in server greeting.\n"));
322 logputs (LOG_VERBOSE, "\n");
323 logputs (LOG_NOTQUIET,
324 _("Write failed, closing control connection.\n"));
330 logputs (LOG_VERBOSE, "\n");
331 logputs (LOG_NOTQUIET, _("The server refuses login.\n"));
334 return FTPLOGREFUSED;
337 logputs (LOG_VERBOSE, "\n");
338 logputs (LOG_NOTQUIET, _("Login incorrect.\n"));
344 if (!opt.server_response)
345 logputs (LOG_VERBOSE, _("Logged in!\n"));
352 /* Third: Get the system type */
353 if (!opt.server_response)
354 logprintf (LOG_VERBOSE, "==> SYST ... ");
355 err = ftp_syst (csock, &con->rs);
360 logputs (LOG_VERBOSE, "\n");
361 logputs (LOG_NOTQUIET, _("\
362 Error in server response, closing control connection.\n"));
368 logputs (LOG_VERBOSE, "\n");
369 logputs (LOG_NOTQUIET,
370 _("Server error, can't determine system type.\n"));
373 /* Everything is OK. */
379 if (!opt.server_response && err != FTPSRVERR)
380 logputs (LOG_VERBOSE, _("done. "));
382 /* Fourth: Find the initial ftp directory */
384 if (!opt.server_response)
385 logprintf (LOG_VERBOSE, "==> PWD ... ");
386 err = ftp_pwd (csock, &con->id);
391 logputs (LOG_VERBOSE, "\n");
392 logputs (LOG_NOTQUIET, _("\
393 Error in server response, closing control connection.\n"));
399 /* PWD unsupported -- assume "/". */
400 xfree_null (con->id);
401 con->id = xstrdup ("/");
404 /* Everything is OK. */
410 /* VMS will report something like "PUB$DEVICE:[INITIAL.FOLDER]".
411 Convert it to "/INITIAL/FOLDER" */
412 if (con->rs == ST_VMS)
414 char *path = strchr (con->id, '[');
415 char *pathend = path ? strchr (path + 1, ']') : NULL;
416 if (!path || !pathend)
417 DEBUGP (("Initial VMS directory not in the form [...]!\n"));
420 char *idir = con->id;
421 DEBUGP (("Preprocessing the initial VMS directory\n"));
422 DEBUGP ((" old = '%s'\n", con->id));
423 /* We do the conversion in-place by copying the stuff
424 between [ and ] to the beginning, and changing dots
425 to slashes at the same time. */
427 for (++path; path < pathend; path++, idir++)
428 *idir = *path == '.' ? '/' : *path;
430 DEBUGP ((" new = '%s'\n\n", con->id));
433 if (!opt.server_response)
434 logputs (LOG_VERBOSE, _("done.\n"));
436 /* Fifth: Set the FTP type. */
437 type_char = ftp_process_type (u->params);
438 if (!opt.server_response)
439 logprintf (LOG_VERBOSE, "==> TYPE %c ... ", type_char);
440 err = ftp_type (csock, type_char);
441 /* FTPRERR, WRITEFAILED, FTPUNKNOWNTYPE */
445 logputs (LOG_VERBOSE, "\n");
446 logputs (LOG_NOTQUIET, _("\
447 Error in server response, closing control connection.\n"));
453 logputs (LOG_VERBOSE, "\n");
454 logputs (LOG_NOTQUIET,
455 _("Write failed, closing control connection.\n"));
461 logputs (LOG_VERBOSE, "\n");
462 logprintf (LOG_NOTQUIET,
463 _("Unknown type `%c', closing control connection.\n"),
469 /* Everything is OK. */
475 if (!opt.server_response)
476 logputs (LOG_VERBOSE, _("done. "));
482 logputs (LOG_VERBOSE, _("==> CWD not needed.\n"));
485 char *target = u->dir;
487 DEBUGP (("changing working directory\n"));
489 /* Change working directory. To change to a non-absolute
490 Unix directory, we need to prepend initial directory
491 (con->id) to it. Absolute directories "just work".
493 A relative directory is one that does not begin with '/'
494 and, on non-Unix OS'es, one that doesn't begin with
497 This is not done for OS400, which doesn't use
498 "/"-delimited directories, nor does it support directory
499 hierarchies. "CWD foo" followed by "CWD bar" leaves us
500 in "bar", not in "foo/bar", as would be customary
504 && !(con->rs != ST_UNIX
505 && ISALPHA (target[0])
507 && con->rs != ST_OS400)
509 int idlen = strlen (con->id);
512 /* Strip trailing slash(es) from con->id. */
513 while (idlen > 0 && con->id[idlen - 1] == '/')
515 p = ntarget = (char *)alloca (idlen + 1 + strlen (u->dir) + 1);
516 memcpy (p, con->id, idlen);
521 DEBUGP (("Prepended initial PWD to relative path:\n"));
522 DEBUGP ((" pwd: '%s'\n old: '%s'\n new: '%s'\n",
523 con->id, target, ntarget));
527 /* If the FTP host runs VMS, we will have to convert the absolute
528 directory path in UNIX notation to absolute directory path in
529 VMS notation as VMS FTP servers do not like UNIX notation of
530 absolute paths. "VMS notation" is [dir.subdir.subsubdir]. */
532 if (con->rs == ST_VMS)
535 char *ntarget = (char *)alloca (strlen (target) + 2);
536 /* We use a converted initial dir, so directories in
537 TARGET will be separated with slashes, something like
538 "/INITIAL/FOLDER/DIR/SUBDIR". Convert that to
539 "[INITIAL.FOLDER.DIR.SUBDIR]". */
540 strcpy (ntarget, target);
541 assert (*ntarget == '/');
543 for (tmpp = ntarget + 1; *tmpp; tmpp++)
548 DEBUGP (("Changed file name to VMS syntax:\n"));
549 DEBUGP ((" Unix: '%s'\n VMS: '%s'\n", target, ntarget));
553 if (!opt.server_response)
554 logprintf (LOG_VERBOSE, "==> CWD %s ... ", target);
555 err = ftp_cwd (csock, target);
556 /* FTPRERR, WRITEFAILED, FTPNSFOD */
560 logputs (LOG_VERBOSE, "\n");
561 logputs (LOG_NOTQUIET, _("\
562 Error in server response, closing control connection.\n"));
568 logputs (LOG_VERBOSE, "\n");
569 logputs (LOG_NOTQUIET,
570 _("Write failed, closing control connection.\n"));
576 logputs (LOG_VERBOSE, "\n");
577 logprintf (LOG_NOTQUIET, _("No such directory `%s'.\n\n"),
590 if (!opt.server_response)
591 logputs (LOG_VERBOSE, _("done.\n"));
594 else /* do not CWD */
595 logputs (LOG_VERBOSE, _("==> CWD not required.\n"));
597 if ((cmd & DO_RETR) && restval && *len == 0)
601 if (!opt.server_response)
602 logprintf (LOG_VERBOSE, "==> SIZE %s ... ", u->file);
605 err = ftp_size (csock, u->file, len);
611 logputs (LOG_VERBOSE, "\n");
612 logputs (LOG_NOTQUIET, _("\
613 Error in server response, closing control connection.\n"));
619 /* Everything is OK. */
625 if (!opt.server_response)
626 logputs (LOG_VERBOSE, _("done.\n"));
629 /* If anything is to be retrieved, PORT (or PASV) must be sent. */
630 if (cmd & (DO_LIST | DO_RETR))
632 if (opt.ftp_pasv > 0)
634 ip_address passive_addr;
636 err = ftp_do_pasv (csock, &passive_addr, &passive_port);
637 /* FTPRERR, WRITEFAILED, FTPNOPASV, FTPINVPASV */
641 logputs (LOG_VERBOSE, "\n");
642 logputs (LOG_NOTQUIET, _("\
643 Error in server response, closing control connection.\n"));
649 logputs (LOG_VERBOSE, "\n");
650 logputs (LOG_NOTQUIET,
651 _("Write failed, closing control connection.\n"));
657 logputs (LOG_VERBOSE, "\n");
658 logputs (LOG_NOTQUIET, _("Cannot initiate PASV transfer.\n"));
661 logputs (LOG_VERBOSE, "\n");
662 logputs (LOG_NOTQUIET, _("Cannot parse PASV response.\n"));
673 DEBUGP (("trying to connect to %s port %d\n",
674 pretty_print_address (&passive_addr),
676 dtsock = connect_to_ip (&passive_addr, passive_port, NULL);
679 int save_errno = errno;
682 logprintf (LOG_VERBOSE, _("couldn't connect to %s port %d: %s\n"),
683 pretty_print_address (&passive_addr), passive_port,
684 strerror (save_errno));
685 return (retryable_socket_connect_error (save_errno)
686 ? CONERROR : CONIMPOSSIBLE);
689 pasv_mode_open = 1; /* Flag to avoid accept port */
690 if (!opt.server_response)
691 logputs (LOG_VERBOSE, _("done. "));
695 if (!pasv_mode_open) /* Try to use a port command if PASV failed */
697 err = ftp_do_port (csock, &local_sock);
698 /* FTPRERR, WRITEFAILED, bindport (FTPSYSERR), HOSTERR,
703 logputs (LOG_VERBOSE, "\n");
704 logputs (LOG_NOTQUIET, _("\
705 Error in server response, closing control connection.\n"));
709 fd_close (local_sock);
713 logputs (LOG_VERBOSE, "\n");
714 logputs (LOG_NOTQUIET,
715 _("Write failed, closing control connection.\n"));
719 fd_close (local_sock);
723 logputs (LOG_VERBOSE, "\n");
724 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
728 fd_close (local_sock);
732 logputs (LOG_VERBOSE, "\n");
733 logprintf (LOG_NOTQUIET, _("Bind error (%s).\n"),
739 logputs (LOG_VERBOSE, "\n");
740 logputs (LOG_NOTQUIET, _("Invalid PORT.\n"));
744 fd_close (local_sock);
754 if (!opt.server_response)
755 logputs (LOG_VERBOSE, _("done. "));
757 } /* cmd & (DO_LIST | DO_RETR) */
759 /* Restart if needed. */
760 if (restval && (cmd & DO_RETR))
762 if (!opt.server_response)
763 logprintf (LOG_VERBOSE, "==> REST %s ... ", number_to_static_string (restval));
764 err = ftp_rest (csock, restval);
766 /* FTPRERR, WRITEFAILED, FTPRESTFAIL */
770 logputs (LOG_VERBOSE, "\n");
771 logputs (LOG_NOTQUIET, _("\
772 Error in server response, closing control connection.\n"));
776 fd_close (local_sock);
780 logputs (LOG_VERBOSE, "\n");
781 logputs (LOG_NOTQUIET,
782 _("Write failed, closing control connection.\n"));
786 fd_close (local_sock);
790 logputs (LOG_VERBOSE, _("\nREST failed, starting from scratch.\n"));
800 if (err != FTPRESTFAIL && !opt.server_response)
801 logputs (LOG_VERBOSE, _("done. "));
802 } /* restval && cmd & DO_RETR */
806 /* If we're in spider mode, don't really retrieve anything. The
807 fact that we got to this point should be proof enough that
808 the file exists, vaguely akin to HTTP's concept of a "HEAD"
815 fd_close (local_sock);
821 if (!opt.server_response)
824 logputs (LOG_VERBOSE, "\n");
825 logprintf (LOG_VERBOSE, "==> RETR %s ... ", u->file);
829 err = ftp_retr (csock, u->file);
830 /* FTPRERR, WRITEFAILED, FTPNSFOD */
834 logputs (LOG_VERBOSE, "\n");
835 logputs (LOG_NOTQUIET, _("\
836 Error in server response, closing control connection.\n"));
840 fd_close (local_sock);
844 logputs (LOG_VERBOSE, "\n");
845 logputs (LOG_NOTQUIET,
846 _("Write failed, closing control connection.\n"));
850 fd_close (local_sock);
854 logputs (LOG_VERBOSE, "\n");
855 logprintf (LOG_NOTQUIET, _("No such file `%s'.\n\n"), u->file);
857 fd_close (local_sock);
868 if (!opt.server_response)
869 logputs (LOG_VERBOSE, _("done.\n"));
870 expected_bytes = ftp_expected_bytes (ftp_last_respline);
875 if (!opt.server_response)
876 logputs (LOG_VERBOSE, "==> LIST ... ");
877 /* As Maciej W. Rozycki (macro@ds2.pg.gda.pl) says, `LIST'
878 without arguments is better than `LIST .'; confirmed by
880 err = ftp_list (csock, NULL);
881 /* FTPRERR, WRITEFAILED */
885 logputs (LOG_VERBOSE, "\n");
886 logputs (LOG_NOTQUIET, _("\
887 Error in server response, closing control connection.\n"));
891 fd_close (local_sock);
895 logputs (LOG_VERBOSE, "\n");
896 logputs (LOG_NOTQUIET,
897 _("Write failed, closing control connection.\n"));
901 fd_close (local_sock);
905 logputs (LOG_VERBOSE, "\n");
906 logprintf (LOG_NOTQUIET, _("No such file or directory `%s'.\n\n"),
909 fd_close (local_sock);
919 if (!opt.server_response)
920 logputs (LOG_VERBOSE, _("done.\n"));
921 expected_bytes = ftp_expected_bytes (ftp_last_respline);
922 } /* cmd & DO_LIST */
924 if (!(cmd & (DO_LIST | DO_RETR)) || (opt.spider && !(cmd & DO_LIST)))
927 /* Some FTP servers return the total length of file after REST
928 command, others just return the remaining size. */
929 if (*len && restval && expected_bytes
930 && (expected_bytes == *len - restval))
932 DEBUGP (("Lying FTP server found, adjusting.\n"));
933 expected_bytes = *len;
936 /* If no transmission was required, then everything is OK. */
937 if (!pasv_mode_open) /* we are not using pasive mode so we need
940 /* Wait for the server to connect to the address we're waiting
942 dtsock = accept_connection (local_sock);
945 logprintf (LOG_NOTQUIET, "accept: %s\n", strerror (errno));
950 /* Open the file -- if output_stream is set, use it instead. */
951 if (!output_stream || con->cmd & DO_LIST)
953 mkalldirs (con->target);
955 rotate_backups (con->target);
958 fp = fopen (con->target, "ab");
959 else if (opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct
960 || opt.output_document)
961 fp = fopen (con->target, "wb");
964 fp = fopen_excl (con->target, 0);
965 if (!fp && errno == EEXIST)
967 /* We cannot just invent a new name and use it (which is
968 what functions like unique_create typically do)
969 because we told the user we'd use this name.
970 Instead, return and retry the download. */
971 logprintf (LOG_NOTQUIET, _("%s has sprung into existence.\n"),
976 fd_close (local_sock);
977 return FOPEN_EXCL_ERR;
982 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target, strerror (errno));
986 fd_close (local_sock);
995 logprintf (LOG_VERBOSE, _("Length: %s"), legible (*len));
997 logprintf (LOG_VERBOSE, _(" [%s to go]"), legible (*len - restval));
998 logputs (LOG_VERBOSE, "\n");
999 expected_bytes = *len; /* for get_contents/show_progress */
1001 else if (expected_bytes)
1003 logprintf (LOG_VERBOSE, _("Length: %s"), legible (expected_bytes));
1005 logprintf (LOG_VERBOSE, _(" [%s to go]"),
1006 legible (expected_bytes - restval));
1007 logputs (LOG_VERBOSE, _(" (unauthoritative)\n"));
1010 /* Get the contents of the document. */
1012 if (restval && rest_failed)
1013 flags |= rb_skip_startpos;
1016 res = fd_read_body (dtsock, fp,
1017 expected_bytes ? expected_bytes - restval : 0,
1018 restval, &rd_size, len, &con->dltime, flags);
1020 tms = time_str (NULL);
1021 tmrate = retr_rate (rd_size, con->dltime, 0);
1022 /* Close data connection socket. */
1024 fd_close (local_sock);
1025 /* Close the local file. */
1027 /* Close or flush the file. We have to be careful to check for
1028 error here. Checking the result of fwrite() is not enough --
1029 errors could go unnoticed! */
1031 if (!output_stream || con->cmd & DO_LIST)
1032 flush_res = fclose (fp);
1034 flush_res = fflush (fp);
1035 if (flush_res == EOF)
1039 /* If get_contents couldn't write to fp, bail out. */
1042 logprintf (LOG_NOTQUIET, _("%s: %s, closing control connection.\n"),
1043 con->target, strerror (errno));
1050 logprintf (LOG_NOTQUIET, _("%s (%s) - Data connection: %s; "),
1051 tms, tmrate, strerror (errno));
1052 if (opt.server_response)
1053 logputs (LOG_ALWAYS, "\n");
1056 /* Get the server to tell us if everything is retrieved. */
1057 err = ftp_response (csock, &respline);
1061 /* The control connection is decidedly closed. Print the time
1062 only if it hasn't already been printed. */
1064 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
1065 logputs (LOG_NOTQUIET, _("Control connection closed.\n"));
1066 /* If there is an error on the control connection, close it, but
1067 return FTPRETRINT, since there is a possibility that the
1068 whole file was retrieved nevertheless (but that is for
1069 ftp_loop_internal to decide). */
1073 } /* err != FTPOK */
1074 /* If retrieval failed for any reason, return FTPRETRINT, but do not
1075 close socket, since the control connection is still alive. If
1076 there is something wrong with the control connection, it will
1077 become apparent later. */
1078 if (*respline != '2')
1082 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
1083 logputs (LOG_NOTQUIET, _("Data transfer aborted.\n"));
1090 /* What now? The data connection was erroneous, whereas the
1091 response says everything is OK. We shall play it safe. */
1095 if (!(cmd & LEAVE_PENDING))
1097 /* Closing the socket is faster than sending 'QUIT' and the
1098 effect is the same. */
1102 /* If it was a listing, and opt.server_response is true,
1104 if (opt.server_response && (con->cmd & DO_LIST))
1106 mkalldirs (con->target);
1107 fp = fopen (con->target, "r");
1109 logprintf (LOG_ALWAYS, "%s: %s\n", con->target, strerror (errno));
1113 /* The lines are being read with read_whole_line because of
1114 no-buffering on opt.lfile. */
1115 while ((line = read_whole_line (fp)))
1117 logprintf (LOG_ALWAYS, "%s\n", line);
1122 } /* con->cmd & DO_LIST && server_response */
1124 return RETRFINISHED;
1127 /* A one-file FTP loop. This is the part where FTP retrieval is
1128 retried, and retried, and retried, and...
1130 This loop either gets commands from con, or (if ON_YOUR_OWN is
1131 set), makes them up to retrieve the file given by the URL. */
1133 ftp_loop_internal (struct url *u, struct fileinfo *f, ccon *con)
1136 wgint restval, len = 0;
1138 char *tmrate = NULL;
1143 con->target = url_file_name (u);
1145 if (opt.noclobber && file_exists_p (con->target))
1147 logprintf (LOG_VERBOSE,
1148 _("File `%s' already there, not retrieving.\n"), con->target);
1149 /* If the file is there, we suppose it's retrieved OK. */
1153 /* Remove it if it's a link. */
1154 remove_link (con->target);
1155 if (!opt.output_document)
1158 locf = opt.output_document;
1162 if (con->st & ON_YOUR_OWN)
1163 con->st = ON_YOUR_OWN;
1165 orig_lp = con->cmd & LEAVE_PENDING ? 1 : 0;
1170 /* Increment the pass counter. */
1172 sleep_between_retrievals (count);
1173 if (con->st & ON_YOUR_OWN)
1176 con->cmd |= (DO_RETR | LEAVE_PENDING);
1177 if (con->csock != -1)
1178 con->cmd &= ~ (DO_LOGIN | DO_CWD);
1180 con->cmd |= (DO_LOGIN | DO_CWD);
1182 else /* not on your own */
1184 if (con->csock != -1)
1185 con->cmd &= ~DO_LOGIN;
1187 con->cmd |= DO_LOGIN;
1188 if (con->st & DONE_CWD)
1189 con->cmd &= ~DO_CWD;
1194 /* Decide whether or not to restart. */
1197 restval = len; /* start where the previous run left off */
1198 else if (opt.always_rest
1199 && stat (locf, &st) == 0
1200 && S_ISREG (st.st_mode))
1201 restval = st.st_size;
1203 /* Get the current time string. */
1204 tms = time_str (NULL);
1205 /* Print fetch message, if opt.verbose. */
1208 char *hurl = url_string (u, 1);
1212 sprintf (tmp, _("(try:%2d)"), count);
1213 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1214 tms, hurl, tmp, locf);
1216 ws_changetitle (hurl);
1220 /* Send getftp the proper length, if fileinfo was provided. */
1225 err = getftp (u, &len, restval, con);
1227 if (con->csock != -1)
1228 con->st &= ~DONE_CWD;
1230 con->st |= DONE_CWD;
1234 case HOSTERR: case CONIMPOSSIBLE: case FWRITEERR: case FOPENERR:
1235 case FTPNSFOD: case FTPLOGINC: case FTPNOPASV: case CONTNOTSUPPORTED:
1236 /* Fatal errors, give up. */
1239 case CONSOCKERR: case CONERROR: case FTPSRVERR: case FTPRERR:
1240 case WRITEFAILED: case FTPUNKNOWNTYPE: case FTPSYSERR:
1241 case FTPPORTERR: case FTPLOGREFUSED: case FTPINVPASV:
1242 case FOPEN_EXCL_ERR:
1243 printwhat (count, opt.ntry);
1244 /* non-fatal errors */
1245 if (err == FOPEN_EXCL_ERR)
1247 /* Re-determine the file name. */
1248 xfree_null (con->target);
1249 con->target = url_file_name (u);
1255 /* If the control connection was closed, the retrieval
1256 will be considered OK if f->size == len. */
1257 if (!f || len != f->size)
1259 printwhat (count, opt.ntry);
1270 tms = time_str (NULL);
1272 tmrate = retr_rate (len - restval, con->dltime, 0);
1274 /* If we get out of the switch above without continue'ing, we've
1275 successfully downloaded a file. Remember this fact. */
1276 downloaded_file (FILE_DOWNLOADED_NORMALLY, locf);
1278 if (con->st & ON_YOUR_OWN)
1280 fd_close (con->csock);
1284 logprintf (LOG_VERBOSE, _("%s (%s) - `%s' saved [%s]\n\n"),
1285 tms, tmrate, locf, number_to_static_string (len));
1286 if (!opt.verbose && !opt.quiet)
1288 /* Need to hide the password from the URL. The `if' is here
1289 so that we don't do the needless allocation every
1291 char *hurl = url_string (u, 1);
1292 logprintf (LOG_NONVERBOSE, "%s URL: %s [%s] -> \"%s\" [%d]\n",
1293 tms, hurl, number_to_static_string (len), locf, count);
1297 if ((con->cmd & DO_LIST))
1298 /* This is a directory listing file. */
1300 if (!opt.remove_listing)
1301 /* --dont-remove-listing was specified, so do count this towards the
1302 number of bytes and files downloaded. */
1304 total_downloaded_bytes += len;
1308 /* Deletion of listing files is not controlled by --delete-after, but
1309 by the more specific option --dont-remove-listing, and the code
1310 to do this deletion is in another function. */
1312 else if (!opt.spider)
1313 /* This is not a directory listing file. */
1315 /* Unlike directory listing files, don't pretend normal files weren't
1316 downloaded if they're going to be deleted. People seeding proxies,
1317 for instance, may want to know how many bytes and files they've
1318 downloaded through it. */
1319 total_downloaded_bytes += len;
1322 if (opt.delete_after)
1324 DEBUGP (("Removing file due to --delete-after in"
1325 " ftp_loop_internal():\n"));
1326 logprintf (LOG_VERBOSE, _("Removing %s.\n"), locf);
1328 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1332 /* Restore the original leave-pendingness. */
1334 con->cmd |= LEAVE_PENDING;
1336 con->cmd &= ~LEAVE_PENDING;
1338 } while (!opt.ntry || (count < opt.ntry));
1340 if (con->csock != -1 && (con->st & ON_YOUR_OWN))
1342 fd_close (con->csock);
1348 /* Return the directory listing in a reusable format. The directory
1349 is specifed in u->dir. */
1351 ftp_get_listing (struct url *u, ccon *con, struct fileinfo **f)
1354 char *uf; /* url file name */
1355 char *lf; /* list file name */
1356 char *old_target = con->target;
1358 con->st &= ~ON_YOUR_OWN;
1359 con->cmd |= (DO_LIST | LEAVE_PENDING);
1360 con->cmd &= ~DO_RETR;
1362 /* Find the listing file name. We do it by taking the file name of
1363 the URL and replacing the last component with the listing file
1365 uf = url_file_name (u);
1366 lf = file_merge (uf, LIST_FILENAME);
1368 DEBUGP ((_("Using `%s' as listing tmp file.\n"), lf));
1371 err = ftp_loop_internal (u, NULL, con);
1372 con->target = old_target;
1375 *f = ftp_parse_ls (lf, con->rs);
1378 if (opt.remove_listing)
1381 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1383 logprintf (LOG_VERBOSE, _("Removed `%s'.\n"), lf);
1386 con->cmd &= ~DO_LIST;
1390 static uerr_t ftp_retrieve_dirs PARAMS ((struct url *, struct fileinfo *,
1392 static uerr_t ftp_retrieve_glob PARAMS ((struct url *, ccon *, int));
1393 static struct fileinfo *delelement PARAMS ((struct fileinfo *,
1394 struct fileinfo **));
1395 static void freefileinfo PARAMS ((struct fileinfo *f));
1397 /* Retrieve a list of files given in struct fileinfo linked list. If
1398 a file is a symbolic link, do not retrieve it, but rather try to
1399 set up a similar link on the local disk, if the symlinks are
1402 If opt.recursive is set, after all files have been retrieved,
1403 ftp_retrieve_dirs will be called to retrieve the directories. */
1405 ftp_retrieve_list (struct url *u, struct fileinfo *f, ccon *con)
1407 static int depth = 0;
1409 struct fileinfo *orig;
1414 /* Increase the depth. */
1416 if (opt.reclevel != INFINITE_RECURSION && depth > opt.reclevel)
1418 DEBUGP ((_("Recursion depth %d exceeded max. depth %d.\n"),
1419 depth, opt.reclevel));
1427 con->st &= ~ON_YOUR_OWN;
1428 if (!(con->st & DONE_CWD))
1431 con->cmd &= ~DO_CWD;
1432 con->cmd |= (DO_RETR | LEAVE_PENDING);
1435 con->cmd |= DO_LOGIN;
1437 con->cmd &= ~DO_LOGIN;
1439 err = RETROK; /* in case it's not used */
1443 char *old_target, *ofile;
1445 if (opt.quota && total_downloaded_bytes > opt.quota)
1450 old_target = con->target;
1452 ofile = xstrdup (u->file);
1453 url_set_file (u, f->name);
1455 con->target = url_file_name (u);
1459 if (opt.timestamping && f->type == FT_PLAINFILE)
1462 /* If conversion of HTML files retrieved via FTP is ever implemented,
1463 we'll need to stat() <file>.orig here when -K has been specified.
1464 I'm not implementing it now since files on an FTP server are much
1465 more likely than files on an HTTP server to legitimately have a
1467 if (!stat (con->target, &st))
1471 /* Else, get it from the file. */
1472 local_size = st.st_size;
1475 /* Modification time granularity is 2 seconds for Windows, so
1476 increase local time by 1 second for later comparison. */
1479 /* Compare file sizes only for servers that tell us correct
1480 values. Assumme sizes being equal for servers that lie
1482 cor_val = (con->rs == ST_UNIX || con->rs == ST_WINNT);
1483 eq_size = cor_val ? (local_size == f->size) : 1 ;
1484 if (f->tstamp <= tml && eq_size)
1486 /* Remote file is older, file sizes can be compared and
1488 logprintf (LOG_VERBOSE, _("\
1489 Remote file no newer than local file `%s' -- not retrieving.\n"), con->target);
1494 /* Remote file is newer or sizes cannot be matched */
1495 logprintf (LOG_VERBOSE, _("\
1496 Remote file is newer than local file `%s' -- retrieving.\n\n"),
1501 /* Sizes do not match */
1502 logprintf (LOG_VERBOSE, _("\
1503 The sizes do not match (local %s) -- retrieving.\n\n"),
1504 number_to_static_string (local_size));
1507 } /* opt.timestamping && f->type == FT_PLAINFILE */
1511 /* If opt.retr_symlinks is defined, we treat symlinks as
1512 if they were normal files. There is currently no way
1513 to distinguish whether they might be directories, and
1515 if (!opt.retr_symlinks)
1519 logputs (LOG_NOTQUIET,
1520 _("Invalid name of the symlink, skipping.\n"));
1524 /* Check whether we already have the correct
1526 int rc = lstat (con->target, &st);
1529 size_t len = strlen (f->linkto) + 1;
1530 if (S_ISLNK (st.st_mode))
1532 char *link_target = (char *)alloca (len);
1533 size_t n = readlink (con->target, link_target, len);
1535 && (memcmp (link_target, f->linkto, n) == 0))
1537 logprintf (LOG_VERBOSE, _("\
1538 Already have correct symlink %s -> %s\n\n"),
1539 con->target, f->linkto);
1545 logprintf (LOG_VERBOSE, _("Creating symlink %s -> %s\n"),
1546 con->target, f->linkto);
1547 /* Unlink before creating symlink! */
1548 unlink (con->target);
1549 if (symlink (f->linkto, con->target) == -1)
1550 logprintf (LOG_NOTQUIET, "symlink: %s\n",
1552 logputs (LOG_VERBOSE, "\n");
1553 } /* have f->linkto */
1554 #else /* not HAVE_SYMLINK */
1555 logprintf (LOG_NOTQUIET,
1556 _("Symlinks not supported, skipping symlink `%s'.\n"),
1558 #endif /* not HAVE_SYMLINK */
1560 else /* opt.retr_symlinks */
1563 err = ftp_loop_internal (u, f, con);
1564 } /* opt.retr_symlinks */
1568 logprintf (LOG_NOTQUIET, _("Skipping directory `%s'.\n"),
1572 /* Call the retrieve loop. */
1574 err = ftp_loop_internal (u, f, con);
1577 logprintf (LOG_NOTQUIET, _("%s: unknown/unsupported file type.\n"),
1582 /* Set the time-stamp information to the local file. Symlinks
1583 are not to be stamped because it sets the stamp on the
1585 if (!(f->type == FT_SYMLINK && !opt.retr_symlinks)
1588 && file_exists_p (con->target))
1590 /* #### This code repeats in http.c and ftp.c. Move it to a
1592 const char *fl = NULL;
1593 if (opt.output_document)
1595 if (output_stream_regular)
1596 fl = opt.output_document;
1601 touch (fl, f->tstamp);
1603 else if (f->tstamp == -1)
1604 logprintf (LOG_NOTQUIET, _("%s: corrupt time-stamp.\n"), con->target);
1606 if (f->perms && f->type == FT_PLAINFILE && dlthis)
1608 if (opt.preserve_perm)
1609 chmod (con->target, f->perms);
1612 DEBUGP (("Unrecognized permissions for %s.\n", con->target));
1614 xfree (con->target);
1615 con->target = old_target;
1617 url_set_file (u, ofile);
1620 /* Break on fatals. */
1621 if (err == QUOTEXC || err == HOSTERR || err == FWRITEERR)
1623 con->cmd &= ~ (DO_CWD | DO_LOGIN);
1627 /* We do not want to call ftp_retrieve_dirs here */
1628 if (opt.recursive &&
1629 !(opt.reclevel != INFINITE_RECURSION && depth >= opt.reclevel))
1630 err = ftp_retrieve_dirs (u, orig, con);
1631 else if (opt.recursive)
1632 DEBUGP ((_("Will not retrieve dirs since depth is %d (max %d).\n"),
1633 depth, opt.reclevel));
1638 /* Retrieve the directories given in a file list. This function works
1639 by simply going through the linked list and calling
1640 ftp_retrieve_glob on each directory entry. The function knows
1641 about excluded directories. */
1643 ftp_retrieve_dirs (struct url *u, struct fileinfo *f, ccon *con)
1645 char *container = NULL;
1646 int container_size = 0;
1648 for (; f; f = f->next)
1651 char *odir, *newdir;
1653 if (opt.quota && total_downloaded_bytes > opt.quota)
1655 if (f->type != FT_DIRECTORY)
1658 /* Allocate u->dir off stack, but reallocate only if a larger
1659 string is needed. It's a pity there's no "realloca" for an
1660 item on the bottom of the stack. */
1661 size = strlen (u->dir) + 1 + strlen (f->name) + 1;
1662 if (size > container_size)
1663 container = (char *)alloca (size);
1668 || (*odir == '/' && *(odir + 1) == '\0'))
1669 /* If ODIR is empty or just "/", simply append f->name to
1670 ODIR. (In the former case, to preserve u->dir being
1671 relative; in the latter case, to avoid double slash.) */
1672 sprintf (newdir, "%s%s", odir, f->name);
1674 /* Else, use a separator. */
1675 sprintf (newdir, "%s/%s", odir, f->name);
1677 DEBUGP (("Composing new CWD relative to the initial directory.\n"));
1678 DEBUGP ((" odir = '%s'\n f->name = '%s'\n newdir = '%s'\n\n",
1679 odir, f->name, newdir));
1680 if (!accdir (newdir, ALLABS))
1682 logprintf (LOG_VERBOSE, _("\
1683 Not descending to `%s' as it is excluded/not-included.\n"), newdir);
1687 con->st &= ~DONE_CWD;
1689 odir = xstrdup (u->dir); /* because url_set_dir will free
1691 url_set_dir (u, newdir);
1692 ftp_retrieve_glob (u, con, GETALL);
1693 url_set_dir (u, odir);
1696 /* Set the time-stamp? */
1699 if (opt.quota && total_downloaded_bytes > opt.quota)
1705 /* Return non-zero if S has a leading '/' or contains '../' */
1707 has_insecure_name_p (const char *s)
1712 if (strstr (s, "../") != 0)
1718 /* A near-top-level function to retrieve the files in a directory.
1719 The function calls ftp_get_listing, to get a linked list of files.
1720 Then it weeds out the file names that do not match the pattern.
1721 ftp_retrieve_list is called with this updated list as an argument.
1723 If the argument ACTION is GETONE, just download the file (but first
1724 get the listing, so that the time-stamp is heeded); if it's GLOBALL,
1725 use globbing; if it's GETALL, download the whole directory. */
1727 ftp_retrieve_glob (struct url *u, ccon *con, int action)
1729 struct fileinfo *f, *start;
1732 con->cmd |= LEAVE_PENDING;
1734 res = ftp_get_listing (u, con, &start);
1737 /* First: weed out that do not conform the global rules given in
1738 opt.accepts and opt.rejects. */
1739 if (opt.accepts || opt.rejects)
1744 if (f->type != FT_DIRECTORY && !acceptable (f->name))
1746 logprintf (LOG_VERBOSE, _("Rejecting `%s'.\n"), f->name);
1747 f = delelement (f, &start);
1753 /* Remove all files with possible harmful names */
1757 if (has_insecure_name_p (f->name))
1759 logprintf (LOG_VERBOSE, _("Rejecting `%s'.\n"), f->name);
1760 f = delelement (f, &start);
1765 /* Now weed out the files that do not match our globbing pattern.
1766 If we are dealing with a globbing pattern, that is. */
1767 if (*u->file && (action == GLOBALL || action == GETONE))
1774 matchres = fnmatch (u->file, f->name, 0);
1777 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target,
1781 if (matchres == FNM_NOMATCH)
1782 f = delelement (f, &start); /* delete the element from the list */
1784 f = f->next; /* leave the element in the list */
1788 freefileinfo (start);
1789 return RETRBADPATTERN;
1795 /* Just get everything. */
1796 ftp_retrieve_list (u, start, con);
1800 if (action == GLOBALL)
1803 /* #### This message SUCKS. We should see what was the
1804 reason that nothing was retrieved. */
1805 logprintf (LOG_VERBOSE, _("No matches on pattern `%s'.\n"), u->file);
1807 else /* GETONE or GETALL */
1809 /* Let's try retrieving it anyway. */
1810 con->st |= ON_YOUR_OWN;
1811 res = ftp_loop_internal (u, NULL, con);
1815 freefileinfo (start);
1816 if (opt.quota && total_downloaded_bytes > opt.quota)
1819 /* #### Should we return `res' here? */
1823 /* The wrapper that calls an appropriate routine according to contents
1824 of URL. Inherently, its capabilities are limited on what can be
1825 encoded into a URL. */
1827 ftp_loop (struct url *u, int *dt, struct url *proxy)
1829 ccon con; /* FTP connection */
1834 memset (&con, 0, sizeof (con));
1837 con.st = ON_YOUR_OWN;
1841 res = RETROK; /* in case it's not used */
1843 /* If the file name is empty, the user probably wants a directory
1844 index. We'll provide one, properly HTML-ized. Unless
1845 opt.htmlify is 0, of course. :-) */
1846 if (!*u->file && !opt.recursive)
1849 res = ftp_get_listing (u, &con, &f);
1853 if (opt.htmlify && !opt.spider)
1855 char *filename = (opt.output_document
1856 ? xstrdup (opt.output_document)
1857 : (con.target ? xstrdup (con.target)
1858 : url_file_name (u)));
1859 res = ftp_index (filename, u, f);
1860 if (res == FTPOK && opt.verbose)
1862 if (!opt.output_document)
1866 if (stat (filename, &st) == 0)
1870 logprintf (LOG_NOTQUIET,
1871 _("Wrote HTML-ized index to `%s' [%s].\n"),
1872 filename, number_to_static_string (sz));
1875 logprintf (LOG_NOTQUIET,
1876 _("Wrote HTML-ized index to `%s'.\n"),
1886 int wild = has_wildcards_p (u->file);
1887 if ((opt.ftp_glob && wild) || opt.recursive || opt.timestamping)
1889 /* ftp_retrieve_glob is a catch-all function that gets called
1890 if we need globbing, time-stamping or recursion. Its
1891 third argument is just what we really need. */
1892 res = ftp_retrieve_glob (u, &con,
1893 (opt.ftp_glob && wild) ? GLOBALL : GETONE);
1896 res = ftp_loop_internal (u, NULL, &con);
1902 /* If a connection was left, quench it. */
1903 if (con.csock != -1)
1904 fd_close (con.csock);
1905 xfree_null (con.id);
1907 xfree_null (con.target);
1912 /* Delete an element from the fileinfo linked list. Returns the
1913 address of the next element, or NULL if the list is exhausted. It
1914 can modify the start of the list. */
1915 static struct fileinfo *
1916 delelement (struct fileinfo *f, struct fileinfo **start)
1918 struct fileinfo *prev = f->prev;
1919 struct fileinfo *next = f->next;
1922 xfree_null (f->linkto);
1934 /* Free the fileinfo linked list of files. */
1936 freefileinfo (struct fileinfo *f)
1940 struct fileinfo *next = f->next;