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>
56 #include "convert.h" /* for downloaded_file */
62 /* File where the "ls -al" listing will be saved. */
63 #define LIST_FILENAME ".listing"
65 extern char ftp_last_respline[];
69 int st; /* connection status */
70 int cmd; /* command code */
71 struct rbuf rbuf; /* control connection buffer */
72 double dltime; /* time of the download in msecs */
73 enum stype rs; /* remote system reported by ftp server */
74 char *id; /* initial directory */
75 char *target; /* target file name */
76 struct url *proxy; /* FTWK-style proxy */
80 /* Look for regexp "( *[0-9]+ *byte" (literal parenthesis) anywhere in
81 the string S, and return the number converted to long, if found, 0
84 ftp_expected_bytes (const char *s)
90 while (*s && *s != '(')
94 for (++s; *s && ISSPACE (*s); s++);
102 res = (*s - '0') + 10 * res;
105 while (*s && ISDIGIT (*s));
108 while (*s && ISSPACE (*s))
112 if (TOLOWER (*s) != 'b')
114 if (strncasecmp (s, "byte", 4))
122 /* Retrieves a file with denoted parameters through opening an FTP
123 connection to the server. It always closes the data connection,
124 and closes the control connection in case of error. */
126 getftp (struct url *u, long *len, long restval, ccon *con)
128 int csock, dtsock, res;
131 char *user, *passwd, *respline;
134 int pasv_mode_open = 0;
135 long expected_bytes = 0L;
137 assert (con != NULL);
138 assert (con->target != NULL);
140 /* Debug-check of the sanity of the request by making sure that LIST
141 and RETR are never both requested (since we can handle only one
143 assert (!((cmd & DO_LIST) && (cmd & DO_RETR)));
144 /* Make sure that at least *something* is requested. */
145 assert ((cmd & (DO_LIST | DO_CWD | DO_RETR | DO_LOGIN)) != 0);
149 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 1);
150 user = user ? user : opt.ftp_acc;
151 passwd = passwd ? passwd : opt.ftp_pass;
152 assert (user && passwd);
157 if (!(cmd & DO_LOGIN))
158 csock = RBUF_FD (&con->rbuf);
159 else /* cmd & DO_LOGIN */
162 struct address_list *al;
164 char *host = con->proxy ? con->proxy->host : u->host;
165 int port = con->proxy ? con->proxy->port : u->port;
166 char *logname = user;
170 /* If proxy is in use, log in as username@target-site. */
171 logname = xmalloc (strlen (user) + 1 + strlen (u->host) + 1);
172 sprintf (logname, "%s@%s", user, u->host);
175 /* Login to the server: */
177 /* First: Establish the control connection. */
179 al = lookup_host (host, 0);
182 set_connection_host_name (host);
183 csock = connect_to_many (al, port, 0);
184 set_connection_host_name (NULL);
185 address_list_release (al);
188 return CONNECT_ERROR (errno);
190 if (cmd & LEAVE_PENDING)
191 rbuf_initialize (&con->rbuf, csock);
193 rbuf_uninitialize (&con->rbuf);
195 /* Since this is a new connection, we may safely discard
196 anything left in the buffer. */
197 rbuf_discard (&con->rbuf);
199 /* Second: Login with proper USER/PASS sequence. */
200 logprintf (LOG_VERBOSE, _("Logging in as %s ... "), user);
201 if (opt.server_response)
202 logputs (LOG_ALWAYS, "\n");
203 err = ftp_login (&con->rbuf, logname, passwd);
208 /* FTPRERR, FTPSRVERR, WRITEFAILED, FTPLOGREFUSED, FTPLOGINC */
212 logputs (LOG_VERBOSE, "\n");
213 logputs (LOG_NOTQUIET, _("\
214 Error in server response, closing control connection.\n"));
216 rbuf_uninitialize (&con->rbuf);
220 logputs (LOG_VERBOSE, "\n");
221 logputs (LOG_NOTQUIET, _("Error in server greeting.\n"));
223 rbuf_uninitialize (&con->rbuf);
227 logputs (LOG_VERBOSE, "\n");
228 logputs (LOG_NOTQUIET,
229 _("Write failed, closing control connection.\n"));
231 rbuf_uninitialize (&con->rbuf);
235 logputs (LOG_VERBOSE, "\n");
236 logputs (LOG_NOTQUIET, _("The server refuses login.\n"));
238 rbuf_uninitialize (&con->rbuf);
239 return FTPLOGREFUSED;
242 logputs (LOG_VERBOSE, "\n");
243 logputs (LOG_NOTQUIET, _("Login incorrect.\n"));
245 rbuf_uninitialize (&con->rbuf);
249 if (!opt.server_response)
250 logputs (LOG_VERBOSE, _("Logged in!\n"));
257 /* Third: Get the system type */
258 if (!opt.server_response)
259 logprintf (LOG_VERBOSE, "==> SYST ... ");
260 err = ftp_syst (&con->rbuf, &con->rs);
265 logputs (LOG_VERBOSE, "\n");
266 logputs (LOG_NOTQUIET, _("\
267 Error in server response, closing control connection.\n"));
269 rbuf_uninitialize (&con->rbuf);
273 logputs (LOG_VERBOSE, "\n");
274 logputs (LOG_NOTQUIET,
275 _("Server error, can't determine system type.\n"));
278 /* Everything is OK. */
284 if (!opt.server_response && err != FTPSRVERR)
285 logputs (LOG_VERBOSE, _("done. "));
287 /* Fourth: Find the initial ftp directory */
289 if (!opt.server_response)
290 logprintf (LOG_VERBOSE, "==> PWD ... ");
291 err = ftp_pwd(&con->rbuf, &con->id);
296 logputs (LOG_VERBOSE, "\n");
297 logputs (LOG_NOTQUIET, _("\
298 Error in server response, closing control connection.\n"));
300 rbuf_uninitialize (&con->rbuf);
304 /* PWD unsupported -- assume "/". */
305 FREE_MAYBE (con->id);
306 con->id = xstrdup ("/");
309 /* Everything is OK. */
315 /* VMS will report something like "PUB$DEVICE:[INITIAL.FOLDER]".
316 Convert it to "/INITIAL/FOLDER" */
317 if (con->rs == ST_VMS)
319 char *path = strchr (con->id, '[');
320 char *pathend = path ? strchr (path + 1, ']') : NULL;
321 if (!path || !pathend)
322 DEBUGP (("Initial VMS directory not in the form [...]!\n"));
325 char *idir = con->id;
326 DEBUGP (("Preprocessing the initial VMS directory\n"));
327 DEBUGP ((" old = '%s'\n", con->id));
328 /* We do the conversion in-place by copying the stuff
329 between [ and ] to the beginning, and changing dots
330 to slashes at the same time. */
332 for (++path; path < pathend; path++, idir++)
333 *idir = *path == '.' ? '/' : *path;
335 DEBUGP ((" new = '%s'\n\n", con->id));
338 if (!opt.server_response)
339 logputs (LOG_VERBOSE, _("done.\n"));
341 /* Fifth: Set the FTP type. */
342 type_char = ftp_process_type (u->params);
343 if (!opt.server_response)
344 logprintf (LOG_VERBOSE, "==> TYPE %c ... ", type_char);
345 err = ftp_type (&con->rbuf, type_char);
346 /* FTPRERR, WRITEFAILED, FTPUNKNOWNTYPE */
350 logputs (LOG_VERBOSE, "\n");
351 logputs (LOG_NOTQUIET, _("\
352 Error in server response, closing control connection.\n"));
354 rbuf_uninitialize (&con->rbuf);
358 logputs (LOG_VERBOSE, "\n");
359 logputs (LOG_NOTQUIET,
360 _("Write failed, closing control connection.\n"));
362 rbuf_uninitialize (&con->rbuf);
366 logputs (LOG_VERBOSE, "\n");
367 logprintf (LOG_NOTQUIET,
368 _("Unknown type `%c', closing control connection.\n"),
371 rbuf_uninitialize (&con->rbuf);
374 /* Everything is OK. */
380 if (!opt.server_response)
381 logputs (LOG_VERBOSE, _("done. "));
387 logputs (LOG_VERBOSE, _("==> CWD not needed.\n"));
390 char *target = u->dir;
392 DEBUGP (("changing working directory\n"));
394 /* Change working directory. To change to a non-absolute
395 Unix directory, we need to prepend initial directory
396 (con->id) to it. Absolute directories "just work".
398 A relative directory is one that does not begin with '/'
399 and, on non-Unix OS'es, one that doesn't begin with
402 This is not done for OS400, which doesn't use
403 "/"-delimited directories, nor does it support directory
404 hierarchies. "CWD foo" followed by "CWD bar" leaves us
405 in "bar", not in "foo/bar", as would be customary
409 && !(con->rs != ST_UNIX
410 && ISALPHA (target[0])
412 && con->rs != ST_OS400)
414 int idlen = strlen (con->id);
417 /* Strip trailing slash(es) from con->id. */
418 while (idlen > 0 && con->id[idlen - 1] == '/')
420 p = ntarget = (char *)alloca (idlen + 1 + strlen (u->dir) + 1);
421 memcpy (p, con->id, idlen);
426 DEBUGP (("Prepended initial PWD to relative path:\n"));
427 DEBUGP ((" pwd: '%s'\n old: '%s'\n new: '%s'\n",
428 con->id, target, ntarget));
432 /* If the FTP host runs VMS, we will have to convert the absolute
433 directory path in UNIX notation to absolute directory path in
434 VMS notation as VMS FTP servers do not like UNIX notation of
435 absolute paths. "VMS notation" is [dir.subdir.subsubdir]. */
437 if (con->rs == ST_VMS)
440 char *ntarget = (char *)alloca (strlen (target) + 2);
441 /* We use a converted initial dir, so directories in
442 TARGET will be separated with slashes, something like
443 "/INITIAL/FOLDER/DIR/SUBDIR". Convert that to
444 "[INITIAL.FOLDER.DIR.SUBDIR]". */
445 strcpy (ntarget, target);
446 assert (*ntarget == '/');
448 for (tmpp = ntarget + 1; *tmpp; tmpp++)
453 DEBUGP (("Changed file name to VMS syntax:\n"));
454 DEBUGP ((" Unix: '%s'\n VMS: '%s'\n", target, ntarget));
458 if (!opt.server_response)
459 logprintf (LOG_VERBOSE, "==> CWD %s ... ", target);
460 err = ftp_cwd (&con->rbuf, target);
461 /* FTPRERR, WRITEFAILED, FTPNSFOD */
465 logputs (LOG_VERBOSE, "\n");
466 logputs (LOG_NOTQUIET, _("\
467 Error in server response, closing control connection.\n"));
469 rbuf_uninitialize (&con->rbuf);
473 logputs (LOG_VERBOSE, "\n");
474 logputs (LOG_NOTQUIET,
475 _("Write failed, closing control connection.\n"));
477 rbuf_uninitialize (&con->rbuf);
481 logputs (LOG_VERBOSE, "\n");
482 logprintf (LOG_NOTQUIET, _("No such directory `%s'.\n\n"),
485 rbuf_uninitialize (&con->rbuf);
495 if (!opt.server_response)
496 logputs (LOG_VERBOSE, _("done.\n"));
499 else /* do not CWD */
500 logputs (LOG_VERBOSE, _("==> CWD not required.\n"));
502 if ((cmd & DO_RETR) && restval && *len == 0)
506 if (!opt.server_response)
507 logprintf (LOG_VERBOSE, "==> SIZE %s ... ", u->file);
510 err = ftp_size(&con->rbuf, u->file, len);
516 logputs (LOG_VERBOSE, "\n");
517 logputs (LOG_NOTQUIET, _("\
518 Error in server response, closing control connection.\n"));
520 rbuf_uninitialize (&con->rbuf);
524 /* Everything is OK. */
530 if (!opt.server_response)
531 logputs (LOG_VERBOSE, _("done.\n"));
534 /* If anything is to be retrieved, PORT (or PASV) must be sent. */
535 if (cmd & (DO_LIST | DO_RETR))
537 if (opt.ftp_pasv > 0)
539 ip_address passive_addr;
540 unsigned short passive_port;
541 if (!opt.server_response)
542 logputs (LOG_VERBOSE, "==> PASV ... ");
543 err = ftp_pasv (&con->rbuf, &passive_addr, &passive_port);
544 /* FTPRERR, WRITEFAILED, FTPNOPASV, FTPINVPASV */
548 logputs (LOG_VERBOSE, "\n");
549 logputs (LOG_NOTQUIET, _("\
550 Error in server response, closing control connection.\n"));
552 rbuf_uninitialize (&con->rbuf);
556 logputs (LOG_VERBOSE, "\n");
557 logputs (LOG_NOTQUIET,
558 _("Write failed, closing control connection.\n"));
560 rbuf_uninitialize (&con->rbuf);
564 logputs (LOG_VERBOSE, "\n");
565 logputs (LOG_NOTQUIET, _("Cannot initiate PASV transfer.\n"));
568 logputs (LOG_VERBOSE, "\n");
569 logputs (LOG_NOTQUIET, _("Cannot parse PASV response.\n"));
580 dtsock = connect_to_one (&passive_addr, passive_port, 1);
583 int save_errno = errno;
585 rbuf_uninitialize (&con->rbuf);
586 logprintf (LOG_VERBOSE, _("couldn't connect to %s:%hu: %s\n"),
587 pretty_print_address (&passive_addr), passive_port,
588 strerror (save_errno));
589 return CONNECT_ERROR (save_errno);
592 pasv_mode_open = 1; /* Flag to avoid accept port */
593 if (!opt.server_response)
594 logputs (LOG_VERBOSE, _("done. "));
598 if (!pasv_mode_open) /* Try to use a port command if PASV failed */
600 if (!opt.server_response)
601 logputs (LOG_VERBOSE, "==> PORT ... ");
602 err = ftp_port (&con->rbuf);
603 /* FTPRERR, WRITEFAILED, bindport (CONSOCKERR, CONPORTERR, BINDERR,
604 LISTENERR), HOSTERR, FTPPORTERR */
608 logputs (LOG_VERBOSE, "\n");
609 logputs (LOG_NOTQUIET, _("\
610 Error in server response, closing control connection.\n"));
613 rbuf_uninitialize (&con->rbuf);
617 logputs (LOG_VERBOSE, "\n");
618 logputs (LOG_NOTQUIET,
619 _("Write failed, closing control connection.\n"));
622 rbuf_uninitialize (&con->rbuf);
626 logputs (LOG_VERBOSE, "\n");
627 logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno));
630 rbuf_uninitialize (&con->rbuf);
633 case CONPORTERR: case BINDERR: case LISTENERR:
634 /* What now? These problems are local... */
635 logputs (LOG_VERBOSE, "\n");
636 logprintf (LOG_NOTQUIET, _("Bind error (%s).\n"),
642 logputs (LOG_VERBOSE, "\n");
643 logputs (LOG_NOTQUIET, _("Invalid PORT.\n"));
646 rbuf_uninitialize (&con->rbuf);
656 if (!opt.server_response)
657 logputs (LOG_VERBOSE, _("done. "));
659 } /* cmd & (DO_LIST | DO_RETR) */
661 /* Restart if needed. */
662 if (restval && (cmd & DO_RETR))
664 if (!opt.server_response)
665 logprintf (LOG_VERBOSE, "==> REST %ld ... ", restval);
666 err = ftp_rest (&con->rbuf, restval);
668 /* FTPRERR, WRITEFAILED, FTPRESTFAIL */
672 logputs (LOG_VERBOSE, "\n");
673 logputs (LOG_NOTQUIET, _("\
674 Error in server response, closing control connection.\n"));
677 rbuf_uninitialize (&con->rbuf);
681 logputs (LOG_VERBOSE, "\n");
682 logputs (LOG_NOTQUIET,
683 _("Write failed, closing control connection.\n"));
686 rbuf_uninitialize (&con->rbuf);
690 /* If `-c' is specified and the file already existed when
691 Wget was started, it would be a bad idea for us to start
692 downloading it from scratch, effectively truncating it. */
693 if (opt.always_rest && (cmd & NO_TRUNCATE))
695 logprintf (LOG_NOTQUIET,
696 _("\nREST failed; will not truncate `%s'.\n"),
700 rbuf_uninitialize (&con->rbuf);
701 return CONTNOTSUPPORTED;
703 logputs (LOG_VERBOSE, _("\nREST failed, starting from scratch.\n"));
713 if (err != FTPRESTFAIL && !opt.server_response)
714 logputs (LOG_VERBOSE, _("done. "));
715 } /* restval && cmd & DO_RETR */
719 /* If we're in spider mode, don't really retrieve anything. The
720 fact that we got to this point should be proof enough that
721 the file exists, vaguely akin to HTTP's concept of a "HEAD"
727 rbuf_uninitialize (&con->rbuf);
733 if (!opt.server_response)
736 logputs (LOG_VERBOSE, "\n");
737 logprintf (LOG_VERBOSE, "==> RETR %s ... ", u->file);
740 err = ftp_retr (&con->rbuf, u->file);
741 /* FTPRERR, WRITEFAILED, FTPNSFOD */
745 logputs (LOG_VERBOSE, "\n");
746 logputs (LOG_NOTQUIET, _("\
747 Error in server response, closing control connection.\n"));
750 rbuf_uninitialize (&con->rbuf);
754 logputs (LOG_VERBOSE, "\n");
755 logputs (LOG_NOTQUIET,
756 _("Write failed, closing control connection.\n"));
759 rbuf_uninitialize (&con->rbuf);
763 logputs (LOG_VERBOSE, "\n");
764 logprintf (LOG_NOTQUIET, _("No such file `%s'.\n\n"), u->file);
776 if (!opt.server_response)
777 logputs (LOG_VERBOSE, _("done.\n"));
778 expected_bytes = ftp_expected_bytes (ftp_last_respline);
783 if (!opt.server_response)
784 logputs (LOG_VERBOSE, "==> LIST ... ");
785 /* As Maciej W. Rozycki (macro@ds2.pg.gda.pl) says, `LIST'
786 without arguments is better than `LIST .'; confirmed by
788 err = ftp_list (&con->rbuf, NULL);
789 /* FTPRERR, WRITEFAILED */
793 logputs (LOG_VERBOSE, "\n");
794 logputs (LOG_NOTQUIET, _("\
795 Error in server response, closing control connection.\n"));
798 rbuf_uninitialize (&con->rbuf);
802 logputs (LOG_VERBOSE, "\n");
803 logputs (LOG_NOTQUIET,
804 _("Write failed, closing control connection.\n"));
807 rbuf_uninitialize (&con->rbuf);
811 logputs (LOG_VERBOSE, "\n");
812 logprintf (LOG_NOTQUIET, _("No such file or directory `%s'.\n\n"),
824 if (!opt.server_response)
825 logputs (LOG_VERBOSE, _("done.\n"));
826 expected_bytes = ftp_expected_bytes (ftp_last_respline);
827 } /* cmd & DO_LIST */
829 if (!(cmd & (DO_LIST | DO_RETR)) || (opt.spider && !(cmd & DO_LIST)))
832 /* Some FTP servers return the total length of file after REST
833 command, others just return the remaining size. */
834 if (*len && restval && expected_bytes
835 && (expected_bytes == *len - restval))
837 DEBUGP (("Lying FTP server found, adjusting.\n"));
838 expected_bytes = *len;
841 /* If no transmission was required, then everything is OK. */
842 if (!pasv_mode_open) /* we are not using pasive mode so we need
845 /* Open the data transmission socket by calling acceptport(). */
846 err = acceptport (&dtsock);
847 /* Possible errors: ACCEPTERR. */
848 if (err == ACCEPTERR)
850 logprintf (LOG_NOTQUIET, "accept: %s\n", strerror (errno));
855 /* Open the file -- if opt.dfp is set, use it instead. */
856 if (!opt.dfp || con->cmd & DO_LIST)
858 mkalldirs (con->target);
860 rotate_backups (con->target);
861 /* #### Is this correct? */
862 chmod (con->target, 0600);
864 fp = fopen (con->target, restval ? "ab" : "wb");
867 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target, strerror (errno));
869 rbuf_uninitialize (&con->rbuf);
876 extern int global_download_count;
879 /* Rewind the output document if the download starts over and if
880 this is the first download. See gethttp() for a longer
882 if (!restval && global_download_count == 0 && opt.dfp != stdout)
884 /* This will silently fail for streams that don't correspond
885 to regular files, but that's OK. */
887 /* ftruncate is needed because opt.dfp is opened in append
888 mode if opt.always_rest is set. */
889 ftruncate (fileno (fp), 0);
896 logprintf (LOG_VERBOSE, _("Length: %s"), legible (*len));
898 logprintf (LOG_VERBOSE, _(" [%s to go]"), legible (*len - restval));
899 logputs (LOG_VERBOSE, "\n");
900 expected_bytes = *len; /* for get_contents/show_progress */
902 else if (expected_bytes)
904 logprintf (LOG_VERBOSE, _("Length: %s"), legible (expected_bytes));
906 logprintf (LOG_VERBOSE, _(" [%s to go]"),
907 legible (expected_bytes - restval));
908 logputs (LOG_VERBOSE, _(" (unauthoritative)\n"));
911 /* Get the contents of the document. */
912 res = get_contents (dtsock, fp, len, restval, expected_bytes, &con->rbuf,
914 tms = time_str (NULL);
915 tmrate = retr_rate (*len - restval, con->dltime, 0);
916 /* Close data connection socket. */
918 /* Close the local file. */
920 /* Close or flush the file. We have to be careful to check for
921 error here. Checking the result of fwrite() is not enough --
922 errors could go unnoticed! */
924 if (!opt.dfp || con->cmd & DO_LIST)
925 flush_res = fclose (fp);
927 flush_res = fflush (fp);
928 if (flush_res == EOF)
931 /* If get_contents couldn't write to fp, bail out. */
934 logprintf (LOG_NOTQUIET, _("%s: %s, closing control connection.\n"),
935 con->target, strerror (errno));
937 rbuf_uninitialize (&con->rbuf);
942 logprintf (LOG_NOTQUIET, _("%s (%s) - Data connection: %s; "),
943 tms, tmrate, strerror (errno));
944 if (opt.server_response)
945 logputs (LOG_ALWAYS, "\n");
948 /* Get the server to tell us if everything is retrieved. */
949 err = ftp_response (&con->rbuf, &respline);
950 /* ...and empty the buffer. */
951 rbuf_discard (&con->rbuf);
955 /* The control connection is decidedly closed. Print the time
956 only if it hasn't already been printed. */
958 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
959 logputs (LOG_NOTQUIET, _("Control connection closed.\n"));
960 /* If there is an error on the control connection, close it, but
961 return FTPRETRINT, since there is a possibility that the
962 whole file was retrieved nevertheless (but that is for
963 ftp_loop_internal to decide). */
965 rbuf_uninitialize (&con->rbuf);
968 /* If retrieval failed for any reason, return FTPRETRINT, but do not
969 close socket, since the control connection is still alive. If
970 there is something wrong with the control connection, it will
971 become apparent later. */
972 if (*respline != '2')
976 logprintf (LOG_NOTQUIET, "%s (%s) - ", tms, tmrate);
977 logputs (LOG_NOTQUIET, _("Data transfer aborted.\n"));
984 /* What now? The data connection was erroneous, whereas the
985 response says everything is OK. We shall play it safe. */
989 if (!(cmd & LEAVE_PENDING))
991 /* I should probably send 'QUIT' and check for a reply, but this
992 is faster. #### Is it OK, though? */
994 rbuf_uninitialize (&con->rbuf);
996 /* If it was a listing, and opt.server_response is true,
998 if (opt.server_response && (con->cmd & DO_LIST))
1000 mkalldirs (con->target);
1001 fp = fopen (con->target, "r");
1003 logprintf (LOG_ALWAYS, "%s: %s\n", con->target, strerror (errno));
1007 /* The lines are being read with read_whole_line because of
1008 no-buffering on opt.lfile. */
1009 while ((line = read_whole_line (fp)))
1011 logprintf (LOG_ALWAYS, "%s\n", line);
1016 } /* con->cmd & DO_LIST && server_response */
1018 return RETRFINISHED;
1021 /* A one-file FTP loop. This is the part where FTP retrieval is
1022 retried, and retried, and retried, and...
1024 This loop either gets commands from con, or (if ON_YOUR_OWN is
1025 set), makes them up to retrieve the file given by the URL. */
1027 ftp_loop_internal (struct url *u, struct fileinfo *f, ccon *con)
1032 char *tmrate = NULL;
1037 con->target = url_file_name (u);
1039 if (opt.noclobber && file_exists_p (con->target))
1041 logprintf (LOG_VERBOSE,
1042 _("File `%s' already there, not retrieving.\n"), con->target);
1043 /* If the file is there, we suppose it's retrieved OK. */
1047 /* Remove it if it's a link. */
1048 remove_link (con->target);
1049 if (!opt.output_document)
1052 locf = opt.output_document;
1056 if (con->st & ON_YOUR_OWN)
1057 con->st = ON_YOUR_OWN;
1059 orig_lp = con->cmd & LEAVE_PENDING ? 1 : 0;
1064 /* Increment the pass counter. */
1066 sleep_between_retrievals (count);
1067 if (con->st & ON_YOUR_OWN)
1070 con->cmd |= (DO_RETR | LEAVE_PENDING);
1071 if (rbuf_initialized_p (&con->rbuf))
1072 con->cmd &= ~ (DO_LOGIN | DO_CWD);
1074 con->cmd |= (DO_LOGIN | DO_CWD);
1076 else /* not on your own */
1078 if (rbuf_initialized_p (&con->rbuf))
1079 con->cmd &= ~DO_LOGIN;
1081 con->cmd |= DO_LOGIN;
1082 if (con->st & DONE_CWD)
1083 con->cmd &= ~DO_CWD;
1088 /* Assume no restarting. */
1090 if ((count > 1 || opt.always_rest)
1091 && !(con->cmd & DO_LIST)
1092 && file_exists_p (locf))
1093 if (stat (locf, &st) == 0 && S_ISREG (st.st_mode))
1094 restval = st.st_size;
1096 /* In `-c' is used, check whether the file we're writing to
1097 exists and is of non-zero length. If so, we'll refuse to
1098 truncate it if the server doesn't support continued
1100 if (opt.always_rest && restval > 0)
1101 con->cmd |= NO_TRUNCATE;
1103 /* Get the current time string. */
1104 tms = time_str (NULL);
1105 /* Print fetch message, if opt.verbose. */
1108 char *hurl = url_string (u, 1);
1112 sprintf (tmp, _("(try:%2d)"), count);
1113 logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
1114 tms, hurl, tmp, locf);
1116 ws_changetitle (hurl, 1);
1120 /* Send getftp the proper length, if fileinfo was provided. */
1125 err = getftp (u, &len, restval, con);
1127 if (!rbuf_initialized_p (&con->rbuf))
1128 con->st &= ~DONE_CWD;
1130 con->st |= DONE_CWD;
1134 case HOSTERR: case CONREFUSED: case FWRITEERR: case FOPENERR:
1135 case FTPNSFOD: case FTPLOGINC: case FTPNOPASV: case CONTNOTSUPPORTED:
1136 /* Fatal errors, give up. */
1139 case CONSOCKERR: case CONERROR: case FTPSRVERR: case FTPRERR:
1140 case WRITEFAILED: case FTPUNKNOWNTYPE: case CONPORTERR:
1141 case BINDERR: case LISTENERR: case ACCEPTERR:
1142 case FTPPORTERR: case FTPLOGREFUSED: case FTPINVPASV:
1143 printwhat (count, opt.ntry);
1144 /* non-fatal errors */
1148 /* If the control connection was closed, the retrieval
1149 will be considered OK if f->size == len. */
1150 if (!f || len != f->size)
1152 printwhat (count, opt.ntry);
1164 tms = time_str (NULL);
1166 tmrate = retr_rate (len - restval, con->dltime, 0);
1168 /* If we get out of the switch above without continue'ing, we've
1169 successfully downloaded a file. Remember this fact. */
1170 downloaded_file (FILE_DOWNLOADED_NORMALLY, locf);
1172 if (con->st & ON_YOUR_OWN)
1174 CLOSE (RBUF_FD (&con->rbuf));
1175 rbuf_uninitialize (&con->rbuf);
1178 logprintf (LOG_VERBOSE, _("%s (%s) - `%s' saved [%ld]\n\n"),
1179 tms, tmrate, locf, len);
1180 if (!opt.verbose && !opt.quiet)
1182 /* Need to hide the password from the URL. The `if' is here
1183 so that we don't do the needless allocation every
1185 char *hurl = url_string (u, 1);
1186 logprintf (LOG_NONVERBOSE, "%s URL: %s [%ld] -> \"%s\" [%d]\n",
1187 tms, hurl, len, locf, count);
1191 if ((con->cmd & DO_LIST))
1192 /* This is a directory listing file. */
1194 if (!opt.remove_listing)
1195 /* --dont-remove-listing was specified, so do count this towards the
1196 number of bytes and files downloaded. */
1198 downloaded_increase (len);
1202 /* Deletion of listing files is not controlled by --delete-after, but
1203 by the more specific option --dont-remove-listing, and the code
1204 to do this deletion is in another function. */
1206 else if (!opt.spider)
1207 /* This is not a directory listing file. */
1209 /* Unlike directory listing files, don't pretend normal files weren't
1210 downloaded if they're going to be deleted. People seeding proxies,
1211 for instance, may want to know how many bytes and files they've
1212 downloaded through it. */
1213 downloaded_increase (len);
1216 if (opt.delete_after)
1218 DEBUGP (("Removing file due to --delete-after in"
1219 " ftp_loop_internal():\n"));
1220 logprintf (LOG_VERBOSE, _("Removing %s.\n"), locf);
1222 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1226 /* Restore the original leave-pendingness. */
1228 con->cmd |= LEAVE_PENDING;
1230 con->cmd &= ~LEAVE_PENDING;
1232 } while (!opt.ntry || (count < opt.ntry));
1234 if (rbuf_initialized_p (&con->rbuf) && (con->st & ON_YOUR_OWN))
1236 CLOSE (RBUF_FD (&con->rbuf));
1237 rbuf_uninitialize (&con->rbuf);
1242 /* Return the directory listing in a reusable format. The directory
1243 is specifed in u->dir. */
1245 ftp_get_listing (struct url *u, ccon *con, struct fileinfo **f)
1248 char *uf; /* url file name */
1249 char *lf; /* list file name */
1250 char *old_target = con->target;
1252 con->st &= ~ON_YOUR_OWN;
1253 con->cmd |= (DO_LIST | LEAVE_PENDING);
1254 con->cmd &= ~DO_RETR;
1256 /* Find the listing file name. We do it by taking the file name of
1257 the URL and replacing the last component with the listing file
1259 uf = url_file_name (u);
1260 lf = file_merge (uf, LIST_FILENAME);
1262 DEBUGP ((_("Using `%s' as listing tmp file.\n"), lf));
1265 err = ftp_loop_internal (u, NULL, con);
1266 con->target = old_target;
1269 *f = ftp_parse_ls (lf, con->rs);
1272 if (opt.remove_listing)
1275 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
1277 logprintf (LOG_VERBOSE, _("Removed `%s'.\n"), lf);
1280 con->cmd &= ~DO_LIST;
1284 static uerr_t ftp_retrieve_dirs PARAMS ((struct url *, struct fileinfo *,
1286 static uerr_t ftp_retrieve_glob PARAMS ((struct url *, ccon *, int));
1287 static struct fileinfo *delelement PARAMS ((struct fileinfo *,
1288 struct fileinfo **));
1289 static void freefileinfo PARAMS ((struct fileinfo *f));
1291 /* Retrieve a list of files given in struct fileinfo linked list. If
1292 a file is a symbolic link, do not retrieve it, but rather try to
1293 set up a similar link on the local disk, if the symlinks are
1296 If opt.recursive is set, after all files have been retrieved,
1297 ftp_retrieve_dirs will be called to retrieve the directories. */
1299 ftp_retrieve_list (struct url *u, struct fileinfo *f, ccon *con)
1301 static int depth = 0;
1303 struct fileinfo *orig;
1308 /* Increase the depth. */
1310 if (opt.reclevel != INFINITE_RECURSION && depth > opt.reclevel)
1312 DEBUGP ((_("Recursion depth %d exceeded max. depth %d.\n"),
1313 depth, opt.reclevel));
1321 con->st &= ~ON_YOUR_OWN;
1322 if (!(con->st & DONE_CWD))
1325 con->cmd &= ~DO_CWD;
1326 con->cmd |= (DO_RETR | LEAVE_PENDING);
1328 if (!rbuf_initialized_p (&con->rbuf))
1329 con->cmd |= DO_LOGIN;
1331 con->cmd &= ~DO_LOGIN;
1333 err = RETROK; /* in case it's not used */
1337 char *old_target, *ofile;
1339 if (downloaded_exceeds_quota ())
1344 old_target = con->target;
1346 ofile = xstrdup (u->file);
1347 url_set_file (u, f->name);
1349 con->target = url_file_name (u);
1353 if (opt.timestamping && f->type == FT_PLAINFILE)
1356 /* If conversion of HTML files retrieved via FTP is ever implemented,
1357 we'll need to stat() <file>.orig here when -K has been specified.
1358 I'm not implementing it now since files on an FTP server are much
1359 more likely than files on an HTTP server to legitimately have a
1361 if (!stat (con->target, &st))
1365 /* Else, get it from the file. */
1366 local_size = st.st_size;
1369 /* Modification time granularity is 2 seconds for Windows, so
1370 increase local time by 1 second for later comparison. */
1373 /* Compare file sizes only for servers that tell us correct
1374 values. Assumme sizes being equal for servers that lie
1376 cor_val = (con->rs == ST_UNIX || con->rs == ST_WINNT);
1377 eq_size = cor_val ? (local_size == f->size) : 1 ;
1378 if (f->tstamp <= tml && eq_size)
1380 /* Remote file is older, file sizes can be compared and
1382 logprintf (LOG_VERBOSE, _("\
1383 Remote file no newer than local file `%s' -- not retrieving.\n"), con->target);
1388 /* Remote file is newer or sizes cannot be matched */
1389 logprintf (LOG_VERBOSE, _("\
1390 Remote file is newer than local file `%s' -- retrieving.\n\n"),
1395 /* Sizes do not match */
1396 logprintf (LOG_VERBOSE, _("\
1397 The sizes do not match (local %ld) -- retrieving.\n\n"), local_size);
1400 } /* opt.timestamping && f->type == FT_PLAINFILE */
1404 /* If opt.retr_symlinks is defined, we treat symlinks as
1405 if they were normal files. There is currently no way
1406 to distinguish whether they might be directories, and
1408 if (!opt.retr_symlinks)
1412 logputs (LOG_NOTQUIET,
1413 _("Invalid name of the symlink, skipping.\n"));
1417 /* Check whether we already have the correct
1419 int rc = lstat (con->target, &st);
1422 size_t len = strlen (f->linkto) + 1;
1423 if (S_ISLNK (st.st_mode))
1425 char *link_target = (char *)alloca (len);
1426 size_t n = readlink (con->target, link_target, len);
1428 && (memcmp (link_target, f->linkto, n) == 0))
1430 logprintf (LOG_VERBOSE, _("\
1431 Already have correct symlink %s -> %s\n\n"),
1432 con->target, f->linkto);
1438 logprintf (LOG_VERBOSE, _("Creating symlink %s -> %s\n"),
1439 con->target, f->linkto);
1440 /* Unlink before creating symlink! */
1441 unlink (con->target);
1442 if (symlink (f->linkto, con->target) == -1)
1443 logprintf (LOG_NOTQUIET, "symlink: %s\n",
1445 logputs (LOG_VERBOSE, "\n");
1446 } /* have f->linkto */
1447 #else /* not HAVE_SYMLINK */
1448 logprintf (LOG_NOTQUIET,
1449 _("Symlinks not supported, skipping symlink `%s'.\n"),
1451 #endif /* not HAVE_SYMLINK */
1453 else /* opt.retr_symlinks */
1456 err = ftp_loop_internal (u, f, con);
1457 } /* opt.retr_symlinks */
1461 logprintf (LOG_NOTQUIET, _("Skipping directory `%s'.\n"),
1465 /* Call the retrieve loop. */
1467 err = ftp_loop_internal (u, f, con);
1470 logprintf (LOG_NOTQUIET, _("%s: unknown/unsupported file type.\n"),
1475 /* Set the time-stamp information to the local file. Symlinks
1476 are not to be stamped because it sets the stamp on the
1478 if (!(f->type == FT_SYMLINK && !opt.retr_symlinks)
1481 && file_exists_p (con->target))
1483 /* #### This code repeats in http.c and ftp.c. Move it to a
1485 const char *fl = NULL;
1486 if (opt.output_document)
1488 if (opt.od_known_regular)
1489 fl = opt.output_document;
1494 touch (fl, f->tstamp);
1496 else if (f->tstamp == -1)
1497 logprintf (LOG_NOTQUIET, _("%s: corrupt time-stamp.\n"), con->target);
1499 if (f->perms && f->type == FT_PLAINFILE && dlthis)
1500 chmod (con->target, f->perms);
1502 DEBUGP (("Unrecognized permissions for %s.\n", con->target));
1504 xfree (con->target);
1505 con->target = old_target;
1507 url_set_file (u, ofile);
1510 /* Break on fatals. */
1511 if (err == QUOTEXC || err == HOSTERR || err == FWRITEERR)
1513 con->cmd &= ~ (DO_CWD | DO_LOGIN);
1517 /* We do not want to call ftp_retrieve_dirs here */
1518 if (opt.recursive &&
1519 !(opt.reclevel != INFINITE_RECURSION && depth >= opt.reclevel))
1520 err = ftp_retrieve_dirs (u, orig, con);
1521 else if (opt.recursive)
1522 DEBUGP ((_("Will not retrieve dirs since depth is %d (max %d).\n"),
1523 depth, opt.reclevel));
1528 /* Retrieve the directories given in a file list. This function works
1529 by simply going through the linked list and calling
1530 ftp_retrieve_glob on each directory entry. The function knows
1531 about excluded directories. */
1533 ftp_retrieve_dirs (struct url *u, struct fileinfo *f, ccon *con)
1535 char *container = NULL;
1536 int container_size = 0;
1538 for (; f; f = f->next)
1541 char *odir, *newdir;
1543 if (downloaded_exceeds_quota ())
1545 if (f->type != FT_DIRECTORY)
1548 /* Allocate u->dir off stack, but reallocate only if a larger
1549 string is needed. It's a pity there's no "realloca" for an
1550 item on the bottom of the stack. */
1551 size = strlen (u->dir) + 1 + strlen (f->name) + 1;
1552 if (size > container_size)
1553 container = (char *)alloca (size);
1558 || (*odir == '/' && *(odir + 1) == '\0'))
1559 /* If ODIR is empty or just "/", simply append f->name to
1560 ODIR. (In the former case, to preserve u->dir being
1561 relative; in the latter case, to avoid double slash.) */
1562 sprintf (newdir, "%s%s", odir, f->name);
1564 /* Else, use a separator. */
1565 sprintf (newdir, "%s/%s", odir, f->name);
1567 DEBUGP (("Composing new CWD relative to the initial directory.\n"));
1568 DEBUGP ((" odir = '%s'\n f->name = '%s'\n newdir = '%s'\n\n",
1569 odir, f->name, newdir));
1570 if (!accdir (newdir, ALLABS))
1572 logprintf (LOG_VERBOSE, _("\
1573 Not descending to `%s' as it is excluded/not-included.\n"), newdir);
1577 con->st &= ~DONE_CWD;
1579 odir = xstrdup (u->dir); /* because url_set_dir will free
1581 url_set_dir (u, newdir);
1582 ftp_retrieve_glob (u, con, GETALL);
1583 url_set_dir (u, odir);
1586 /* Set the time-stamp? */
1589 if (opt.quota && opt.downloaded > opt.quota)
1595 /* Return non-zero if S has a leading '/' or contains '../' */
1597 has_insecure_name_p (const char *s)
1602 if (strstr(s, "../") != 0)
1608 /* A near-top-level function to retrieve the files in a directory.
1609 The function calls ftp_get_listing, to get a linked list of files.
1610 Then it weeds out the file names that do not match the pattern.
1611 ftp_retrieve_list is called with this updated list as an argument.
1613 If the argument ACTION is GETONE, just download the file (but first
1614 get the listing, so that the time-stamp is heeded); if it's GLOBALL,
1615 use globbing; if it's GETALL, download the whole directory. */
1617 ftp_retrieve_glob (struct url *u, ccon *con, int action)
1619 struct fileinfo *f, *orig, *start;
1622 con->cmd |= LEAVE_PENDING;
1624 res = ftp_get_listing (u, con, &orig);
1628 /* First: weed out that do not conform the global rules given in
1629 opt.accepts and opt.rejects. */
1630 if (opt.accepts || opt.rejects)
1635 if (f->type != FT_DIRECTORY && !acceptable (f->name))
1637 logprintf (LOG_VERBOSE, _("Rejecting `%s'.\n"), f->name);
1638 f = delelement (f, &start);
1644 /* Remove all files with possible harmful names */
1648 if (has_insecure_name_p (f->name))
1650 logprintf (LOG_VERBOSE, _("Rejecting `%s'.\n"), f->name);
1651 f = delelement (f, &start);
1656 /* Now weed out the files that do not match our globbing pattern.
1657 If we are dealing with a globbing pattern, that is. */
1658 if (*u->file && (action == GLOBALL || action == GETONE))
1665 matchres = fnmatch (u->file, f->name, 0);
1668 logprintf (LOG_NOTQUIET, "%s: %s\n", con->target,
1672 if (matchres == FNM_NOMATCH)
1673 f = delelement (f, &start); /* delete the element from the list */
1675 f = f->next; /* leave the element in the list */
1679 freefileinfo (start);
1680 return RETRBADPATTERN;
1686 /* Just get everything. */
1687 ftp_retrieve_list (u, start, con);
1691 if (action == GLOBALL)
1694 /* #### This message SUCKS. We should see what was the
1695 reason that nothing was retrieved. */
1696 logprintf (LOG_VERBOSE, _("No matches on pattern `%s'.\n"), u->file);
1698 else /* GETONE or GETALL */
1700 /* Let's try retrieving it anyway. */
1701 con->st |= ON_YOUR_OWN;
1702 res = ftp_loop_internal (u, NULL, con);
1706 freefileinfo (start);
1707 if (downloaded_exceeds_quota ())
1710 /* #### Should we return `res' here? */
1714 /* The wrapper that calls an appropriate routine according to contents
1715 of URL. Inherently, its capabilities are limited on what can be
1716 encoded into a URL. */
1718 ftp_loop (struct url *u, int *dt, struct url *proxy)
1720 ccon con; /* FTP connection */
1725 memset (&con, 0, sizeof (con));
1727 rbuf_uninitialize (&con.rbuf);
1728 con.st = ON_YOUR_OWN;
1732 res = RETROK; /* in case it's not used */
1734 /* If the file name is empty, the user probably wants a directory
1735 index. We'll provide one, properly HTML-ized. Unless
1736 opt.htmlify is 0, of course. :-) */
1737 if (!*u->file && !opt.recursive)
1740 res = ftp_get_listing (u, &con, &f);
1744 if (opt.htmlify && !opt.spider)
1746 char *filename = (opt.output_document
1747 ? xstrdup (opt.output_document)
1748 : (con.target ? xstrdup (con.target)
1749 : url_file_name (u)));
1750 res = ftp_index (filename, u, f);
1751 if (res == FTPOK && opt.verbose)
1753 if (!opt.output_document)
1757 if (stat (filename, &st) == 0)
1761 logprintf (LOG_NOTQUIET,
1762 _("Wrote HTML-ized index to `%s' [%ld].\n"),
1766 logprintf (LOG_NOTQUIET,
1767 _("Wrote HTML-ized index to `%s'.\n"),
1777 int wild = has_wildcards_p (u->file);
1778 if ((opt.ftp_glob && wild) || opt.recursive || opt.timestamping)
1780 /* ftp_retrieve_glob is a catch-all function that gets called
1781 if we need globbing, time-stamping or recursion. Its
1782 third argument is just what we really need. */
1783 res = ftp_retrieve_glob (u, &con,
1784 (opt.ftp_glob && wild) ? GLOBALL : GETONE);
1787 res = ftp_loop_internal (u, NULL, &con);
1793 /* If a connection was left, quench it. */
1794 if (rbuf_initialized_p (&con.rbuf))
1795 CLOSE (RBUF_FD (&con.rbuf));
1796 FREE_MAYBE (con.id);
1798 FREE_MAYBE (con.target);
1803 /* Delete an element from the fileinfo linked list. Returns the
1804 address of the next element, or NULL if the list is exhausted. It
1805 can modify the start of the list. */
1806 static struct fileinfo *
1807 delelement (struct fileinfo *f, struct fileinfo **start)
1809 struct fileinfo *prev = f->prev;
1810 struct fileinfo *next = f->next;
1813 FREE_MAYBE (f->linkto);
1825 /* Free the fileinfo linked list of files. */
1827 freefileinfo (struct fileinfo *f)
1831 struct fileinfo *next = f->next;