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[];
71 int st; /* connection status */
72 int cmd; /* command code */
73 int csock; /* control connection socket */
74 double dltime; /* time of the download in msecs */
75 enum stype rs; /* remote system reported by ftp server */
76 char *id; /* initial directory */
77 char *target; /* target file name */
78 struct url *proxy; /* FTWK-style proxy */
82 /* Look for regexp "( *[0-9]+ *byte" (literal parenthesis) anywhere in
83 the string S, and return the number converted to long, if found, 0
86 ftp_expected_bytes (const char *s)
92 while (*s && *s != '(')
96 for (++s; *s && ISSPACE (*s); s++);
104 res = (*s - '0') + 10 * res;
107 while (*s && ISDIGIT (*s));
110 while (*s && ISSPACE (*s))
114 if (TOLOWER (*s) != 'b')
116 if (strncasecmp (s, "byte", 4))
126 * This function sets up a passive data connection with the FTP server.
127 * It is merely a wrapper around ftp_epsv, ftp_lpsv and ftp_pasv.
130 ftp_do_pasv (int csock, ip_address *addr, int *port)
134 /* We need to determine the address family and need to call
135 getpeername, so while we're at it, store the address to ADDR.
136 ftp_pasv and ftp_lpsv can simply override it. */
137 if (!socket_ip_address (csock, addr, ENDPOINT_PEER))
140 /* If our control connection is over IPv6, then we first try EPSV and then
141 * LPSV if the former is not supported. If the control connection is over
142 * IPv4, we simply issue the good old PASV request. */
146 if (!opt.server_response)
147 logputs (LOG_VERBOSE, "==> PASV ... ");
148 err = ftp_pasv (csock, addr, port);
151 if (!opt.server_response)
152 logputs (LOG_VERBOSE, "==> EPSV ... ");
153 err = ftp_epsv (csock, addr, port);
155 /* If EPSV is not supported try LPSV */
156 if (err == FTPNOPASV)
158 if (!opt.server_response)
159 logputs (LOG_VERBOSE, "==> LPSV ... ");
160 err = ftp_lpsv (csock, addr, port);
171 * This function sets up an active data connection with the FTP server.
172 * It is merely a wrapper around ftp_eprt, ftp_lprt and ftp_port.
175 ftp_do_port (int csock, int *local_sock)
180 if (!socket_ip_address (csock, &cip, ENDPOINT_PEER))
183 /* If our control connection is over IPv6, then we first try EPRT and then
184 * LPRT if the former is not supported. If the control connection is over
185 * IPv4, we simply issue the good old PORT request. */
189 if (!opt.server_response)
190 logputs (LOG_VERBOSE, "==> PORT ... ");
191 err = ftp_port (csock, local_sock);
194 if (!opt.server_response)
195 logputs (LOG_VERBOSE, "==> EPRT ... ");
196 err = ftp_eprt (csock, local_sock);
198 /* If EPRT is not supported try LPRT */
199 if (err == FTPPORTERR)
201 if (!opt.server_response)
202 logputs (LOG_VERBOSE, "==> LPRT ... ");
203 err = ftp_lprt (csock, local_sock);
214 ftp_do_pasv (int csock, ip_address *addr, int *port)
216 if (!opt.server_response)
217 logputs (LOG_VERBOSE, "==> PASV ... ");
218 return ftp_pasv (csock, addr, port);
222 ftp_do_port (int csock, int *local_sock)
224 if (!opt.server_response)
225 logputs (LOG_VERBOSE, "==> PORT ... ");
226 return ftp_port (csock, local_sock);
230 /* Retrieves a file with denoted parameters through opening an FTP
231 connection to the server. It always closes the data connection,
232 and closes the control connection in case of error. */
234 getftp (struct url *u, long *len, long restval, ccon *con)
236 int csock, dtsock, local_sock, res;
239 char *user, *passwd, *respline;
242 int pasv_mode_open = 0;
243 long expected_bytes = 0L;
245 assert (con != NULL);
246 assert (con->target != NULL);
248 /* Debug-check of the sanity of the request by making sure that LIST
249 and RETR are never both requested (since we can handle only one
251 assert (!((cmd & DO_LIST) && (cmd & DO_RETR)));
252 /* Make sure that at least *something* is requested. */
253 assert ((cmd & (DO_LIST | DO_CWD | DO_RETR | DO_LOGIN)) != 0);
257 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 1);
258 user = user ? user : opt.ftp_acc;
259 passwd = passwd ? passwd : opt.ftp_pass;
260 assert (user && passwd);
266 if (!(cmd & DO_LOGIN))
268 else /* cmd & DO_LOGIN */
271 char *host = con->proxy ? con->proxy->host : u->host;
272 int port = con->proxy ? con->proxy->port : u->port;
273 char *logname = user;
277 /* If proxy is in use, log in as username@target-site. */
278 logname = xmalloc (strlen (user) + 1 + strlen (u->host) + 1);
279 sprintf (logname, "%s@%s", user, u->host);
282 /* Login to the server: */
284 /* First: Establish the control connection. */
286 csock = connect_to_host (host, port);
290 return (retryable_socket_connect_error (errno)
291 ? CONERROR : CONIMPOSSIBLE);
293 if (cmd & LEAVE_PENDING)
298 /* Second: Login with proper USER/PASS sequence. */
299 logprintf (LOG_VERBOSE, _("Logging in as %s ... "), user);
300 if (opt.server_response)
301 logputs (LOG_ALWAYS, "\n");
302 err = ftp_login (csock, logname, passwd);
307 /* FTPRERR, FTPSRVERR, WRITEFAILED, FTPLOGREFUSED, FTPLOGINC */
311 logputs (LOG_VERBOSE, "\n");
312 logputs (LOG_NOTQUIET, _("\
313 Error in server response, closing control connection.\n"));
319 logputs (LOG_VERBOSE, "\n");
320 logputs (LOG_NOTQUIET, _("Error in server greeting.\n"));
326 logputs (LOG_VERBOSE, "\n");
327 logputs (LOG_NOTQUIET,
328 _("Write failed, closing control connection.\n"));
334 logputs (LOG_VERBOSE, "\n");
335 logputs (LOG_NOTQUIET, _("The server refuses login.\n"));
338 return FTPLOGREFUSED;
341 logputs (LOG_VERBOSE, "\n");
342 logputs (LOG_NOTQUIET, _("Login incorrect.\n"));
348 if (!opt.server_response)
349 logputs (LOG_VERBOSE, _("Logged in!\n"));
356 /* Third: Get the system type */
357 if (!opt.server_response)
358 logprintf (LOG_VERBOSE, "==> SYST ... ");
359 err = ftp_syst (csock, &con->rs);
364 logputs (LOG_VERBOSE, "\n");
365 logputs (LOG_NOTQUIET, _("\
366 Error in server response, closing control connection.\n"));
372 logputs (LOG_VERBOSE, "\n");
373 logputs (LOG_NOTQUIET,
374 _("Server error, can't determine system type.\n"));
377 /* Everything is OK. */
383 if (!opt.server_response && err != FTPSRVERR)
384 logputs (LOG_VERBOSE, _("done. "));
386 /* Fourth: Find the initial ftp directory */
388 if (!opt.server_response)
389 logprintf (LOG_VERBOSE, "==> PWD ... ");
390 err = ftp_pwd (csock, &con->id);
395 logputs (LOG_VERBOSE, "\n");
396 logputs (LOG_NOTQUIET, _("\
397 Error in server response, closing control connection.\n"));
403 /* PWD unsupported -- assume "/". */
404 xfree_null (con->id);
405 con->id = xstrdup ("/");
408 /* Everything is OK. */
414 /* VMS will report something like "PUB$DEVICE:[INITIAL.FOLDER]".
415 Convert it to "/INITIAL/FOLDER" */
416 if (con->rs == ST_VMS)
418 char *path = strchr (con->id, '[');
419 char *pathend = path ? strchr (path + 1, ']') : NULL;
420 if (!path || !pathend)
421 DEBUGP (("Initial VMS directory not in the form [...]!\n"));
424 char *idir = con->id;
425 DEBUGP (("Preprocessing the initial VMS directory\n"));
426 DEBUGP ((" old = '%s'\n", con->id));
427 /* We do the conversion in-place by copying the stuff
428 between [ and ] to the beginning, and changing dots
429 to slashes at the same time. */
431 for (++path; path < pathend; path++, idir++)
432 *idir = *path == '.' ? '/' : *path;
434 DEBUGP ((" new = '%s'\n\n", con->id));
437 if (!opt.server_response)
438 logputs (LOG_VERBOSE, _("done.\n"));
440 /* Fifth: Set the FTP type. */
441 type_char = ftp_process_type (u->params);
442 if (!opt.server_response)
443 logprintf (LOG_VERBOSE, "==> TYPE %c ... ", type_char);
444 err = ftp_type (csock, type_char);
445 /* FTPRERR, WRITEFAILED, FTPUNKNOWNTYPE */
449 logputs (LOG_VERBOSE, "\n");
450 logputs (LOG_NOTQUIET, _("\
451 Error in server response, closing control connection.\n"));
457 logputs (LOG_VERBOSE, "\n");
458 logputs (LOG_NOTQUIET,
459 _("Write failed, closing control connection.\n"));
465 logputs (LOG_VERBOSE, "\n");
466 logprintf (LOG_NOTQUIET,
467 _("Unknown type `%c', closing control connection.\n"),
473 /* Everything is OK. */
479 if (!opt.server_response)
480 logputs (LOG_VERBOSE, _("done. "));
486 logputs (LOG_VERBOSE, _("==> CWD not needed.\n"));
489 char *target = u->dir;
491 DEBUGP (("changing working directory\n"));
493 /* Change working directory. To change to a non-absolute
494 Unix directory, we need to prepend initial directory
495 (con->id) to it. Absolute directories "just work".
497 A relative directory is one that does not begin with '/'
498 and, on non-Unix OS'es, one that doesn't begin with
501 This is not done for OS400, which doesn't use
502 "/"-delimited directories, nor does it support directory
503 hierarchies. "CWD foo" followed by "CWD bar" leaves us
504 in "bar", not in "foo/bar", as would be customary
508 && !(con->rs != ST_UNIX
509 && ISALPHA (target[0])
511 && con->rs != ST_OS400)
513 int idlen = strlen (con->id);
516 /* Strip trailing slash(es) from con->id. */
517 while (idlen > 0 && con->id[idlen - 1] == '/')
519 p = ntarget = (char *)alloca (idlen + 1 + strlen (u->dir) + 1);
520 memcpy (p, con->id, idlen);
525 DEBUGP (("Prepended initial PWD to relative path:\n"));
526 DEBUGP ((" pwd: '%s'\n old: '%s'\n new: '%s'\n",
527 con->id, target, ntarget));
531 /* If the FTP host runs VMS, we will have to convert the absolute
532 directory path in UNIX notation to absolute directory path in
533 VMS notation as VMS FTP servers do not like UNIX notation of
534 absolute paths. "VMS notation" is [dir.subdir.subsubdir]. */
536 if (con->rs == ST_VMS)
539 char *ntarget = (char *)alloca (strlen (target) + 2);
540 /* We use a converted initial dir, so directories in
541 TARGET will be separated with slashes, something like
542 "/INITIAL/FOLDER/DIR/SUBDIR". Convert that to
543 "[INITIAL.FOLDER.DIR.SUBDIR]". */
544 strcpy (ntarget, target);
545 assert (*ntarget == '/');
547 for (tmpp = ntarget + 1; *tmpp; tmpp++)
552 DEBUGP (("Changed file name to VMS syntax:\n"));
553 DEBUGP ((" Unix: '%s'\n VMS: '%s'\n", target, ntarget));
557 if (!opt.server_response)
558 logprintf (LOG_VERBOSE, "==> CWD %s ... ", target);
559 err = ftp_cwd (csock, target);
560 /* FTPRERR, WRITEFAILED, FTPNSFOD */
564 logputs (LOG_VERBOSE, "\n");
565 logputs (LOG_NOTQUIET, _("\
566 Error in server response, closing control connection.\n"));
572 logputs (LOG_VERBOSE, "\n");
573 logputs (LOG_NOTQUIET,
574 _("Write failed, closing control connection.\n"));
580 logputs (LOG_VERBOSE, "\n");
581 logprintf (LOG_NOTQUIET, _("No such directory `%s'.\n\n"),
594 if (!opt.server_response)
595 logputs (LOG_VERBOSE, _("done.\n"));
598 else /* do not CWD */
599 logputs (LOG_VERBOSE, _("==> CWD not required.\n"));
601 if ((cmd & DO_RETR) && restval && *len == 0)
605 if (!opt.server_response)
606 logprintf (LOG_VERBOSE, "==> SIZE %s ... ", u->file);
609 err = ftp_size (csock, u->file, len);
615 logputs (LOG_VERBOSE, "\n");
616 logputs (LOG_NOTQUIET, _("\
617 Error in server response, closing control connection.\n"));
623 /* Everything is OK. */
629 if (!opt.server_response)
630 logputs (LOG_VERBOSE, _("done.\n"));
633 /* If anything is to be retrieved, PORT (or PASV) must be sent. */
634 if (cmd & (DO_LIST | DO_RETR))
636 if (opt.ftp_pasv > 0)
638 ip_address passive_addr;
640 err = ftp_do_pasv (csock, &passive_addr, &passive_port);
641 /* FTPRERR, WRITEFAILED, FTPNOPASV, FTPINVPASV */
645 logputs (LOG_VERBOSE, "\n");
646 logputs (LOG_NOTQUIET, _("\
647 Error in server response, closing control connection.\n"));
653 logputs (LOG_VERBOSE, "\n");
654 logputs (LOG_NOTQUIET,
655 _("Write failed, closing control connection.\n"));
661 logputs (LOG_VERBOSE, "\n");
662 logputs (LOG_NOTQUIET, _("Cannot initiate PASV transfer.\n"));
665 logputs (LOG_VERBOSE, "\n");
666 logputs (LOG_NOTQUIET, _("Cannot parse PASV response.\n"));
677 DEBUGP (("trying to connect to %s port %d\n",
678 pretty_print_address (&passive_addr),
680 dtsock = connect_to_ip (&passive_addr, passive_port, NULL);
683 int save_errno = errno;
686 logprintf (LOG_VERBOSE, _("couldn't connect to %s port %d: %s\n"),
687 pretty_print_address (&passive_addr), passive_port,
688 strerror (save_errno));
689 return (retryable_socket_connect_error (save_errno)
690 ? CONERROR : CONIMPOSSIBLE);
693 pasv_mode_open = 1; /* Flag to avoid accept port */
694 if (!opt.server_response)
695 logputs (LOG_VERBOSE, _("done. "));
699 if (!pasv_mode_open) /* Try to use a port command if PASV failed */
701 err = ftp_do_port (csock, &local_sock);
702 /* FTPRERR, WRITEFAILED, bindport (FTPSYSERR), HOSTERR,
707 logputs (LOG_VERBOSE, "\n");
708 logputs (LOG_NOTQUIET, _("\
709 Error in server response, closing control connection.\n"));
713 fd_close (local_sock);
717 logputs (LOG_VERBOSE, "\n");
718 logputs (LOG_NOTQUIET,
719 _("Write failed, closing control connection.\n"));
723 fd_close (local_sock);
727 logputs (LOG_VERBOSE, "\n");
728 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
732 fd_close (local_sock);
736 logputs (LOG_VERBOSE, "\n");
737 logprintf (LOG_NOTQUIET, _("Bind error (%s).\n"),
743 logputs (LOG_VERBOSE, "\n");
744 logputs (LOG_NOTQUIET, _("Invalid PORT.\n"));
748 fd_close (local_sock);
758 if (!opt.server_response)
759 logputs (LOG_VERBOSE, _("done. "));
761 } /* cmd & (DO_LIST | DO_RETR) */
763 /* Restart if needed. */
764 if (restval && (cmd & DO_RETR))
766 if (!opt.server_response)
767 logprintf (LOG_VERBOSE, "==> REST %ld ... ", restval);
768 err = ftp_rest (csock, restval);
770 /* FTPRERR, WRITEFAILED, FTPRESTFAIL */
774 logputs (LOG_VERBOSE, "\n");
775 logputs (LOG_NOTQUIET, _("\
776 Error in server response, closing control connection.\n"));
780 fd_close (local_sock);
784 logputs (LOG_VERBOSE, "\n");
785 logputs (LOG_NOTQUIET,
786 _("Write failed, closing control connection.\n"));
790 fd_close (local_sock);
794 /* If `-c' is specified and the file already existed when
795 Wget was started, it would be a bad idea for us to start
796 downloading it from scratch, effectively truncating it. */
797 if (opt.always_rest && (cmd & NO_TRUNCATE))
799 logprintf (LOG_NOTQUIET,
800 _("\nREST failed; will not truncate `%s'.\n"),
805 fd_close (local_sock);
806 return CONTNOTSUPPORTED;
808 logputs (LOG_VERBOSE, _("\nREST failed, starting from scratch.\n"));
818 if (err != FTPRESTFAIL && !opt.server_response)
819 logputs (LOG_VERBOSE, _("done. "));
820 } /* restval && cmd & DO_RETR */
824 /* If we're in spider mode, don't really retrieve anything. The
825 fact that we got to this point should be proof enough that
826 the file exists, vaguely akin to HTTP's concept of a "HEAD"
833 fd_close (local_sock);
839 if (!opt.server_response)
842 logputs (LOG_VERBOSE, "\n");
843 logprintf (LOG_VERBOSE, "==> RETR %s ... ", u->file);
847 err = ftp_retr (csock, u->file);
848 /* FTPRERR, WRITEFAILED, FTPNSFOD */
852 logputs (LOG_VERBOSE, "\n");
853 logputs (LOG_NOTQUIET, _("\
854 Error in server response, closing control connection.\n"));
858 fd_close (local_sock);
862 logputs (LOG_VERBOSE, "\n");
863 logputs (LOG_NOTQUIET,
864 _("Write failed, closing control connection.\n"));
868 fd_close (local_sock);
872 logputs (LOG_VERBOSE, "\n");
873 logprintf (LOG_NOTQUIET, _("No such file `%s'.\n\n"), u->file);
875 fd_close (local_sock);
886 if (!opt.server_response)
887 logputs (LOG_VERBOSE, _("done.\n"));
888 expected_bytes = ftp_expected_bytes (ftp_last_respline);
893 if (!opt.server_response)
894 logputs (LOG_VERBOSE, "==> LIST ... ");
895 /* As Maciej W. Rozycki (macro@ds2.pg.gda.pl) says, `LIST'
896 without arguments is better than `LIST .'; confirmed by
898 err = ftp_list (csock, NULL);
899 /* FTPRERR, WRITEFAILED */
903 logputs (LOG_VERBOSE, "\n");
904 logputs (LOG_NOTQUIET, _("\
905 Error in server response, closing control connection.\n"));
909 fd_close (local_sock);
913 logputs (LOG_VERBOSE, "\n");
914 logputs (LOG_NOTQUIET,
915 _("Write failed, closing control connection.\n"));
919 fd_close (local_sock);
923 logputs (LOG_VERBOSE, "\n");
924 logprintf (LOG_NOTQUIET, _("No such file or directory `%s'.\n\n"),
927 fd_close (local_sock);
937 if (!opt.server_response)
938 logputs (LOG_VERBOSE, _("done.\n"));
939 expected_bytes = ftp_expected_bytes (ftp_last_respline);
940 } /* cmd & DO_LIST */
942 if (!(cmd & (DO_LIST | DO_RETR)) || (opt.spider && !(cmd & DO_LIST)))
945 /* Some FTP servers return the total length of file after REST
946 command, others just return the remaining size. */
947 if (*len && restval && expected_bytes
948 && (expected_bytes == *len - restval))
950 DEBUGP (("Lying FTP server found, adjusting.\n"));
951 expected_bytes = *len;
954 /* If no transmission was required, then everything is OK. */
955 if (!pasv_mode_open) /* we are not using pasive mode so we need
958 /* Wait for the server to connect to the address we're waiting
960 dtsock = accept_connection (local_sock);
963 logprintf (LOG_NOTQUIET, "accept: %s\n", strerror (errno));
968 /* Open the file -- if opt.dfp is set, use it instead. */
969 if (!opt.dfp || con->cmd & DO_LIST)
971 mkalldirs (con->target);
973 rotate_backups (con->target);
974 /* #### Is this correct? */
975 chmod (con->target, 0600);
977 fp = fopen (con->target, restval ? "ab" : "wb");
980 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target, strerror (errno));
984 fd_close (local_sock);
990 extern int global_download_count;
993 /* Rewind the output document if the download starts over and if
994 this is the first download. See gethttp() for a longer
996 if (!restval && global_download_count == 0 && opt.dfp != stdout)
998 /* This will silently fail for streams that don't correspond
999 to regular files, but that's OK. */
1001 /* ftruncate is needed because opt.dfp is opened in append
1002 mode if opt.always_rest is set. */
1003 ftruncate (fileno (fp), 0);
1010 logprintf (LOG_VERBOSE, _("Length: %s"), legible (*len));
1012 logprintf (LOG_VERBOSE, _(" [%s to go]"), legible (*len - restval));
1013 logputs (LOG_VERBOSE, "\n");
1014 expected_bytes = *len; /* for get_contents/show_progress */
1016 else if (expected_bytes)
1018 logprintf (LOG_VERBOSE, _("Length: %s"), legible (expected_bytes));
1020 logprintf (LOG_VERBOSE, _(" [%s to go]"),
1021 legible (expected_bytes - restval));
1022 logputs (LOG_VERBOSE, _(" (unauthoritative)\n"));
1025 /* Get the contents of the document. */
1026 res = fd_read_body (dtsock, fp, len, restval, expected_bytes, 0,
1028 tms = time_str (NULL);
1029 tmrate = retr_rate (*len - restval, con->dltime, 0);
1030 /* Close data connection socket. */
1032 fd_close (local_sock);
1033 /* Close the local file. */
1035 /* Close or flush the file. We have to be careful to check for
1036 error here. Checking the result of fwrite() is not enough --
1037 errors could go unnoticed! */
1039 if (!opt.dfp || con->cmd & DO_LIST)
1040 flush_res = fclose (fp);
1042 flush_res = fflush (fp);
1043 if (flush_res == EOF)
1047 /* If get_contents couldn't write to fp, bail out. */
1050 logprintf (LOG_NOTQUIET, _("%s: %s, closing control connection.\n"),
1051 con->target, strerror (errno));
1058 logprintf (LOG_NOTQUIET, _("%s (%s) - Data connection: %s; "),
1059 tms, tmrate, strerror (errno));
1060 if (opt.server_response)
1061 logputs (LOG_ALWAYS, "\n");
1064 /* Get the server to tell us if everything is retrieved. */
1065 err = ftp_response (csock, &respline);
1069 /* The control connection is decidedly closed. Print the time
1070 only if it hasn't already been printed. */
1072 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
1073 logputs (LOG_NOTQUIET, _("Control connection closed.\n"));
1074 /* If there is an error on the control connection, close it, but
1075 return FTPRETRINT, since there is a possibility that the
1076 whole file was retrieved nevertheless (but that is for
1077 ftp_loop_internal to decide). */
1081 } /* err != FTPOK */
1082 /* If retrieval failed for any reason, return FTPRETRINT, but do not
1083 close socket, since the control connection is still alive. If
1084 there is something wrong with the control connection, it will
1085 become apparent later. */
1086 if (*respline != '2')
1090 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
1091 logputs (LOG_NOTQUIET, _("Data transfer aborted.\n"));
1098 /* What now? The data connection was erroneous, whereas the
1099 response says everything is OK. We shall play it safe. */
1103 if (!(cmd & LEAVE_PENDING))
1105 /* I should probably send 'QUIT' and check for a reply, but this
1106 is faster. #### Is it OK, though? */
1110 /* If it was a listing, and opt.server_response is true,
1112 if (opt.server_response && (con->cmd & DO_LIST))
1114 mkalldirs (con->target);
1115 fp = fopen (con->target, "r");
1117 logprintf (LOG_ALWAYS, "%s: %s\n", con->target, strerror (errno));
1121 /* The lines are being read with read_whole_line because of
1122 no-buffering on opt.lfile. */
1123 while ((line = read_whole_line (fp)))
1125 logprintf (LOG_ALWAYS, "%s\n", line);
1130 } /* con->cmd & DO_LIST && server_response */
1132 return RETRFINISHED;
1135 /* A one-file FTP loop. This is the part where FTP retrieval is
1136 retried, and retried, and retried, and...
1138 This loop either gets commands from con, or (if ON_YOUR_OWN is
1139 set), makes them up to retrieve the file given by the URL. */
1141 ftp_loop_internal (struct url *u, struct fileinfo *f, ccon *con)
1146 char *tmrate = NULL;
1151 con->target = url_file_name (u);
1153 if (opt.noclobber && file_exists_p (con->target))
1155 logprintf (LOG_VERBOSE,
1156 _("File `%s' already there, not retrieving.\n"), con->target);
1157 /* If the file is there, we suppose it's retrieved OK. */
1161 /* Remove it if it's a link. */
1162 remove_link (con->target);
1163 if (!opt.output_document)
1166 locf = opt.output_document;
1170 if (con->st & ON_YOUR_OWN)
1171 con->st = ON_YOUR_OWN;
1173 orig_lp = con->cmd & LEAVE_PENDING ? 1 : 0;
1178 /* Increment the pass counter. */
1180 sleep_between_retrievals (count);
1181 if (con->st & ON_YOUR_OWN)
1184 con->cmd |= (DO_RETR | LEAVE_PENDING);
1185 if (con->csock != -1)
1186 con->cmd &= ~ (DO_LOGIN | DO_CWD);
1188 con->cmd |= (DO_LOGIN | DO_CWD);
1190 else /* not on your own */
1192 if (con->csock != -1)
1193 con->cmd &= ~DO_LOGIN;
1195 con->cmd |= DO_LOGIN;
1196 if (con->st & DONE_CWD)
1197 con->cmd &= ~DO_CWD;
1202 /* Assume no restarting. */
1204 if ((count > 1 || opt.always_rest)
1205 && !(con->cmd & DO_LIST)
1206 && file_exists_p (locf))
1207 if (stat (locf, &st) == 0 && S_ISREG (st.st_mode))
1208 restval = st.st_size;
1210 /* In `-c' is used, check whether the file we're writing to
1211 exists and is of non-zero length. If so, we'll refuse to
1212 truncate it if the server doesn't support continued
1214 if (opt.always_rest && restval > 0)
1215 con->cmd |= NO_TRUNCATE;
1217 /* Get the current time string. */
1218 tms = time_str (NULL);
1219 /* Print fetch message, if opt.verbose. */
1222 char *hurl = url_string (u, 1);
1226 sprintf (tmp, _("(try:%2d)"), count);
1227 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1228 tms, hurl, tmp, locf);
1230 ws_changetitle (hurl, 1);
1234 /* Send getftp the proper length, if fileinfo was provided. */
1239 err = getftp (u, &len, restval, con);
1241 if (con->csock != -1)
1242 con->st &= ~DONE_CWD;
1244 con->st |= DONE_CWD;
1248 case HOSTERR: case CONIMPOSSIBLE: case FWRITEERR: case FOPENERR:
1249 case FTPNSFOD: case FTPLOGINC: case FTPNOPASV: case CONTNOTSUPPORTED:
1250 /* Fatal errors, give up. */
1253 case CONSOCKERR: case CONERROR: case FTPSRVERR: case FTPRERR:
1254 case WRITEFAILED: case FTPUNKNOWNTYPE: case FTPSYSERR:
1255 case FTPPORTERR: case FTPLOGREFUSED: case FTPINVPASV:
1256 printwhat (count, opt.ntry);
1257 /* non-fatal errors */
1261 /* If the control connection was closed, the retrieval
1262 will be considered OK if f->size == len. */
1263 if (!f || len != f->size)
1265 printwhat (count, opt.ntry);
1277 tms = time_str (NULL);
1279 tmrate = retr_rate (len - restval, con->dltime, 0);
1281 /* If we get out of the switch above without continue'ing, we've
1282 successfully downloaded a file. Remember this fact. */
1283 downloaded_file (FILE_DOWNLOADED_NORMALLY, locf);
1285 if (con->st & ON_YOUR_OWN)
1287 fd_close (con->csock);
1291 logprintf (LOG_VERBOSE, _("%s (%s) - `%s' saved [%ld]\n\n"),
1292 tms, tmrate, locf, len);
1293 if (!opt.verbose && !opt.quiet)
1295 /* Need to hide the password from the URL. The `if' is here
1296 so that we don't do the needless allocation every
1298 char *hurl = url_string (u, 1);
1299 logprintf (LOG_NONVERBOSE, "%s URL: %s [%ld] -> \"%s\" [%d]\n",
1300 tms, hurl, len, locf, count);
1304 if ((con->cmd & DO_LIST))
1305 /* This is a directory listing file. */
1307 if (!opt.remove_listing)
1308 /* --dont-remove-listing was specified, so do count this towards the
1309 number of bytes and files downloaded. */
1311 total_downloaded_bytes += len;
1315 /* Deletion of listing files is not controlled by --delete-after, but
1316 by the more specific option --dont-remove-listing, and the code
1317 to do this deletion is in another function. */
1319 else if (!opt.spider)
1320 /* This is not a directory listing file. */
1322 /* Unlike directory listing files, don't pretend normal files weren't
1323 downloaded if they're going to be deleted. People seeding proxies,
1324 for instance, may want to know how many bytes and files they've
1325 downloaded through it. */
1326 total_downloaded_bytes += len;
1329 if (opt.delete_after)
1331 DEBUGP (("Removing file due to --delete-after in"
1332 " ftp_loop_internal():\n"));
1333 logprintf (LOG_VERBOSE, _("Removing %s.\n"), locf);
1335 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1339 /* Restore the original leave-pendingness. */
1341 con->cmd |= LEAVE_PENDING;
1343 con->cmd &= ~LEAVE_PENDING;
1345 } while (!opt.ntry || (count < opt.ntry));
1347 if (con->csock != -1 && (con->st & ON_YOUR_OWN))
1349 fd_close (con->csock);
1355 /* Return the directory listing in a reusable format. The directory
1356 is specifed in u->dir. */
1358 ftp_get_listing (struct url *u, ccon *con, struct fileinfo **f)
1361 char *uf; /* url file name */
1362 char *lf; /* list file name */
1363 char *old_target = con->target;
1365 con->st &= ~ON_YOUR_OWN;
1366 con->cmd |= (DO_LIST | LEAVE_PENDING);
1367 con->cmd &= ~DO_RETR;
1369 /* Find the listing file name. We do it by taking the file name of
1370 the URL and replacing the last component with the listing file
1372 uf = url_file_name (u);
1373 lf = file_merge (uf, LIST_FILENAME);
1375 DEBUGP ((_("Using `%s' as listing tmp file.\n"), lf));
1378 err = ftp_loop_internal (u, NULL, con);
1379 con->target = old_target;
1382 *f = ftp_parse_ls (lf, con->rs);
1385 if (opt.remove_listing)
1388 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1390 logprintf (LOG_VERBOSE, _("Removed `%s'.\n"), lf);
1393 con->cmd &= ~DO_LIST;
1397 static uerr_t ftp_retrieve_dirs PARAMS ((struct url *, struct fileinfo *,
1399 static uerr_t ftp_retrieve_glob PARAMS ((struct url *, ccon *, int));
1400 static struct fileinfo *delelement PARAMS ((struct fileinfo *,
1401 struct fileinfo **));
1402 static void freefileinfo PARAMS ((struct fileinfo *f));
1404 /* Retrieve a list of files given in struct fileinfo linked list. If
1405 a file is a symbolic link, do not retrieve it, but rather try to
1406 set up a similar link on the local disk, if the symlinks are
1409 If opt.recursive is set, after all files have been retrieved,
1410 ftp_retrieve_dirs will be called to retrieve the directories. */
1412 ftp_retrieve_list (struct url *u, struct fileinfo *f, ccon *con)
1414 static int depth = 0;
1416 struct fileinfo *orig;
1421 /* Increase the depth. */
1423 if (opt.reclevel != INFINITE_RECURSION && depth > opt.reclevel)
1425 DEBUGP ((_("Recursion depth %d exceeded max. depth %d.\n"),
1426 depth, opt.reclevel));
1434 con->st &= ~ON_YOUR_OWN;
1435 if (!(con->st & DONE_CWD))
1438 con->cmd &= ~DO_CWD;
1439 con->cmd |= (DO_RETR | LEAVE_PENDING);
1442 con->cmd |= DO_LOGIN;
1444 con->cmd &= ~DO_LOGIN;
1446 err = RETROK; /* in case it's not used */
1450 char *old_target, *ofile;
1452 if (opt.quota && total_downloaded_bytes > opt.quota)
1457 old_target = con->target;
1459 ofile = xstrdup (u->file);
1460 url_set_file (u, f->name);
1462 con->target = url_file_name (u);
1466 if (opt.timestamping && f->type == FT_PLAINFILE)
1469 /* If conversion of HTML files retrieved via FTP is ever implemented,
1470 we'll need to stat() <file>.orig here when -K has been specified.
1471 I'm not implementing it now since files on an FTP server are much
1472 more likely than files on an HTTP server to legitimately have a
1474 if (!stat (con->target, &st))
1478 /* Else, get it from the file. */
1479 local_size = st.st_size;
1482 /* Modification time granularity is 2 seconds for Windows, so
1483 increase local time by 1 second for later comparison. */
1486 /* Compare file sizes only for servers that tell us correct
1487 values. Assumme sizes being equal for servers that lie
1489 cor_val = (con->rs == ST_UNIX || con->rs == ST_WINNT);
1490 eq_size = cor_val ? (local_size == f->size) : 1 ;
1491 if (f->tstamp <= tml && eq_size)
1493 /* Remote file is older, file sizes can be compared and
1495 logprintf (LOG_VERBOSE, _("\
1496 Remote file no newer than local file `%s' -- not retrieving.\n"), con->target);
1501 /* Remote file is newer or sizes cannot be matched */
1502 logprintf (LOG_VERBOSE, _("\
1503 Remote file is newer than local file `%s' -- retrieving.\n\n"),
1508 /* Sizes do not match */
1509 logprintf (LOG_VERBOSE, _("\
1510 The sizes do not match (local %ld) -- retrieving.\n\n"), local_size);
1513 } /* opt.timestamping && f->type == FT_PLAINFILE */
1517 /* If opt.retr_symlinks is defined, we treat symlinks as
1518 if they were normal files. There is currently no way
1519 to distinguish whether they might be directories, and
1521 if (!opt.retr_symlinks)
1525 logputs (LOG_NOTQUIET,
1526 _("Invalid name of the symlink, skipping.\n"));
1530 /* Check whether we already have the correct
1532 int rc = lstat (con->target, &st);
1535 size_t len = strlen (f->linkto) + 1;
1536 if (S_ISLNK (st.st_mode))
1538 char *link_target = (char *)alloca (len);
1539 size_t n = readlink (con->target, link_target, len);
1541 && (memcmp (link_target, f->linkto, n) == 0))
1543 logprintf (LOG_VERBOSE, _("\
1544 Already have correct symlink %s -> %s\n\n"),
1545 con->target, f->linkto);
1551 logprintf (LOG_VERBOSE, _("Creating symlink %s -> %s\n"),
1552 con->target, f->linkto);
1553 /* Unlink before creating symlink! */
1554 unlink (con->target);
1555 if (symlink (f->linkto, con->target) == -1)
1556 logprintf (LOG_NOTQUIET, "symlink: %s\n",
1558 logputs (LOG_VERBOSE, "\n");
1559 } /* have f->linkto */
1560 #else /* not HAVE_SYMLINK */
1561 logprintf (LOG_NOTQUIET,
1562 _("Symlinks not supported, skipping symlink `%s'.\n"),
1564 #endif /* not HAVE_SYMLINK */
1566 else /* opt.retr_symlinks */
1569 err = ftp_loop_internal (u, f, con);
1570 } /* opt.retr_symlinks */
1574 logprintf (LOG_NOTQUIET, _("Skipping directory `%s'.\n"),
1578 /* Call the retrieve loop. */
1580 err = ftp_loop_internal (u, f, con);
1583 logprintf (LOG_NOTQUIET, _("%s: unknown/unsupported file type.\n"),
1588 /* Set the time-stamp information to the local file. Symlinks
1589 are not to be stamped because it sets the stamp on the
1591 if (!(f->type == FT_SYMLINK && !opt.retr_symlinks)
1594 && file_exists_p (con->target))
1596 /* #### This code repeats in http.c and ftp.c. Move it to a
1598 const char *fl = NULL;
1599 if (opt.output_document)
1601 if (opt.od_known_regular)
1602 fl = opt.output_document;
1607 touch (fl, f->tstamp);
1609 else if (f->tstamp == -1)
1610 logprintf (LOG_NOTQUIET, _("%s: corrupt time-stamp.\n"), con->target);
1612 if (f->perms && f->type == FT_PLAINFILE && dlthis)
1614 if (opt.preserve_perm)
1615 chmod (con->target, f->perms);
1618 DEBUGP (("Unrecognized permissions for %s.\n", con->target));
1620 xfree (con->target);
1621 con->target = old_target;
1623 url_set_file (u, ofile);
1626 /* Break on fatals. */
1627 if (err == QUOTEXC || err == HOSTERR || err == FWRITEERR)
1629 con->cmd &= ~ (DO_CWD | DO_LOGIN);
1633 /* We do not want to call ftp_retrieve_dirs here */
1634 if (opt.recursive &&
1635 !(opt.reclevel != INFINITE_RECURSION && depth >= opt.reclevel))
1636 err = ftp_retrieve_dirs (u, orig, con);
1637 else if (opt.recursive)
1638 DEBUGP ((_("Will not retrieve dirs since depth is %d (max %d).\n"),
1639 depth, opt.reclevel));
1644 /* Retrieve the directories given in a file list. This function works
1645 by simply going through the linked list and calling
1646 ftp_retrieve_glob on each directory entry. The function knows
1647 about excluded directories. */
1649 ftp_retrieve_dirs (struct url *u, struct fileinfo *f, ccon *con)
1651 char *container = NULL;
1652 int container_size = 0;
1654 for (; f; f = f->next)
1657 char *odir, *newdir;
1659 if (opt.quota && total_downloaded_bytes > opt.quota)
1661 if (f->type != FT_DIRECTORY)
1664 /* Allocate u->dir off stack, but reallocate only if a larger
1665 string is needed. It's a pity there's no "realloca" for an
1666 item on the bottom of the stack. */
1667 size = strlen (u->dir) + 1 + strlen (f->name) + 1;
1668 if (size > container_size)
1669 container = (char *)alloca (size);
1674 || (*odir == '/' && *(odir + 1) == '\0'))
1675 /* If ODIR is empty or just "/", simply append f->name to
1676 ODIR. (In the former case, to preserve u->dir being
1677 relative; in the latter case, to avoid double slash.) */
1678 sprintf (newdir, "%s%s", odir, f->name);
1680 /* Else, use a separator. */
1681 sprintf (newdir, "%s/%s", odir, f->name);
1683 DEBUGP (("Composing new CWD relative to the initial directory.\n"));
1684 DEBUGP ((" odir = '%s'\n f->name = '%s'\n newdir = '%s'\n\n",
1685 odir, f->name, newdir));
1686 if (!accdir (newdir, ALLABS))
1688 logprintf (LOG_VERBOSE, _("\
1689 Not descending to `%s' as it is excluded/not-included.\n"), newdir);
1693 con->st &= ~DONE_CWD;
1695 odir = xstrdup (u->dir); /* because url_set_dir will free
1697 url_set_dir (u, newdir);
1698 ftp_retrieve_glob (u, con, GETALL);
1699 url_set_dir (u, odir);
1702 /* Set the time-stamp? */
1705 if (opt.quota && total_downloaded_bytes > opt.quota)
1711 /* Return non-zero if S has a leading '/' or contains '../' */
1713 has_insecure_name_p (const char *s)
1718 if (strstr (s, "../") != 0)
1724 /* A near-top-level function to retrieve the files in a directory.
1725 The function calls ftp_get_listing, to get a linked list of files.
1726 Then it weeds out the file names that do not match the pattern.
1727 ftp_retrieve_list is called with this updated list as an argument.
1729 If the argument ACTION is GETONE, just download the file (but first
1730 get the listing, so that the time-stamp is heeded); if it's GLOBALL,
1731 use globbing; if it's GETALL, download the whole directory. */
1733 ftp_retrieve_glob (struct url *u, ccon *con, int action)
1735 struct fileinfo *f, *start;
1738 con->cmd |= LEAVE_PENDING;
1740 res = ftp_get_listing (u, con, &start);
1743 /* First: weed out that do not conform the global rules given in
1744 opt.accepts and opt.rejects. */
1745 if (opt.accepts || opt.rejects)
1750 if (f->type != FT_DIRECTORY && !acceptable (f->name))
1752 logprintf (LOG_VERBOSE, _("Rejecting `%s'.\n"), f->name);
1753 f = delelement (f, &start);
1759 /* Remove all files with possible harmful names */
1763 if (has_insecure_name_p (f->name))
1765 logprintf (LOG_VERBOSE, _("Rejecting `%s'.\n"), f->name);
1766 f = delelement (f, &start);
1771 /* Now weed out the files that do not match our globbing pattern.
1772 If we are dealing with a globbing pattern, that is. */
1773 if (*u->file && (action == GLOBALL || action == GETONE))
1780 matchres = fnmatch (u->file, f->name, 0);
1783 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target,
1787 if (matchres == FNM_NOMATCH)
1788 f = delelement (f, &start); /* delete the element from the list */
1790 f = f->next; /* leave the element in the list */
1794 freefileinfo (start);
1795 return RETRBADPATTERN;
1801 /* Just get everything. */
1802 ftp_retrieve_list (u, start, con);
1806 if (action == GLOBALL)
1809 /* #### This message SUCKS. We should see what was the
1810 reason that nothing was retrieved. */
1811 logprintf (LOG_VERBOSE, _("No matches on pattern `%s'.\n"), u->file);
1813 else /* GETONE or GETALL */
1815 /* Let's try retrieving it anyway. */
1816 con->st |= ON_YOUR_OWN;
1817 res = ftp_loop_internal (u, NULL, con);
1821 freefileinfo (start);
1822 if (opt.quota && total_downloaded_bytes > opt.quota)
1825 /* #### Should we return `res' here? */
1829 /* The wrapper that calls an appropriate routine according to contents
1830 of URL. Inherently, its capabilities are limited on what can be
1831 encoded into a URL. */
1833 ftp_loop (struct url *u, int *dt, struct url *proxy)
1835 ccon con; /* FTP connection */
1840 memset (&con, 0, sizeof (con));
1843 con.st = ON_YOUR_OWN;
1847 res = RETROK; /* in case it's not used */
1849 /* If the file name is empty, the user probably wants a directory
1850 index. We'll provide one, properly HTML-ized. Unless
1851 opt.htmlify is 0, of course. :-) */
1852 if (!*u->file && !opt.recursive)
1855 res = ftp_get_listing (u, &con, &f);
1859 if (opt.htmlify && !opt.spider)
1861 char *filename = (opt.output_document
1862 ? xstrdup (opt.output_document)
1863 : (con.target ? xstrdup (con.target)
1864 : url_file_name (u)));
1865 res = ftp_index (filename, u, f);
1866 if (res == FTPOK && opt.verbose)
1868 if (!opt.output_document)
1872 if (stat (filename, &st) == 0)
1876 logprintf (LOG_NOTQUIET,
1877 _("Wrote HTML-ized index to `%s' [%ld].\n"),
1881 logprintf (LOG_NOTQUIET,
1882 _("Wrote HTML-ized index to `%s'.\n"),
1892 int wild = has_wildcards_p (u->file);
1893 if ((opt.ftp_glob && wild) || opt.recursive || opt.timestamping)
1895 /* ftp_retrieve_glob is a catch-all function that gets called
1896 if we need globbing, time-stamping or recursion. Its
1897 third argument is just what we really need. */
1898 res = ftp_retrieve_glob (u, &con,
1899 (opt.ftp_glob && wild) ? GLOBALL : GETONE);
1902 res = ftp_loop_internal (u, NULL, &con);
1908 /* If a connection was left, quench it. */
1909 if (con.csock != -1)
1910 fd_close (con.csock);
1911 xfree_null (con.id);
1913 xfree_null (con.target);
1918 /* Delete an element from the fileinfo linked list. Returns the
1919 address of the next element, or NULL if the list is exhausted. It
1920 can modify the start of the list. */
1921 static struct fileinfo *
1922 delelement (struct fileinfo *f, struct fileinfo **start)
1924 struct fileinfo *prev = f->prev;
1925 struct fileinfo *next = f->next;
1928 xfree_null (f->linkto);
1940 /* Free the fileinfo linked list of files. */
1942 freefileinfo (struct fileinfo *f)
1946 struct fileinfo *next = f->next;