1 /* File Transfer Protocol support.
2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007, 2008 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 3 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, see <http://www.gnu.org/licenses/>.
20 Additional permission under GNU GPL version 3 section 7
22 If you modify this program, or any covered work, by linking or
23 combining it with the OpenSSL project's OpenSSL library (or a
24 modified version of that library), containing parts covered by the
25 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
26 grants you additional permission to convey the resulting work.
27 Corresponding Source for a non-source form of such a combination
28 shall include the source code for the parts of OpenSSL used as well
29 as that of the covered work. */
50 #include "convert.h" /* for downloaded_file */
51 #include "recur.h" /* for INFINITE_RECURSION */
53 /* File where the "ls -al" listing will be saved. */
55 #define LIST_FILENAME "_listing"
57 #define LIST_FILENAME ".listing"
62 int st; /* connection status */
63 int cmd; /* command code */
64 int csock; /* control connection socket */
65 double dltime; /* time of the download in msecs */
66 enum stype rs; /* remote system reported by ftp server */
67 char *id; /* initial directory */
68 char *target; /* target file name */
69 struct url *proxy; /* FTWK-style proxy */
73 /* Look for regexp "( *[0-9]+ *byte" (literal parenthesis) anywhere in
74 the string S, and return the number converted to wgint, if found, 0
77 ftp_expected_bytes (const char *s)
83 while (*s && *s != '(')
87 ++s; /* skip the '(' */
88 res = str_to_wgint (s, (char **) &s, 10);
91 while (*s && c_isspace (*s))
95 if (c_tolower (*s) != 'b')
97 if (strncasecmp (s, "byte", 4))
107 * This function sets up a passive data connection with the FTP server.
108 * It is merely a wrapper around ftp_epsv, ftp_lpsv and ftp_pasv.
111 ftp_do_pasv (int csock, ip_address *addr, int *port)
115 /* We need to determine the address family and need to call
116 getpeername, so while we're at it, store the address to ADDR.
117 ftp_pasv and ftp_lpsv can simply override it. */
118 if (!socket_ip_address (csock, addr, ENDPOINT_PEER))
121 /* If our control connection is over IPv6, then we first try EPSV and then
122 * LPSV if the former is not supported. If the control connection is over
123 * IPv4, we simply issue the good old PASV request. */
124 switch (addr->family)
127 if (!opt.server_response)
128 logputs (LOG_VERBOSE, "==> PASV ... ");
129 err = ftp_pasv (csock, addr, port);
132 if (!opt.server_response)
133 logputs (LOG_VERBOSE, "==> EPSV ... ");
134 err = ftp_epsv (csock, addr, port);
136 /* If EPSV is not supported try LPSV */
137 if (err == FTPNOPASV)
139 if (!opt.server_response)
140 logputs (LOG_VERBOSE, "==> LPSV ... ");
141 err = ftp_lpsv (csock, addr, port);
152 * This function sets up an active data connection with the FTP server.
153 * It is merely a wrapper around ftp_eprt, ftp_lprt and ftp_port.
156 ftp_do_port (int csock, int *local_sock)
161 if (!socket_ip_address (csock, &cip, ENDPOINT_PEER))
164 /* If our control connection is over IPv6, then we first try EPRT and then
165 * LPRT if the former is not supported. If the control connection is over
166 * IPv4, we simply issue the good old PORT request. */
170 if (!opt.server_response)
171 logputs (LOG_VERBOSE, "==> PORT ... ");
172 err = ftp_port (csock, local_sock);
175 if (!opt.server_response)
176 logputs (LOG_VERBOSE, "==> EPRT ... ");
177 err = ftp_eprt (csock, local_sock);
179 /* If EPRT is not supported try LPRT */
180 if (err == FTPPORTERR)
182 if (!opt.server_response)
183 logputs (LOG_VERBOSE, "==> LPRT ... ");
184 err = ftp_lprt (csock, local_sock);
195 ftp_do_pasv (int csock, ip_address *addr, int *port)
197 if (!opt.server_response)
198 logputs (LOG_VERBOSE, "==> PASV ... ");
199 return ftp_pasv (csock, addr, port);
203 ftp_do_port (int csock, int *local_sock)
205 if (!opt.server_response)
206 logputs (LOG_VERBOSE, "==> PORT ... ");
207 return ftp_port (csock, local_sock);
212 print_length (wgint size, wgint start, bool authoritative)
214 logprintf (LOG_VERBOSE, _("Length: %s"), number_to_static_string (size));
216 logprintf (LOG_VERBOSE, " (%s)", human_readable (size));
220 logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
221 number_to_static_string (size - start),
222 human_readable (size - start));
224 logprintf (LOG_VERBOSE, _(", %s remaining"),
225 number_to_static_string (size - start));
227 logputs (LOG_VERBOSE, !authoritative ? _(" (unauthoritative)\n") : "\n");
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, wgint *len, wgint restval, ccon *con)
236 int csock, dtsock, local_sock, res;
237 uerr_t err = RETROK; /* appease the compiler */
239 char *user, *passwd, *respline;
243 bool pasv_mode_open = false;
244 wgint expected_bytes = 0;
245 bool rest_failed = false;
249 assert (con != NULL);
250 assert (con->target != NULL);
252 /* Debug-check of the sanity of the request by making sure that LIST
253 and RETR are never both requested (since we can handle only one
255 assert (!((cmd & DO_LIST) && (cmd & DO_RETR)));
256 /* Make sure that at least *something* is requested. */
257 assert ((cmd & (DO_LIST | DO_CWD | DO_RETR | DO_LOGIN)) != 0);
261 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 1);
262 user = user ? user : (opt.ftp_user ? opt.ftp_user : opt.user);
263 if (!user) user = "anonymous";
264 passwd = passwd ? passwd : (opt.ftp_passwd ? opt.ftp_passwd : opt.passwd);
265 if (!passwd) passwd = "-wget@";
271 if (!(cmd & DO_LOGIN))
273 else /* cmd & DO_LOGIN */
276 char *host = con->proxy ? con->proxy->host : u->host;
277 int port = con->proxy ? con->proxy->port : u->port;
278 char *logname = user;
282 /* If proxy is in use, log in as username@target-site. */
283 logname = concat_strings (user, "@", u->host, (char *) 0);
286 /* Login to the server: */
288 /* First: Establish the control connection. */
290 csock = connect_to_host (host, port);
294 return (retryable_socket_connect_error (errno)
295 ? CONERROR : CONIMPOSSIBLE);
297 if (cmd & LEAVE_PENDING)
302 /* Second: Login with proper USER/PASS sequence. */
303 logprintf (LOG_VERBOSE, _("Logging in as %s ... "), escnonprint (user));
304 if (opt.server_response)
305 logputs (LOG_ALWAYS, "\n");
306 err = ftp_login (csock, logname, passwd);
311 /* FTPRERR, FTPSRVERR, WRITEFAILED, FTPLOGREFUSED, FTPLOGINC */
315 logputs (LOG_VERBOSE, "\n");
316 logputs (LOG_NOTQUIET, _("\
317 Error in server response, closing control connection.\n"));
322 logputs (LOG_VERBOSE, "\n");
323 logputs (LOG_NOTQUIET, _("Error in server greeting.\n"));
328 logputs (LOG_VERBOSE, "\n");
329 logputs (LOG_NOTQUIET,
330 _("Write failed, closing control connection.\n"));
335 logputs (LOG_VERBOSE, "\n");
336 logputs (LOG_NOTQUIET, _("The server refuses login.\n"));
339 return FTPLOGREFUSED;
341 logputs (LOG_VERBOSE, "\n");
342 logputs (LOG_NOTQUIET, _("Login incorrect.\n"));
347 if (!opt.server_response)
348 logputs (LOG_VERBOSE, _("Logged in!\n"));
353 /* Third: Get the system type */
354 if (!opt.server_response)
355 logprintf (LOG_VERBOSE, "==> SYST ... ");
356 err = ftp_syst (csock, &con->rs);
361 logputs (LOG_VERBOSE, "\n");
362 logputs (LOG_NOTQUIET, _("\
363 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. */
378 if (!opt.server_response && err != FTPSRVERR)
379 logputs (LOG_VERBOSE, _("done. "));
381 /* Fourth: Find the initial ftp directory */
383 if (!opt.server_response)
384 logprintf (LOG_VERBOSE, "==> PWD ... ");
385 err = ftp_pwd (csock, &con->id);
390 logputs (LOG_VERBOSE, "\n");
391 logputs (LOG_NOTQUIET, _("\
392 Error in server response, closing control connection.\n"));
397 /* PWD unsupported -- assume "/". */
398 xfree_null (con->id);
399 con->id = xstrdup ("/");
402 /* Everything is OK. */
407 /* VMS will report something like "PUB$DEVICE:[INITIAL.FOLDER]".
408 Convert it to "/INITIAL/FOLDER" */
409 if (con->rs == ST_VMS)
411 char *path = strchr (con->id, '[');
412 char *pathend = path ? strchr (path + 1, ']') : NULL;
413 if (!path || !pathend)
414 DEBUGP (("Initial VMS directory not in the form [...]!\n"));
417 char *idir = con->id;
418 DEBUGP (("Preprocessing the initial VMS directory\n"));
419 DEBUGP ((" old = '%s'\n", con->id));
420 /* We do the conversion in-place by copying the stuff
421 between [ and ] to the beginning, and changing dots
422 to slashes at the same time. */
424 for (++path; path < pathend; path++, idir++)
425 *idir = *path == '.' ? '/' : *path;
427 DEBUGP ((" new = '%s'\n\n", con->id));
430 if (!opt.server_response)
431 logputs (LOG_VERBOSE, _("done.\n"));
433 /* Fifth: Set the FTP type. */
434 type_char = ftp_process_type (u->params);
435 if (!opt.server_response)
436 logprintf (LOG_VERBOSE, "==> TYPE %c ... ", type_char);
437 err = ftp_type (csock, type_char);
438 /* FTPRERR, WRITEFAILED, FTPUNKNOWNTYPE */
442 logputs (LOG_VERBOSE, "\n");
443 logputs (LOG_NOTQUIET, _("\
444 Error in server response, closing control connection.\n"));
449 logputs (LOG_VERBOSE, "\n");
450 logputs (LOG_NOTQUIET,
451 _("Write failed, closing control connection.\n"));
456 logputs (LOG_VERBOSE, "\n");
457 logprintf (LOG_NOTQUIET,
458 _("Unknown type `%c', closing control connection.\n"),
464 /* Everything is OK. */
469 if (!opt.server_response)
470 logputs (LOG_VERBOSE, _("done. "));
476 logputs (LOG_VERBOSE, _("==> CWD not needed.\n"));
479 char *target = u->dir;
481 DEBUGP (("changing working directory\n"));
483 /* Change working directory. To change to a non-absolute
484 Unix directory, we need to prepend initial directory
485 (con->id) to it. Absolute directories "just work".
487 A relative directory is one that does not begin with '/'
488 and, on non-Unix OS'es, one that doesn't begin with
491 This is not done for OS400, which doesn't use
492 "/"-delimited directories, nor does it support directory
493 hierarchies. "CWD foo" followed by "CWD bar" leaves us
494 in "bar", not in "foo/bar", as would be customary
498 && !(con->rs != ST_UNIX
499 && c_isalpha (target[0])
501 && con->rs != ST_OS400)
503 int idlen = strlen (con->id);
506 /* Strip trailing slash(es) from con->id. */
507 while (idlen > 0 && con->id[idlen - 1] == '/')
509 p = ntarget = (char *)alloca (idlen + 1 + strlen (u->dir) + 1);
510 memcpy (p, con->id, idlen);
515 DEBUGP (("Prepended initial PWD to relative path:\n"));
516 DEBUGP ((" pwd: '%s'\n old: '%s'\n new: '%s'\n",
517 con->id, target, ntarget));
521 /* If the FTP host runs VMS, we will have to convert the absolute
522 directory path in UNIX notation to absolute directory path in
523 VMS notation as VMS FTP servers do not like UNIX notation of
524 absolute paths. "VMS notation" is [dir.subdir.subsubdir]. */
526 if (con->rs == ST_VMS)
529 char *ntarget = (char *)alloca (strlen (target) + 2);
530 /* We use a converted initial dir, so directories in
531 TARGET will be separated with slashes, something like
532 "/INITIAL/FOLDER/DIR/SUBDIR". Convert that to
533 "[INITIAL.FOLDER.DIR.SUBDIR]". */
534 strcpy (ntarget, target);
535 assert (*ntarget == '/');
537 for (tmpp = ntarget + 1; *tmpp; tmpp++)
542 DEBUGP (("Changed file name to VMS syntax:\n"));
543 DEBUGP ((" Unix: '%s'\n VMS: '%s'\n", target, ntarget));
547 if (!opt.server_response)
548 logprintf (LOG_VERBOSE, "==> CWD %s ... ", escnonprint (target));
549 err = ftp_cwd (csock, target);
550 /* FTPRERR, WRITEFAILED, FTPNSFOD */
554 logputs (LOG_VERBOSE, "\n");
555 logputs (LOG_NOTQUIET, _("\
556 Error in server response, closing control connection.\n"));
561 logputs (LOG_VERBOSE, "\n");
562 logputs (LOG_NOTQUIET,
563 _("Write failed, closing control connection.\n"));
568 logputs (LOG_VERBOSE, "\n");
569 logprintf (LOG_NOTQUIET, _("No such directory %s.\n\n"),
570 quote (escnonprint (u->dir)));
579 if (!opt.server_response)
580 logputs (LOG_VERBOSE, _("done.\n"));
583 else /* do not CWD */
584 logputs (LOG_VERBOSE, _("==> CWD not required.\n"));
586 if ((cmd & DO_RETR) && *len == 0)
590 if (!opt.server_response)
591 logprintf (LOG_VERBOSE, "==> SIZE %s ... ", escnonprint (u->file));
594 err = ftp_size (csock, u->file, len);
600 logputs (LOG_VERBOSE, "\n");
601 logputs (LOG_NOTQUIET, _("\
602 Error in server response, closing control connection.\n"));
607 /* Everything is OK. */
612 if (!opt.server_response)
613 logprintf (LOG_VERBOSE, *len ? "%s\n" : _("done.\n"),
614 number_to_static_string (*len));
617 /* If anything is to be retrieved, PORT (or PASV) must be sent. */
618 if (cmd & (DO_LIST | DO_RETR))
622 ip_address passive_addr;
624 err = ftp_do_pasv (csock, &passive_addr, &passive_port);
625 /* FTPRERR, WRITEFAILED, FTPNOPASV, FTPINVPASV */
629 logputs (LOG_VERBOSE, "\n");
630 logputs (LOG_NOTQUIET, _("\
631 Error in server response, closing control connection.\n"));
636 logputs (LOG_VERBOSE, "\n");
637 logputs (LOG_NOTQUIET,
638 _("Write failed, closing control connection.\n"));
643 logputs (LOG_VERBOSE, "\n");
644 logputs (LOG_NOTQUIET, _("Cannot initiate PASV transfer.\n"));
647 logputs (LOG_VERBOSE, "\n");
648 logputs (LOG_NOTQUIET, _("Cannot parse PASV response.\n"));
657 DEBUGP (("trying to connect to %s port %d\n",
658 print_address (&passive_addr), passive_port));
659 dtsock = connect_to_ip (&passive_addr, passive_port, NULL);
662 int save_errno = errno;
665 logprintf (LOG_VERBOSE, _("couldn't connect to %s port %d: %s\n"),
666 print_address (&passive_addr), passive_port,
667 strerror (save_errno));
668 return (retryable_socket_connect_error (save_errno)
669 ? CONERROR : CONIMPOSSIBLE);
672 pasv_mode_open = true; /* Flag to avoid accept port */
673 if (!opt.server_response)
674 logputs (LOG_VERBOSE, _("done. "));
678 if (!pasv_mode_open) /* Try to use a port command if PASV failed */
680 err = ftp_do_port (csock, &local_sock);
681 /* FTPRERR, WRITEFAILED, bindport (FTPSYSERR), HOSTERR,
686 logputs (LOG_VERBOSE, "\n");
687 logputs (LOG_NOTQUIET, _("\
688 Error in server response, closing control connection.\n"));
692 fd_close (local_sock);
695 logputs (LOG_VERBOSE, "\n");
696 logputs (LOG_NOTQUIET,
697 _("Write failed, closing control connection.\n"));
701 fd_close (local_sock);
704 logputs (LOG_VERBOSE, "\n");
705 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
709 fd_close (local_sock);
712 logputs (LOG_VERBOSE, "\n");
713 logprintf (LOG_NOTQUIET, _("Bind error (%s).\n"),
718 logputs (LOG_VERBOSE, "\n");
719 logputs (LOG_NOTQUIET, _("Invalid PORT.\n"));
723 fd_close (local_sock);
730 if (!opt.server_response)
731 logputs (LOG_VERBOSE, _("done. "));
733 } /* cmd & (DO_LIST | DO_RETR) */
735 /* Restart if needed. */
736 if (restval && (cmd & DO_RETR))
738 if (!opt.server_response)
739 logprintf (LOG_VERBOSE, "==> REST %s ... ",
740 number_to_static_string (restval));
741 err = ftp_rest (csock, restval);
743 /* FTPRERR, WRITEFAILED, FTPRESTFAIL */
747 logputs (LOG_VERBOSE, "\n");
748 logputs (LOG_NOTQUIET, _("\
749 Error in server response, closing control connection.\n"));
753 fd_close (local_sock);
756 logputs (LOG_VERBOSE, "\n");
757 logputs (LOG_NOTQUIET,
758 _("Write failed, closing control connection.\n"));
762 fd_close (local_sock);
765 logputs (LOG_VERBOSE, _("\nREST failed, starting from scratch.\n"));
773 if (err != FTPRESTFAIL && !opt.server_response)
774 logputs (LOG_VERBOSE, _("done. "));
775 } /* restval && cmd & DO_RETR */
779 /* If we're in spider mode, don't really retrieve anything. The
780 fact that we got to this point should be proof enough that
781 the file exists, vaguely akin to HTTP's concept of a "HEAD"
788 fd_close (local_sock);
794 if (!opt.server_response)
797 logputs (LOG_VERBOSE, "\n");
798 logprintf (LOG_VERBOSE, "==> RETR %s ... ", escnonprint (u->file));
802 err = ftp_retr (csock, u->file);
803 /* FTPRERR, WRITEFAILED, FTPNSFOD */
807 logputs (LOG_VERBOSE, "\n");
808 logputs (LOG_NOTQUIET, _("\
809 Error in server response, closing control connection.\n"));
813 fd_close (local_sock);
816 logputs (LOG_VERBOSE, "\n");
817 logputs (LOG_NOTQUIET,
818 _("Write failed, closing control connection.\n"));
822 fd_close (local_sock);
825 logputs (LOG_VERBOSE, "\n");
826 logprintf (LOG_NOTQUIET, _("No such file %s.\n\n"),
827 quote (escnonprint (u->file)));
829 fd_close (local_sock);
837 if (!opt.server_response)
838 logputs (LOG_VERBOSE, _("done.\n"));
839 expected_bytes = ftp_expected_bytes (ftp_last_respline);
844 if (!opt.server_response)
845 logputs (LOG_VERBOSE, "==> LIST ... ");
846 /* As Maciej W. Rozycki (macro@ds2.pg.gda.pl) says, `LIST'
847 without arguments is better than `LIST .'; confirmed by
849 err = ftp_list (csock, NULL);
850 /* FTPRERR, WRITEFAILED */
854 logputs (LOG_VERBOSE, "\n");
855 logputs (LOG_NOTQUIET, _("\
856 Error in server response, closing control connection.\n"));
860 fd_close (local_sock);
863 logputs (LOG_VERBOSE, "\n");
864 logputs (LOG_NOTQUIET,
865 _("Write failed, closing control connection.\n"));
869 fd_close (local_sock);
872 logputs (LOG_VERBOSE, "\n");
873 logprintf (LOG_NOTQUIET, _("No such file or directory %s.\n\n"),
876 fd_close (local_sock);
883 if (!opt.server_response)
884 logputs (LOG_VERBOSE, _("done.\n"));
885 expected_bytes = ftp_expected_bytes (ftp_last_respline);
886 } /* cmd & DO_LIST */
888 if (!(cmd & (DO_LIST | DO_RETR)) || (opt.spider && !(cmd & DO_LIST)))
891 /* Some FTP servers return the total length of file after REST
892 command, others just return the remaining size. */
893 if (*len && restval && expected_bytes
894 && (expected_bytes == *len - restval))
896 DEBUGP (("Lying FTP server found, adjusting.\n"));
897 expected_bytes = *len;
900 /* If no transmission was required, then everything is OK. */
901 if (!pasv_mode_open) /* we are not using pasive mode so we need
904 /* Wait for the server to connect to the address we're waiting
906 dtsock = accept_connection (local_sock);
909 logprintf (LOG_NOTQUIET, "accept: %s\n", strerror (errno));
914 /* Open the file -- if output_stream is set, use it instead. */
915 if (!output_stream || con->cmd & DO_LIST)
917 mkalldirs (con->target);
919 rotate_backups (con->target);
921 if (restval && !(con->cmd & DO_LIST))
922 fp = fopen (con->target, "ab");
923 else if (opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct
924 || opt.output_document)
925 fp = fopen (con->target, "wb");
928 fp = fopen_excl (con->target, true);
929 if (!fp && errno == EEXIST)
931 /* We cannot just invent a new name and use it (which is
932 what functions like unique_create typically do)
933 because we told the user we'd use this name.
934 Instead, return and retry the download. */
935 logprintf (LOG_NOTQUIET, _("%s has sprung into existence.\n"),
940 fd_close (local_sock);
941 return FOPEN_EXCL_ERR;
946 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target, strerror (errno));
950 fd_close (local_sock);
959 print_length (*len, restval, true);
960 expected_bytes = *len; /* for fd_read_body's progress bar */
962 else if (expected_bytes)
963 print_length (expected_bytes, restval, false);
965 /* Get the contents of the document. */
967 if (restval && rest_failed)
968 flags |= rb_skip_startpos;
971 res = fd_read_body (dtsock, fp,
972 expected_bytes ? expected_bytes - restval : 0,
973 restval, &rd_size, len, &con->dltime, flags);
975 tms = datetime_str (time (NULL));
976 tmrate = retr_rate (rd_size, con->dltime);
977 total_download_time += con->dltime;
979 fd_close (local_sock);
980 /* Close the local file. */
981 if (!output_stream || con->cmd & DO_LIST)
984 /* If fd_read_body couldn't write to fp, bail out. */
987 logprintf (LOG_NOTQUIET, _("%s: %s, closing control connection.\n"),
988 con->target, strerror (errno));
996 logprintf (LOG_NOTQUIET, _("%s (%s) - Data connection: %s; "),
997 tms, tmrate, fd_errstr (dtsock));
998 if (opt.server_response)
999 logputs (LOG_ALWAYS, "\n");
1003 /* Get the server to tell us if everything is retrieved. */
1004 err = ftp_response (csock, &respline);
1007 /* The control connection is decidedly closed. Print the time
1008 only if it hasn't already been printed. */
1010 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
1011 logputs (LOG_NOTQUIET, _("Control connection closed.\n"));
1012 /* If there is an error on the control connection, close it, but
1013 return FTPRETRINT, since there is a possibility that the
1014 whole file was retrieved nevertheless (but that is for
1015 ftp_loop_internal to decide). */
1019 } /* err != FTPOK */
1020 /* If retrieval failed for any reason, return FTPRETRINT, but do not
1021 close socket, since the control connection is still alive. If
1022 there is something wrong with the control connection, it will
1023 become apparent later. */
1024 if (*respline != '2')
1028 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
1029 logputs (LOG_NOTQUIET, _("Data transfer aborted.\n"));
1036 /* What now? The data connection was erroneous, whereas the
1037 response says everything is OK. We shall play it safe. */
1041 if (!(cmd & LEAVE_PENDING))
1043 /* Closing the socket is faster than sending 'QUIT' and the
1044 effect is the same. */
1048 /* If it was a listing, and opt.server_response is true,
1050 if (opt.server_response && (con->cmd & DO_LIST))
1052 mkalldirs (con->target);
1053 fp = fopen (con->target, "r");
1055 logprintf (LOG_ALWAYS, "%s: %s\n", con->target, strerror (errno));
1059 /* The lines are being read with read_whole_line because of
1060 no-buffering on opt.lfile. */
1061 while ((line = read_whole_line (fp)) != NULL)
1063 char *p = strchr (line, '\0');
1064 while (p > line && (p[-1] == '\n' || p[-1] == '\r'))
1066 logprintf (LOG_ALWAYS, "%s\n", escnonprint (line));
1071 } /* con->cmd & DO_LIST && server_response */
1073 return RETRFINISHED;
1076 /* A one-file FTP loop. This is the part where FTP retrieval is
1077 retried, and retried, and retried, and...
1079 This loop either gets commands from con, or (if ON_YOUR_OWN is
1080 set), makes them up to retrieve the file given by the URL. */
1082 ftp_loop_internal (struct url *u, struct fileinfo *f, ccon *con)
1085 wgint restval, len = 0;
1087 const char *tmrate = NULL;
1092 con->target = url_file_name (u);
1094 if (opt.noclobber && file_exists_p (con->target))
1096 logprintf (LOG_VERBOSE,
1097 _("File %s already there; not retrieving.\n"), quote (con->target));
1098 /* If the file is there, we suppose it's retrieved OK. */
1102 /* Remove it if it's a link. */
1103 remove_link (con->target);
1104 if (!opt.output_document)
1107 locf = opt.output_document;
1111 if (con->st & ON_YOUR_OWN)
1112 con->st = ON_YOUR_OWN;
1114 orig_lp = con->cmd & LEAVE_PENDING ? 1 : 0;
1119 /* Increment the pass counter. */
1121 sleep_between_retrievals (count);
1122 if (con->st & ON_YOUR_OWN)
1125 con->cmd |= (DO_RETR | LEAVE_PENDING);
1126 if (con->csock != -1)
1127 con->cmd &= ~ (DO_LOGIN | DO_CWD);
1129 con->cmd |= (DO_LOGIN | DO_CWD);
1131 else /* not on your own */
1133 if (con->csock != -1)
1134 con->cmd &= ~DO_LOGIN;
1136 con->cmd |= DO_LOGIN;
1137 if (con->st & DONE_CWD)
1138 con->cmd &= ~DO_CWD;
1143 /* Decide whether or not to restart. */
1144 if (con->cmd & DO_LIST)
1146 else if (opt.always_rest
1147 && stat (locf, &st) == 0
1148 && S_ISREG (st.st_mode))
1149 /* When -c is used, continue from on-disk size. (Can't use
1150 hstat.len even if count>1 because we don't want a failed
1151 first attempt to clobber existing data.) */
1152 restval = st.st_size;
1154 restval = len; /* start where the previous run left off */
1158 /* Get the current time string. */
1159 tms = datetime_str (time (NULL));
1160 /* Print fetch message, if opt.verbose. */
1163 char *hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
1167 sprintf (tmp, _("(try:%2d)"), count);
1168 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => %s\n",
1169 tms, hurl, tmp, quote (locf));
1171 ws_changetitle (hurl);
1175 /* Send getftp the proper length, if fileinfo was provided. */
1180 err = getftp (u, &len, restval, con);
1182 if (con->csock == -1)
1183 con->st &= ~DONE_CWD;
1185 con->st |= DONE_CWD;
1189 case HOSTERR: case CONIMPOSSIBLE: case FWRITEERR: case FOPENERR:
1190 case FTPNSFOD: case FTPLOGINC: case FTPNOPASV: case CONTNOTSUPPORTED:
1191 /* Fatal errors, give up. */
1193 case CONSOCKERR: case CONERROR: case FTPSRVERR: case FTPRERR:
1194 case WRITEFAILED: case FTPUNKNOWNTYPE: case FTPSYSERR:
1195 case FTPPORTERR: case FTPLOGREFUSED: case FTPINVPASV:
1196 case FOPEN_EXCL_ERR:
1197 printwhat (count, opt.ntry);
1198 /* non-fatal errors */
1199 if (err == FOPEN_EXCL_ERR)
1201 /* Re-determine the file name. */
1202 xfree_null (con->target);
1203 con->target = url_file_name (u);
1208 /* If the control connection was closed, the retrieval
1209 will be considered OK if f->size == len. */
1210 if (!f || len != f->size)
1212 printwhat (count, opt.ntry);
1223 tms = datetime_str (time (NULL));
1225 tmrate = retr_rate (len - restval, con->dltime);
1227 /* If we get out of the switch above without continue'ing, we've
1228 successfully downloaded a file. Remember this fact. */
1229 downloaded_file (FILE_DOWNLOADED_NORMALLY, locf);
1231 if (con->st & ON_YOUR_OWN)
1233 fd_close (con->csock);
1237 logprintf (LOG_VERBOSE, _("%s (%s) - %s saved [%s]\n\n"),
1238 tms, tmrate, quote (locf), number_to_static_string (len));
1239 if (!opt.verbose && !opt.quiet)
1241 /* Need to hide the password from the URL. The `if' is here
1242 so that we don't do the needless allocation every
1244 char *hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
1245 logprintf (LOG_NONVERBOSE, "%s URL: %s [%s] -> \"%s\" [%d]\n",
1246 tms, hurl, number_to_static_string (len), locf, count);
1250 if ((con->cmd & DO_LIST))
1251 /* This is a directory listing file. */
1253 if (!opt.remove_listing)
1254 /* --dont-remove-listing was specified, so do count this towards the
1255 number of bytes and files downloaded. */
1257 total_downloaded_bytes += len;
1261 /* Deletion of listing files is not controlled by --delete-after, but
1262 by the more specific option --dont-remove-listing, and the code
1263 to do this deletion is in another function. */
1265 else if (!opt.spider)
1266 /* This is not a directory listing file. */
1268 /* Unlike directory listing files, don't pretend normal files weren't
1269 downloaded if they're going to be deleted. People seeding proxies,
1270 for instance, may want to know how many bytes and files they've
1271 downloaded through it. */
1272 total_downloaded_bytes += len;
1275 if (opt.delete_after)
1278 Removing file due to --delete-after in ftp_loop_internal():\n"));
1279 logprintf (LOG_VERBOSE, _("Removing %s.\n"), locf);
1281 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1285 /* Restore the original leave-pendingness. */
1287 con->cmd |= LEAVE_PENDING;
1289 con->cmd &= ~LEAVE_PENDING;
1291 } while (!opt.ntry || (count < opt.ntry));
1293 if (con->csock != -1 && (con->st & ON_YOUR_OWN))
1295 fd_close (con->csock);
1301 /* Return the directory listing in a reusable format. The directory
1302 is specifed in u->dir. */
1304 ftp_get_listing (struct url *u, ccon *con, struct fileinfo **f)
1307 char *uf; /* url file name */
1308 char *lf; /* list file name */
1309 char *old_target = con->target;
1311 con->st &= ~ON_YOUR_OWN;
1312 con->cmd |= (DO_LIST | LEAVE_PENDING);
1313 con->cmd &= ~DO_RETR;
1315 /* Find the listing file name. We do it by taking the file name of
1316 the URL and replacing the last component with the listing file
1318 uf = url_file_name (u);
1319 lf = file_merge (uf, LIST_FILENAME);
1321 DEBUGP ((_("Using %s as listing tmp file.\n"), quote (lf)));
1324 err = ftp_loop_internal (u, NULL, con);
1325 con->target = old_target;
1328 *f = ftp_parse_ls (lf, con->rs);
1331 if (opt.remove_listing)
1334 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1336 logprintf (LOG_VERBOSE, _("Removed %s.\n"), quote (lf));
1339 con->cmd &= ~DO_LIST;
1343 static uerr_t ftp_retrieve_dirs (struct url *, struct fileinfo *, ccon *);
1344 static uerr_t ftp_retrieve_glob (struct url *, ccon *, int);
1345 static struct fileinfo *delelement (struct fileinfo *, struct fileinfo **);
1346 static void freefileinfo (struct fileinfo *f);
1348 /* Retrieve a list of files given in struct fileinfo linked list. If
1349 a file is a symbolic link, do not retrieve it, but rather try to
1350 set up a similar link on the local disk, if the symlinks are
1353 If opt.recursive is set, after all files have been retrieved,
1354 ftp_retrieve_dirs will be called to retrieve the directories. */
1356 ftp_retrieve_list (struct url *u, struct fileinfo *f, ccon *con)
1358 static int depth = 0;
1360 struct fileinfo *orig;
1365 /* Increase the depth. */
1367 if (opt.reclevel != INFINITE_RECURSION && depth > opt.reclevel)
1369 DEBUGP ((_("Recursion depth %d exceeded max. depth %d.\n"),
1370 depth, opt.reclevel));
1378 con->st &= ~ON_YOUR_OWN;
1379 if (!(con->st & DONE_CWD))
1382 con->cmd &= ~DO_CWD;
1383 con->cmd |= (DO_RETR | LEAVE_PENDING);
1386 con->cmd |= DO_LOGIN;
1388 con->cmd &= ~DO_LOGIN;
1390 err = RETROK; /* in case it's not used */
1394 char *old_target, *ofile;
1396 if (opt.quota && total_downloaded_bytes > opt.quota)
1401 old_target = con->target;
1403 ofile = xstrdup (u->file);
1404 url_set_file (u, f->name);
1406 con->target = url_file_name (u);
1410 if (opt.timestamping && f->type == FT_PLAINFILE)
1413 /* If conversion of HTML files retrieved via FTP is ever implemented,
1414 we'll need to stat() <file>.orig here when -K has been specified.
1415 I'm not implementing it now since files on an FTP server are much
1416 more likely than files on an HTTP server to legitimately have a
1418 if (!stat (con->target, &st))
1422 /* Else, get it from the file. */
1423 local_size = st.st_size;
1426 /* Modification time granularity is 2 seconds for Windows, so
1427 increase local time by 1 second for later comparison. */
1430 /* Compare file sizes only for servers that tell us correct
1431 values. Assume sizes being equal for servers that lie
1433 cor_val = (con->rs == ST_UNIX || con->rs == ST_WINNT);
1434 eq_size = cor_val ? (local_size == f->size) : true;
1435 if (f->tstamp <= tml && eq_size)
1437 /* Remote file is older, file sizes can be compared and
1439 logprintf (LOG_VERBOSE, _("\
1440 Remote file no newer than local file %s -- not retrieving.\n"), quote (con->target));
1445 /* Remote file is newer or sizes cannot be matched */
1446 logprintf (LOG_VERBOSE, _("\
1447 Remote file is newer than local file %s -- retrieving.\n\n"),
1448 quote (con->target));
1452 /* Sizes do not match */
1453 logprintf (LOG_VERBOSE, _("\
1454 The sizes do not match (local %s) -- retrieving.\n\n"),
1455 number_to_static_string (local_size));
1458 } /* opt.timestamping && f->type == FT_PLAINFILE */
1462 /* If opt.retr_symlinks is defined, we treat symlinks as
1463 if they were normal files. There is currently no way
1464 to distinguish whether they might be directories, and
1466 if (!opt.retr_symlinks)
1470 logputs (LOG_NOTQUIET,
1471 _("Invalid name of the symlink, skipping.\n"));
1475 /* Check whether we already have the correct
1477 int rc = lstat (con->target, &st);
1480 size_t len = strlen (f->linkto) + 1;
1481 if (S_ISLNK (st.st_mode))
1483 char *link_target = (char *)alloca (len);
1484 size_t n = readlink (con->target, link_target, len);
1486 && (memcmp (link_target, f->linkto, n) == 0))
1488 logprintf (LOG_VERBOSE, _("\
1489 Already have correct symlink %s -> %s\n\n"),
1490 con->target, escnonprint (f->linkto));
1496 logprintf (LOG_VERBOSE, _("Creating symlink %s -> %s\n"),
1497 con->target, escnonprint (f->linkto));
1498 /* Unlink before creating symlink! */
1499 unlink (con->target);
1500 if (symlink (f->linkto, con->target) == -1)
1501 logprintf (LOG_NOTQUIET, "symlink: %s\n", strerror (errno));
1502 logputs (LOG_VERBOSE, "\n");
1503 } /* have f->linkto */
1504 #else /* not HAVE_SYMLINK */
1505 logprintf (LOG_NOTQUIET,
1506 _("Symlinks not supported, skipping symlink %s.\n"),
1507 quote (con->target));
1508 #endif /* not HAVE_SYMLINK */
1510 else /* opt.retr_symlinks */
1513 err = ftp_loop_internal (u, f, con);
1514 } /* opt.retr_symlinks */
1518 logprintf (LOG_NOTQUIET, _("Skipping directory %s.\n"),
1519 quote (escnonprint (f->name)));
1522 /* Call the retrieve loop. */
1524 err = ftp_loop_internal (u, f, con);
1527 logprintf (LOG_NOTQUIET, _("%s: unknown/unsupported file type.\n"),
1528 escnonprint (f->name));
1532 /* Set the time-stamp information to the local file. Symlinks
1533 are not to be stamped because it sets the stamp on the
1535 if (!(f->type == FT_SYMLINK && !opt.retr_symlinks)
1538 && file_exists_p (con->target))
1540 /* #### This code repeats in http.c and ftp.c. Move it to a
1542 const char *fl = NULL;
1543 if (opt.output_document)
1545 if (output_stream_regular)
1546 fl = opt.output_document;
1551 touch (fl, f->tstamp);
1553 else if (f->tstamp == -1)
1554 logprintf (LOG_NOTQUIET, _("%s: corrupt time-stamp.\n"), con->target);
1556 if (f->perms && f->type == FT_PLAINFILE && dlthis)
1558 if (opt.preserve_perm)
1559 chmod (con->target, f->perms);
1562 DEBUGP (("Unrecognized permissions for %s.\n", con->target));
1564 xfree (con->target);
1565 con->target = old_target;
1567 url_set_file (u, ofile);
1570 /* Break on fatals. */
1571 if (err == QUOTEXC || err == HOSTERR || err == FWRITEERR)
1573 con->cmd &= ~ (DO_CWD | DO_LOGIN);
1577 /* We do not want to call ftp_retrieve_dirs here */
1578 if (opt.recursive &&
1579 !(opt.reclevel != INFINITE_RECURSION && depth >= opt.reclevel))
1580 err = ftp_retrieve_dirs (u, orig, con);
1581 else if (opt.recursive)
1582 DEBUGP ((_("Will not retrieve dirs since depth is %d (max %d).\n"),
1583 depth, opt.reclevel));
1588 /* Retrieve the directories given in a file list. This function works
1589 by simply going through the linked list and calling
1590 ftp_retrieve_glob on each directory entry. The function knows
1591 about excluded directories. */
1593 ftp_retrieve_dirs (struct url *u, struct fileinfo *f, ccon *con)
1595 char *container = NULL;
1596 int container_size = 0;
1598 for (; f; f = f->next)
1601 char *odir, *newdir;
1603 if (opt.quota && total_downloaded_bytes > opt.quota)
1605 if (f->type != FT_DIRECTORY)
1608 /* Allocate u->dir off stack, but reallocate only if a larger
1609 string is needed. It's a pity there's no "realloca" for an
1610 item on the bottom of the stack. */
1611 size = strlen (u->dir) + 1 + strlen (f->name) + 1;
1612 if (size > container_size)
1613 container = (char *)alloca (size);
1618 || (*odir == '/' && *(odir + 1) == '\0'))
1619 /* If ODIR is empty or just "/", simply append f->name to
1620 ODIR. (In the former case, to preserve u->dir being
1621 relative; in the latter case, to avoid double slash.) */
1622 sprintf (newdir, "%s%s", odir, f->name);
1624 /* Else, use a separator. */
1625 sprintf (newdir, "%s/%s", odir, f->name);
1627 DEBUGP (("Composing new CWD relative to the initial directory.\n"));
1628 DEBUGP ((" odir = '%s'\n f->name = '%s'\n newdir = '%s'\n\n",
1629 odir, f->name, newdir));
1630 if (!accdir (newdir))
1632 logprintf (LOG_VERBOSE, _("\
1633 Not descending to %s as it is excluded/not-included.\n"),
1634 quote (escnonprint (newdir)));
1638 con->st &= ~DONE_CWD;
1640 odir = xstrdup (u->dir); /* because url_set_dir will free
1642 url_set_dir (u, newdir);
1643 ftp_retrieve_glob (u, con, GLOB_GETALL);
1644 url_set_dir (u, odir);
1647 /* Set the time-stamp? */
1650 if (opt.quota && total_downloaded_bytes > opt.quota)
1656 /* Return true if S has a leading '/' or contains '../' */
1658 has_insecure_name_p (const char *s)
1663 if (strstr (s, "../") != 0)
1669 /* A near-top-level function to retrieve the files in a directory.
1670 The function calls ftp_get_listing, to get a linked list of files.
1671 Then it weeds out the file names that do not match the pattern.
1672 ftp_retrieve_list is called with this updated list as an argument.
1674 If the argument ACTION is GLOB_GETONE, just download the file (but
1675 first get the listing, so that the time-stamp is heeded); if it's
1676 GLOB_GLOBALL, use globbing; if it's GLOB_GETALL, download the whole
1679 ftp_retrieve_glob (struct url *u, ccon *con, int action)
1681 struct fileinfo *f, *start;
1684 con->cmd |= LEAVE_PENDING;
1686 res = ftp_get_listing (u, con, &start);
1689 /* First: weed out that do not conform the global rules given in
1690 opt.accepts and opt.rejects. */
1691 if (opt.accepts || opt.rejects)
1696 if (f->type != FT_DIRECTORY && !acceptable (f->name))
1698 logprintf (LOG_VERBOSE, _("Rejecting %s.\n"),
1699 quote (escnonprint (f->name)));
1700 f = delelement (f, &start);
1706 /* Remove all files with possible harmful names */
1710 if (has_insecure_name_p (f->name))
1712 logprintf (LOG_VERBOSE, _("Rejecting %s.\n"),
1713 quote (escnonprint (f->name)));
1714 f = delelement (f, &start);
1719 /* Now weed out the files that do not match our globbing pattern.
1720 If we are dealing with a globbing pattern, that is. */
1723 if (action == GLOB_GLOBALL)
1725 int (*matcher) (const char *, const char *, int)
1726 = opt.ignore_case ? fnmatch_nocase : fnmatch;
1732 matchres = matcher (u->file, f->name, 0);
1735 logprintf (LOG_NOTQUIET, _("Error matching %s against %s: %s\n"),
1736 u->file, escnonprint (f->name), strerror (errno));
1739 if (matchres == FNM_NOMATCH)
1740 f = delelement (f, &start); /* delete the element from the list */
1742 f = f->next; /* leave the element in the list */
1746 freefileinfo (start);
1747 return RETRBADPATTERN;
1750 else if (action == GLOB_GETONE)
1752 int (*cmp) (const char *, const char *)
1753 = opt.ignore_case ? strcasecmp : strcmp;
1757 if (0 != cmp(u->file, f->name))
1758 f = delelement (f, &start);
1766 /* Just get everything. */
1767 ftp_retrieve_list (u, start, con);
1771 if (action == GLOB_GLOBALL)
1774 /* #### This message SUCKS. We should see what was the
1775 reason that nothing was retrieved. */
1776 logprintf (LOG_VERBOSE, _("No matches on pattern %s.\n"),
1777 quote (escnonprint (u->file)));
1779 else /* GLOB_GETONE or GLOB_GETALL */
1781 /* Let's try retrieving it anyway. */
1782 con->st |= ON_YOUR_OWN;
1783 res = ftp_loop_internal (u, NULL, con);
1787 freefileinfo (start);
1788 if (opt.quota && total_downloaded_bytes > opt.quota)
1791 /* #### Should we return `res' here? */
1795 /* The wrapper that calls an appropriate routine according to contents
1796 of URL. Inherently, its capabilities are limited on what can be
1797 encoded into a URL. */
1799 ftp_loop (struct url *u, int *dt, struct url *proxy, bool recursive, bool glob)
1801 ccon con; /* FTP connection */
1809 con.st = ON_YOUR_OWN;
1814 /* If the file name is empty, the user probably wants a directory
1815 index. We'll provide one, properly HTML-ized. Unless
1816 opt.htmlify is 0, of course. :-) */
1817 if (!*u->file && !recursive)
1820 res = ftp_get_listing (u, &con, &f);
1824 if (opt.htmlify && !opt.spider)
1826 char *filename = (opt.output_document
1827 ? xstrdup (opt.output_document)
1828 : (con.target ? xstrdup (con.target)
1829 : url_file_name (u)));
1830 res = ftp_index (filename, u, f);
1831 if (res == FTPOK && opt.verbose)
1833 if (!opt.output_document)
1837 if (stat (filename, &st) == 0)
1841 logprintf (LOG_NOTQUIET,
1842 _("Wrote HTML-ized index to %s [%s].\n"),
1843 quote (filename), number_to_static_string (sz));
1846 logprintf (LOG_NOTQUIET,
1847 _("Wrote HTML-ized index to %s.\n"),
1857 bool ispattern = false;
1860 /* Treat the URL as a pattern if the file name part of the
1861 URL path contains wildcards. (Don't check for u->file
1862 because it is unescaped and therefore doesn't leave users
1863 the option to escape literal '*' as %2A.) */
1864 char *file_part = strrchr (u->path, '/');
1866 file_part = u->path;
1867 ispattern = has_wildcards_p (file_part);
1869 if (ispattern || recursive || opt.timestamping)
1871 /* ftp_retrieve_glob is a catch-all function that gets called
1872 if we need globbing, time-stamping or recursion. Its
1873 third argument is just what we really need. */
1874 res = ftp_retrieve_glob (u, &con,
1875 ispattern ? GLOB_GLOBALL : GLOB_GETONE);
1878 res = ftp_loop_internal (u, NULL, &con);
1884 /* If a connection was left, quench it. */
1885 if (con.csock != -1)
1886 fd_close (con.csock);
1887 xfree_null (con.id);
1889 xfree_null (con.target);
1894 /* Delete an element from the fileinfo linked list. Returns the
1895 address of the next element, or NULL if the list is exhausted. It
1896 can modify the start of the list. */
1897 static struct fileinfo *
1898 delelement (struct fileinfo *f, struct fileinfo **start)
1900 struct fileinfo *prev = f->prev;
1901 struct fileinfo *next = f->next;
1904 xfree_null (f->linkto);
1916 /* Free the fileinfo linked list of files. */
1918 freefileinfo (struct fileinfo *f)
1922 struct fileinfo *next = f->next;