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 (at
10 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. */
37 #endif /* HAVE_UNISTD_H */
55 /* Total size of downloaded files. Used to enforce quota. */
56 SUM_SIZE_INT total_downloaded_bytes;
58 /* Total download time in seconds. */
59 double total_download_time;
61 /* If non-NULL, the stream to which output should be written. This
62 stream is initialized when `-O' is used. */
65 /* Whether output_document is a regular file we can manipulate,
66 i.e. not `-' or a device file. */
67 bool output_stream_regular;
76 limit_bandwidth_reset (void)
81 /* Limit the bandwidth by pausing the download for an amount of time.
82 BYTES is the number of bytes received from the network, and TIMER
83 is the timer that started at the beginning of download. */
86 limit_bandwidth (wgint bytes, struct ptimer *timer)
88 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
91 limit_data.chunk_bytes += bytes;
93 /* Calculate the amount of time we expect downloading the chunk
94 should take. If in reality it took less time, sleep to
95 compensate for the difference. */
96 expected = (double) limit_data.chunk_bytes / opt.limit_rate;
98 if (expected > delta_t)
100 double slp = expected - delta_t + limit_data.sleep_adjust;
104 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
105 slp * 1000, number_to_static_string (limit_data.chunk_bytes),
109 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
110 slp * 1000, number_to_static_string (limit_data.chunk_bytes),
111 limit_data.sleep_adjust));
113 t0 = ptimer_read (timer);
115 t1 = ptimer_measure (timer);
117 /* Due to scheduling, we probably slept slightly longer (or
118 shorter) than desired. Calculate the difference between the
119 desired and the actual sleep, and adjust the next sleep by
121 limit_data.sleep_adjust = slp - (t1 - t0);
122 /* If sleep_adjust is very large, it's likely due to suspension
123 and not clock inaccuracy. Don't enforce those. */
124 if (limit_data.sleep_adjust > 0.5)
125 limit_data.sleep_adjust = 0.5;
126 else if (limit_data.sleep_adjust < -0.5)
127 limit_data.sleep_adjust = -0.5;
130 limit_data.chunk_bytes = 0;
131 limit_data.chunk_start = ptimer_read (timer);
135 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
138 /* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
139 amount of data and decrease SKIP. Increment *TOTAL by the amount
143 write_data (FILE *out, const char *buf, int bufsize, wgint *skip,
162 fwrite (buf, 1, bufsize, out);
165 /* Immediately flush the downloaded data. This should not hinder
166 performance: fast downloads will arrive in large 16K chunks
167 (which stdio would write out immediately anyway), and slow
168 downloads wouldn't be limited by disk speed. */
170 return !ferror (out);
173 /* Read the contents of file descriptor FD until it the connection
174 terminates or a read error occurs. The data is read in portions of
175 up to 16K and written to OUT as it arrives. If opt.verbose is set,
176 the progress is shown.
178 TOREAD is the amount of data expected to arrive, normally only used
179 by the progress gauge.
181 STARTPOS is the position from which the download starts, used by
182 the progress gauge. If QTYREAD is non-NULL, the value it points to
183 is incremented by the amount of data read from the network. If
184 QTYWRITTEN is non-NULL, the value it points to is incremented by
185 the amount of data written to disk. The time it took to download
186 the data is stored to ELAPSED.
188 The function exits and returns the amount of data read. In case of
189 error while reading data, -1 is returned. In case of error while
190 writing data, -2 is returned. */
193 fd_read_body (int fd, FILE *out, wgint toread, wgint startpos,
194 wgint *qtyread, wgint *qtywritten, double *elapsed, int flags)
198 static char dlbuf[16384];
199 int dlbufsize = sizeof (dlbuf);
201 struct ptimer *timer = NULL;
202 double last_successful_read_tm = 0;
204 /* The progress gauge, set according to the user preferences. */
205 void *progress = NULL;
207 /* Non-zero if the progress gauge is interactive, i.e. if it can
208 continually update the display. When true, smaller timeout
209 values are used so that the gauge can update the display when
210 data arrives slowly. */
211 bool progress_interactive = false;
213 bool exact = !!(flags & rb_read_exactly);
216 /* How much data we've read/written. */
218 wgint sum_written = 0;
220 if (flags & rb_skip_startpos)
225 /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
226 argument to progress_create because the indicator doesn't
227 (yet) know about "skipping" data. */
228 progress = progress_create (skip ? 0 : startpos, startpos + toread);
229 progress_interactive = progress_interactive_p (progress);
233 limit_bandwidth_reset ();
235 /* A timer is needed for tracking progress, for throttling, and for
236 tracking elapsed time. If either of these are requested, start
238 if (progress || opt.limit_rate || elapsed)
240 timer = ptimer_new ();
241 last_successful_read_tm = 0;
244 /* Use a smaller buffer for low requested bandwidths. For example,
245 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
246 data and then sleep for 8s. With buffer size equal to the limit,
247 we never have to sleep for more than one second. */
248 if (opt.limit_rate && opt.limit_rate < dlbufsize)
249 dlbufsize = opt.limit_rate;
251 /* Read from FD while there is data to read. Normally toread==0
252 means that it is unknown how much data is to arrive. However, if
253 EXACT is set, then toread==0 means what it says: that no data
255 while (!exact || (sum_read < toread))
257 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
258 double tmout = opt.read_timeout;
259 if (progress_interactive)
261 /* For interactive progress gauges, always specify a ~1s
262 timeout, so that the gauge can be updated regularly even
263 when the data arrives very slowly or stalls. */
265 if (opt.read_timeout)
268 waittm = ptimer_read (timer) - last_successful_read_tm;
269 if (waittm + tmout > opt.read_timeout)
271 /* Don't let total idle time exceed read timeout. */
272 tmout = opt.read_timeout - waittm;
275 /* We've already exceeded the timeout. */
276 ret = -1, errno = ETIMEDOUT;
282 ret = fd_read (fd, dlbuf, rdsize, tmout);
284 if (progress_interactive && ret < 0 && errno == ETIMEDOUT)
285 ret = 0; /* interactive timeout, handled above */
287 break; /* EOF or read error */
289 if (progress || opt.limit_rate)
291 ptimer_measure (timer);
293 last_successful_read_tm = ptimer_read (timer);
299 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
307 limit_bandwidth (ret, timer);
310 progress_update (progress, ret, ptimer_read (timer));
312 if (toread > 0 && !opt.quiet)
313 ws_percenttitle (100.0 *
314 (startpos + sum_read) / (startpos + toread));
322 progress_finish (progress, ptimer_read (timer));
325 *elapsed = ptimer_read (timer);
327 ptimer_destroy (timer);
330 *qtyread += sum_read;
332 *qtywritten += sum_written;
337 /* Read a hunk of data from FD, up until a terminator. The hunk is
338 limited by whatever the TERMINATOR callback chooses as its
339 terminator. For example, if terminator stops at newline, the hunk
340 will consist of a line of data; if terminator stops at two
341 newlines, it can be used to read the head of an HTTP response.
342 Upon determining the boundary, the function returns the data (up to
343 the terminator) in malloc-allocated storage.
345 In case of read error, NULL is returned. In case of EOF and no
346 data read, NULL is returned and errno set to 0. In case of having
347 read some data, but encountering EOF before seeing the terminator,
348 the data that has been read is returned, but it will (obviously)
349 not contain the terminator.
351 The TERMINATOR function is called with three arguments: the
352 beginning of the data read so far, the beginning of the current
353 block of peeked-at data, and the length of the current block.
354 Depending on its needs, the function is free to choose whether to
355 analyze all data or just the newly arrived data. If TERMINATOR
356 returns NULL, it means that the terminator has not been seen.
357 Otherwise it should return a pointer to the charactre immediately
358 following the terminator.
360 The idea is to be able to read a line of input, or otherwise a hunk
361 of text, such as the head of an HTTP request, without crossing the
362 boundary, so that the next call to fd_read etc. reads the data
363 after the hunk. To achieve that, this function does the following:
365 1. Peek at incoming data.
367 2. Determine whether the peeked data, along with the previously
368 read data, includes the terminator.
370 2a. If yes, read the data until the end of the terminator, and
373 2b. If no, read the peeked data and goto 1.
375 The function is careful to assume as little as possible about the
376 implementation of peeking. For example, every peek is followed by
377 a read. If the read returns a different amount of data, the
378 process is retried until all data arrives safely.
380 SIZEHINT is the buffer size sufficient to hold all the data in the
381 typical case (it is used as the initial buffer size). MAXSIZE is
382 the maximum amount of memory this function is allowed to allocate,
383 or 0 if no upper limit is to be enforced.
385 This function should be used as a building block for other
386 functions -- see fd_read_line as a simple example. */
389 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
391 long bufsize = sizehint;
392 char *hunk = xmalloc (bufsize);
393 int tail = 0; /* tail position in HUNK */
395 assert (maxsize >= bufsize);
400 int pklen, rdlen, remain;
402 /* First, peek at the available data. */
404 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
410 end = terminator (hunk, hunk + tail, pklen);
413 /* The data contains the terminator: we'll drain the data up
414 to the end of the terminator. */
415 remain = end - (hunk + tail);
416 assert (remain >= 0);
419 /* No more data needs to be read. */
423 if (bufsize - 1 < tail + remain)
425 bufsize = tail + remain + 1;
426 hunk = xrealloc (hunk, bufsize);
430 /* No terminator: simply read the data we know is (or should
434 /* Now, read the data. Note that we make no assumptions about
435 how much data we'll get. (Some TCP stacks are notorious for
436 read returning less data than the previous MSG_PEEK.) */
438 rdlen = fd_read (fd, hunk + tail, remain, 0);
451 /* EOF without anything having been read */
457 /* EOF seen: return the data we've read. */
460 if (end && rdlen == remain)
461 /* The terminator was seen and the remaining data drained --
462 we got what we came for. */
465 /* Keep looping until all the data arrives. */
467 if (tail == bufsize - 1)
469 /* Double the buffer size, but refuse to allocate more than
471 if (maxsize && bufsize >= maxsize)
478 if (maxsize && bufsize > maxsize)
480 hunk = xrealloc (hunk, bufsize);
486 line_terminator (const char *start, const char *peeked, int peeklen)
488 const char *p = memchr (peeked, '\n', peeklen);
490 /* p+1 because the line must include '\n' */
495 /* The maximum size of the single line we agree to accept. This is
496 not meant to impose an arbitrary limit, but to protect the user
497 from Wget slurping up available memory upon encountering malicious
498 or buggy server output. Define it to 0 to remove the limit. */
499 #define FD_READ_LINE_MAX 4096
501 /* Read one line from FD and return it. The line is allocated using
502 malloc, but is never larger than FD_READ_LINE_MAX.
504 If an error occurs, or if no data can be read, NULL is returned.
505 In the former case errno indicates the error condition, and in the
506 latter case, errno is NULL. */
509 fd_read_line (int fd)
511 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
514 /* Return a printed representation of the download rate, along with
515 the units appropriate for the download speed. */
518 retr_rate (wgint bytes, double secs)
521 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
524 double dlrate = calc_rate (bytes, secs, &units);
525 /* Use more digits for smaller numbers (regardless of unit used),
526 e.g. "1022", "247", "12.5", "2.38". */
527 sprintf (res, "%.*f %s",
528 dlrate >= 99.95 ? 0 : dlrate >= 9.995 ? 1 : 2,
529 dlrate, rate_names[units]);
534 /* Calculate the download rate and trim it as appropriate for the
535 speed. Appropriate means that if rate is greater than 1K/s,
536 kilobytes are used, and if rate is greater than 1MB/s, megabytes
539 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
543 calc_rate (wgint bytes, double secs, int *units)
551 /* If elapsed time is exactly zero, it means we're under the
552 resolution of the timer. This can easily happen on systems
553 that use time() for the timer. Since the interval lies between
554 0 and the timer's resolution, assume half the resolution. */
555 secs = ptimer_resolution () / 2.0;
557 dlrate = bytes / secs;
560 else if (dlrate < 1024.0 * 1024.0)
561 *units = 1, dlrate /= 1024.0;
562 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
563 *units = 2, dlrate /= (1024.0 * 1024.0);
565 /* Maybe someone will need this, one day. */
566 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
572 #define SUSPEND_POST_DATA do { \
573 post_data_suspended = true; \
574 saved_post_data = opt.post_data; \
575 saved_post_file_name = opt.post_file_name; \
576 opt.post_data = NULL; \
577 opt.post_file_name = NULL; \
580 #define RESTORE_POST_DATA do { \
581 if (post_data_suspended) \
583 opt.post_data = saved_post_data; \
584 opt.post_file_name = saved_post_file_name; \
585 post_data_suspended = false; \
589 static char *getproxy (struct url *);
591 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
594 /* #### This function should be rewritten so it doesn't return from
598 retrieve_url (const char *origurl, char **file, char **newloc,
599 const char *refurl, int *dt, bool recursive)
603 bool location_changed;
605 char *mynewloc, *proxy;
606 struct url *u, *proxy_url;
607 int up_error_code; /* url parse error code */
609 int redirection_count = 0;
611 bool post_data_suspended = false;
612 char *saved_post_data = NULL;
613 char *saved_post_file_name = NULL;
615 /* If dt is NULL, use local storage. */
621 url = xstrdup (origurl);
627 u = url_parse (url, &up_error_code);
630 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
636 refurl = opt.referer;
645 proxy = getproxy (u);
648 /* Parse the proxy URL. */
649 proxy_url = url_parse (proxy, &up_error_code);
652 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
653 proxy, url_error (up_error_code));
658 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
660 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
661 url_free (proxy_url);
668 if (u->scheme == SCHEME_HTTP
670 || u->scheme == SCHEME_HTTPS
672 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
674 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
676 else if (u->scheme == SCHEME_FTP)
678 /* If this is a redirection, temporarily turn off opt.ftp_glob
679 and opt.recursive, both being undesirable when following
681 bool oldrec = recursive, glob = opt.ftp_glob;
682 if (redirection_count)
683 oldrec = glob = false;
685 result = ftp_loop (u, dt, proxy_url, recursive, glob);
688 /* There is a possibility of having HTTP being redirected to
689 FTP. In these cases we must decide whether the text is HTML
690 according to the suffix. The HTML suffixes are `.html',
691 `.htm' and a few others, case-insensitive. */
692 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
694 if (has_html_suffix_p (local_file))
701 url_free (proxy_url);
705 location_changed = (result == NEWLOCATION);
706 if (location_changed)
708 char *construced_newloc;
709 struct url *newloc_parsed;
711 assert (mynewloc != NULL);
716 /* The HTTP specs only allow absolute URLs to appear in
717 redirects, but a ton of boneheaded webservers and CGIs out
718 there break the rules and use relative URLs, and popular
719 browsers are lenient about this, so wget should be too. */
720 construced_newloc = uri_merge (url, mynewloc);
722 mynewloc = construced_newloc;
724 /* Now, see if this new location makes sense. */
725 newloc_parsed = url_parse (mynewloc, &up_error_code);
728 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
729 url_error (up_error_code));
737 /* Now mynewloc will become newloc_parsed->url, because if the
738 Location contained relative paths like .././something, we
739 don't want that propagating as url. */
741 mynewloc = xstrdup (newloc_parsed->url);
743 /* Check for max. number of redirections. */
744 if (++redirection_count > opt.max_redirect)
746 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
748 url_free (newloc_parsed);
761 /* If we're being redirected from POST, we don't want to POST
762 again. Many requests answer POST with a redirection to an
763 index page; that redirection is clearly a GET. We "suspend"
764 POST data for the duration of the redirections, and restore
765 it when we're done. */
766 if (!post_data_suspended)
776 register_download (u->url, local_file);
777 if (redirection_count && 0 != strcmp (origurl, u->url))
778 register_redirection (origurl, u->url);
780 register_html (u->url, local_file);
785 *file = local_file ? local_file : NULL;
787 xfree_null (local_file);
791 if (redirection_count)
810 /* Find the URLs in the file and call retrieve_url() for each of them.
811 If HTML is true, treat the file as HTML, and construct the URLs
814 If opt.recursive is set, call retrieve_tree() for each file. */
817 retrieve_from_file (const char *file, bool html, int *count)
820 struct urlpos *url_list, *cur_url;
822 url_list = (html ? get_urls_html (file, NULL, NULL)
823 : get_urls_file (file));
824 status = RETROK; /* Suppose everything is OK. */
825 *count = 0; /* Reset the URL count. */
827 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
829 char *filename = NULL, *new_file = NULL;
832 if (cur_url->ignore_when_downloading)
835 if (opt.quota && total_downloaded_bytes > opt.quota)
840 if ((opt.recursive || opt.page_requisites)
841 && (cur_url->url->scheme != SCHEME_FTP || getproxy (cur_url->url)))
843 int old_follow_ftp = opt.follow_ftp;
845 /* Turn opt.follow_ftp on in case of recursive FTP retrieval */
846 if (cur_url->url->scheme == SCHEME_FTP)
849 status = retrieve_tree (cur_url->url->url);
851 opt.follow_ftp = old_follow_ftp;
854 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt, opt.recursive);
856 if (filename && opt.delete_after && file_exists_p (filename))
859 Removing file due to --delete-after in retrieve_from_file():\n"));
860 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
861 if (unlink (filename))
862 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
866 xfree_null (new_file);
867 xfree_null (filename);
870 /* Free the linked list of URL-s. */
871 free_urlpos (url_list);
876 /* Print `giving up', or `retrying', depending on the impending
877 action. N1 and N2 are the attempt number and the attempt limit. */
879 printwhat (int n1, int n2)
881 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
884 /* If opt.wait or opt.waitretry are specified, and if certain
885 conditions are met, sleep the appropriate number of seconds. See
886 the documentation of --wait and --waitretry for more information.
888 COUNT is the count of current retrieval, beginning with 1. */
891 sleep_between_retrievals (int count)
893 static bool first_retrieval = true;
897 /* Don't sleep before the very first retrieval. */
898 first_retrieval = false;
902 if (opt.waitretry && count > 1)
904 /* If opt.waitretry is specified and this is a retry, wait for
905 COUNT-1 number of seconds, or for opt.waitretry seconds. */
906 if (count <= opt.waitretry)
909 xsleep (opt.waitretry);
913 if (!opt.random_wait || count > 1)
914 /* If random-wait is not specified, or if we are sleeping
915 between retries of the same download, sleep the fixed
920 /* Sleep a random amount of time averaging in opt.wait
921 seconds. The sleeping amount ranges from 0.5*opt.wait to
923 double waitsecs = (0.5 + random_float ()) * opt.wait;
924 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
925 opt.wait, waitsecs));
931 /* Free the linked list of urlpos. */
933 free_urlpos (struct urlpos *l)
937 struct urlpos *next = l->next;
940 xfree_null (l->local_name);
946 /* Rotate FNAME opt.backups times */
948 rotate_backups(const char *fname)
950 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
951 char *from = (char *)alloca (maxlen);
952 char *to = (char *)alloca (maxlen);
956 if (stat (fname, &sb) == 0)
957 if (S_ISREG (sb.st_mode) == 0)
960 for (i = opt.backups; i > 1; i--)
962 sprintf (from, "%s.%d", fname, i - 1);
963 sprintf (to, "%s.%d", fname, i);
967 sprintf (to, "%s.%d", fname, 1);
971 static bool no_proxy_match (const char *, const char **);
973 /* Return the URL of the proxy appropriate for url U. */
976 getproxy (struct url *u)
980 static char rewritten_storage[1024];
984 if (no_proxy_match (u->host, (const char **)opt.no_proxy))
990 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
994 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
998 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1000 case SCHEME_INVALID:
1003 if (!proxy || !*proxy)
1006 /* Handle shorthands. `rewritten_storage' is a kludge to allow
1007 getproxy() to return static storage. */
1008 rewritten_url = rewrite_shorthand_url (proxy);
1011 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
1012 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
1013 proxy = rewritten_storage;
1019 /* Returns true if URL would be downloaded through a proxy. */
1022 url_uses_proxy (const char *url)
1025 struct url *u = url_parse (url, NULL);
1028 ret = getproxy (u) != NULL;
1033 /* Should a host be accessed through proxy, concerning no_proxy? */
1035 no_proxy_match (const char *host, const char **no_proxy)
1040 return sufmatch (no_proxy, host);