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. */
31 #define USE_GNULIB_ALLOC
39 #endif /* HAVE_UNISTD_H */
57 /* Total size of downloaded files. Used to enforce quota. */
58 SUM_SIZE_INT total_downloaded_bytes;
60 /* Total download time in seconds. */
61 double total_download_time;
63 /* If non-NULL, the stream to which output should be written. This
64 stream is initialized when `-O' is used. */
67 /* Whether output_document is a regular file we can manipulate,
68 i.e. not `-' or a device file. */
69 bool output_stream_regular;
78 limit_bandwidth_reset (void)
83 /* Limit the bandwidth by pausing the download for an amount of time.
84 BYTES is the number of bytes received from the network, and TIMER
85 is the timer that started at the beginning of download. */
88 limit_bandwidth (wgint bytes, struct ptimer *timer)
90 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
93 limit_data.chunk_bytes += bytes;
95 /* Calculate the amount of time we expect downloading the chunk
96 should take. If in reality it took less time, sleep to
97 compensate for the difference. */
98 expected = (double) limit_data.chunk_bytes / opt.limit_rate;
100 if (expected > delta_t)
102 double slp = expected - delta_t + limit_data.sleep_adjust;
106 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
107 slp * 1000, number_to_static_string (limit_data.chunk_bytes),
111 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
112 slp * 1000, number_to_static_string (limit_data.chunk_bytes),
113 limit_data.sleep_adjust));
115 t0 = ptimer_read (timer);
117 t1 = ptimer_measure (timer);
119 /* Due to scheduling, we probably slept slightly longer (or
120 shorter) than desired. Calculate the difference between the
121 desired and the actual sleep, and adjust the next sleep by
123 limit_data.sleep_adjust = slp - (t1 - t0);
124 /* If sleep_adjust is very large, it's likely due to suspension
125 and not clock inaccuracy. Don't enforce those. */
126 if (limit_data.sleep_adjust > 0.5)
127 limit_data.sleep_adjust = 0.5;
128 else if (limit_data.sleep_adjust < -0.5)
129 limit_data.sleep_adjust = -0.5;
132 limit_data.chunk_bytes = 0;
133 limit_data.chunk_start = ptimer_read (timer);
137 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
140 /* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
141 amount of data and decrease SKIP. Increment *TOTAL by the amount
145 write_data (FILE *out, const char *buf, int bufsize, wgint *skip,
164 fwrite (buf, 1, bufsize, out);
167 /* Immediately flush the downloaded data. This should not hinder
168 performance: fast downloads will arrive in large 16K chunks
169 (which stdio would write out immediately anyway), and slow
170 downloads wouldn't be limited by disk speed. */
172 return !ferror (out);
175 /* Read the contents of file descriptor FD until it the connection
176 terminates or a read error occurs. The data is read in portions of
177 up to 16K and written to OUT as it arrives. If opt.verbose is set,
178 the progress is shown.
180 TOREAD is the amount of data expected to arrive, normally only used
181 by the progress gauge.
183 STARTPOS is the position from which the download starts, used by
184 the progress gauge. If QTYREAD is non-NULL, the value it points to
185 is incremented by the amount of data read from the network. If
186 QTYWRITTEN is non-NULL, the value it points to is incremented by
187 the amount of data written to disk. The time it took to download
188 the data is stored to ELAPSED.
190 The function exits and returns the amount of data read. In case of
191 error while reading data, -1 is returned. In case of error while
192 writing data, -2 is returned. */
195 fd_read_body (int fd, FILE *out, wgint toread, wgint startpos,
196 wgint *qtyread, wgint *qtywritten, double *elapsed, int flags)
200 static char dlbuf[16384];
201 int dlbufsize = sizeof (dlbuf);
203 struct ptimer *timer = NULL;
204 double last_successful_read_tm = 0;
206 /* The progress gauge, set according to the user preferences. */
207 void *progress = NULL;
209 /* Non-zero if the progress gauge is interactive, i.e. if it can
210 continually update the display. When true, smaller timeout
211 values are used so that the gauge can update the display when
212 data arrives slowly. */
213 bool progress_interactive = false;
215 bool exact = !!(flags & rb_read_exactly);
218 /* How much data we've read/written. */
220 wgint sum_written = 0;
222 if (flags & rb_skip_startpos)
227 /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
228 argument to progress_create because the indicator doesn't
229 (yet) know about "skipping" data. */
230 progress = progress_create (skip ? 0 : startpos, startpos + toread);
231 progress_interactive = progress_interactive_p (progress);
235 limit_bandwidth_reset ();
237 /* A timer is needed for tracking progress, for throttling, and for
238 tracking elapsed time. If either of these are requested, start
240 if (progress || opt.limit_rate || elapsed)
242 timer = ptimer_new ();
243 last_successful_read_tm = 0;
246 /* Use a smaller buffer for low requested bandwidths. For example,
247 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
248 data and then sleep for 8s. With buffer size equal to the limit,
249 we never have to sleep for more than one second. */
250 if (opt.limit_rate && opt.limit_rate < dlbufsize)
251 dlbufsize = opt.limit_rate;
253 /* Read from FD while there is data to read. Normally toread==0
254 means that it is unknown how much data is to arrive. However, if
255 EXACT is set, then toread==0 means what it says: that no data
257 while (!exact || (sum_read < toread))
259 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
260 double tmout = opt.read_timeout;
261 if (progress_interactive)
263 /* For interactive progress gauges, always specify a ~1s
264 timeout, so that the gauge can be updated regularly even
265 when the data arrives very slowly or stalls. */
267 if (opt.read_timeout)
270 waittm = ptimer_read (timer) - last_successful_read_tm;
271 if (waittm + tmout > opt.read_timeout)
273 /* Don't let total idle time exceed read timeout. */
274 tmout = opt.read_timeout - waittm;
277 /* We've already exceeded the timeout. */
278 ret = -1, errno = ETIMEDOUT;
284 ret = fd_read (fd, dlbuf, rdsize, tmout);
286 if (progress_interactive && ret < 0 && errno == ETIMEDOUT)
287 ret = 0; /* interactive timeout, handled above */
289 break; /* EOF or read error */
291 if (progress || opt.limit_rate)
293 ptimer_measure (timer);
295 last_successful_read_tm = ptimer_read (timer);
301 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
309 limit_bandwidth (ret, timer);
312 progress_update (progress, ret, ptimer_read (timer));
314 if (toread > 0 && !opt.quiet)
315 ws_percenttitle (100.0 *
316 (startpos + sum_read) / (startpos + toread));
324 progress_finish (progress, ptimer_read (timer));
327 *elapsed = ptimer_read (timer);
329 ptimer_destroy (timer);
332 *qtyread += sum_read;
334 *qtywritten += sum_written;
339 /* Read a hunk of data from FD, up until a terminator. The hunk is
340 limited by whatever the TERMINATOR callback chooses as its
341 terminator. For example, if terminator stops at newline, the hunk
342 will consist of a line of data; if terminator stops at two
343 newlines, it can be used to read the head of an HTTP response.
344 Upon determining the boundary, the function returns the data (up to
345 the terminator) in malloc-allocated storage.
347 In case of read error, NULL is returned. In case of EOF and no
348 data read, NULL is returned and errno set to 0. In case of having
349 read some data, but encountering EOF before seeing the terminator,
350 the data that has been read is returned, but it will (obviously)
351 not contain the terminator.
353 The TERMINATOR function is called with three arguments: the
354 beginning of the data read so far, the beginning of the current
355 block of peeked-at data, and the length of the current block.
356 Depending on its needs, the function is free to choose whether to
357 analyze all data or just the newly arrived data. If TERMINATOR
358 returns NULL, it means that the terminator has not been seen.
359 Otherwise it should return a pointer to the charactre immediately
360 following the terminator.
362 The idea is to be able to read a line of input, or otherwise a hunk
363 of text, such as the head of an HTTP request, without crossing the
364 boundary, so that the next call to fd_read etc. reads the data
365 after the hunk. To achieve that, this function does the following:
367 1. Peek at incoming data.
369 2. Determine whether the peeked data, along with the previously
370 read data, includes the terminator.
372 2a. If yes, read the data until the end of the terminator, and
375 2b. If no, read the peeked data and goto 1.
377 The function is careful to assume as little as possible about the
378 implementation of peeking. For example, every peek is followed by
379 a read. If the read returns a different amount of data, the
380 process is retried until all data arrives safely.
382 SIZEHINT is the buffer size sufficient to hold all the data in the
383 typical case (it is used as the initial buffer size). MAXSIZE is
384 the maximum amount of memory this function is allowed to allocate,
385 or 0 if no upper limit is to be enforced.
387 This function should be used as a building block for other
388 functions -- see fd_read_line as a simple example. */
391 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
393 long bufsize = sizehint;
394 char *hunk = xmalloc (bufsize);
395 int tail = 0; /* tail position in HUNK */
397 assert (maxsize >= bufsize);
402 int pklen, rdlen, remain;
404 /* First, peek at the available data. */
406 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
412 end = terminator (hunk, hunk + tail, pklen);
415 /* The data contains the terminator: we'll drain the data up
416 to the end of the terminator. */
417 remain = end - (hunk + tail);
418 assert (remain >= 0);
421 /* No more data needs to be read. */
425 if (bufsize - 1 < tail + remain)
427 bufsize = tail + remain + 1;
428 hunk = xrealloc (hunk, bufsize);
432 /* No terminator: simply read the data we know is (or should
436 /* Now, read the data. Note that we make no assumptions about
437 how much data we'll get. (Some TCP stacks are notorious for
438 read returning less data than the previous MSG_PEEK.) */
440 rdlen = fd_read (fd, hunk + tail, remain, 0);
453 /* EOF without anything having been read */
459 /* EOF seen: return the data we've read. */
462 if (end && rdlen == remain)
463 /* The terminator was seen and the remaining data drained --
464 we got what we came for. */
467 /* Keep looping until all the data arrives. */
469 if (tail == bufsize - 1)
471 /* Double the buffer size, but refuse to allocate more than
473 if (maxsize && bufsize >= maxsize)
480 if (maxsize && bufsize > maxsize)
482 hunk = xrealloc (hunk, bufsize);
488 line_terminator (const char *start, const char *peeked, int peeklen)
490 const char *p = memchr (peeked, '\n', peeklen);
492 /* p+1 because the line must include '\n' */
497 /* The maximum size of the single line we agree to accept. This is
498 not meant to impose an arbitrary limit, but to protect the user
499 from Wget slurping up available memory upon encountering malicious
500 or buggy server output. Define it to 0 to remove the limit. */
501 #define FD_READ_LINE_MAX 4096
503 /* Read one line from FD and return it. The line is allocated using
504 malloc, but is never larger than FD_READ_LINE_MAX.
506 If an error occurs, or if no data can be read, NULL is returned.
507 In the former case errno indicates the error condition, and in the
508 latter case, errno is NULL. */
511 fd_read_line (int fd)
513 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
516 /* Return a printed representation of the download rate, along with
517 the units appropriate for the download speed. */
520 retr_rate (wgint bytes, double secs)
523 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
526 double dlrate = calc_rate (bytes, secs, &units);
527 /* Use more digits for smaller numbers (regardless of unit used),
528 e.g. "1022", "247", "12.5", "2.38". */
529 sprintf (res, "%.*f %s",
530 dlrate >= 99.95 ? 0 : dlrate >= 9.995 ? 1 : 2,
531 dlrate, rate_names[units]);
536 /* Calculate the download rate and trim it as appropriate for the
537 speed. Appropriate means that if rate is greater than 1K/s,
538 kilobytes are used, and if rate is greater than 1MB/s, megabytes
541 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
545 calc_rate (wgint bytes, double secs, int *units)
553 /* If elapsed time is exactly zero, it means we're under the
554 resolution of the timer. This can easily happen on systems
555 that use time() for the timer. Since the interval lies between
556 0 and the timer's resolution, assume half the resolution. */
557 secs = ptimer_resolution () / 2.0;
559 dlrate = bytes / secs;
562 else if (dlrate < 1024.0 * 1024.0)
563 *units = 1, dlrate /= 1024.0;
564 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
565 *units = 2, dlrate /= (1024.0 * 1024.0);
567 /* Maybe someone will need this, one day. */
568 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
574 #define SUSPEND_POST_DATA do { \
575 post_data_suspended = true; \
576 saved_post_data = opt.post_data; \
577 saved_post_file_name = opt.post_file_name; \
578 opt.post_data = NULL; \
579 opt.post_file_name = NULL; \
582 #define RESTORE_POST_DATA do { \
583 if (post_data_suspended) \
585 opt.post_data = saved_post_data; \
586 opt.post_file_name = saved_post_file_name; \
587 post_data_suspended = false; \
591 static char *getproxy (struct url *);
593 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
596 /* #### This function should be rewritten so it doesn't return from
600 retrieve_url (const char *origurl, char **file, char **newloc,
601 const char *refurl, int *dt, bool recursive)
605 bool location_changed;
607 char *mynewloc, *proxy;
608 struct url *u, *proxy_url;
609 int up_error_code; /* url parse error code */
611 int redirection_count = 0;
613 bool post_data_suspended = false;
614 char *saved_post_data = NULL;
615 char *saved_post_file_name = NULL;
617 /* If dt is NULL, use local storage. */
623 url = xstrdup (origurl);
629 u = url_parse (url, &up_error_code);
632 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
638 refurl = opt.referer;
647 proxy = getproxy (u);
650 /* Parse the proxy URL. */
651 proxy_url = url_parse (proxy, &up_error_code);
654 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
655 proxy, url_error (up_error_code));
660 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
662 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
663 url_free (proxy_url);
670 if (u->scheme == SCHEME_HTTP
672 || u->scheme == SCHEME_HTTPS
674 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
676 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
678 else if (u->scheme == SCHEME_FTP)
680 /* If this is a redirection, temporarily turn off opt.ftp_glob
681 and opt.recursive, both being undesirable when following
683 bool oldrec = recursive, glob = opt.ftp_glob;
684 if (redirection_count)
685 oldrec = glob = false;
687 result = ftp_loop (u, dt, proxy_url, recursive, glob);
690 /* There is a possibility of having HTTP being redirected to
691 FTP. In these cases we must decide whether the text is HTML
692 according to the suffix. The HTML suffixes are `.html',
693 `.htm' and a few others, case-insensitive. */
694 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
696 if (has_html_suffix_p (local_file))
703 url_free (proxy_url);
707 location_changed = (result == NEWLOCATION);
708 if (location_changed)
710 char *construced_newloc;
711 struct url *newloc_parsed;
713 assert (mynewloc != NULL);
718 /* The HTTP specs only allow absolute URLs to appear in
719 redirects, but a ton of boneheaded webservers and CGIs out
720 there break the rules and use relative URLs, and popular
721 browsers are lenient about this, so wget should be too. */
722 construced_newloc = uri_merge (url, mynewloc);
724 mynewloc = construced_newloc;
726 /* Now, see if this new location makes sense. */
727 newloc_parsed = url_parse (mynewloc, &up_error_code);
730 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
731 url_error (up_error_code));
739 /* Now mynewloc will become newloc_parsed->url, because if the
740 Location contained relative paths like .././something, we
741 don't want that propagating as url. */
743 mynewloc = xstrdup (newloc_parsed->url);
745 /* Check for max. number of redirections. */
746 if (++redirection_count > opt.max_redirect)
748 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
750 url_free (newloc_parsed);
763 /* If we're being redirected from POST, we don't want to POST
764 again. Many requests answer POST with a redirection to an
765 index page; that redirection is clearly a GET. We "suspend"
766 POST data for the duration of the redirections, and restore
767 it when we're done. */
768 if (!post_data_suspended)
778 register_download (u->url, local_file);
779 if (redirection_count && 0 != strcmp (origurl, u->url))
780 register_redirection (origurl, u->url);
782 register_html (u->url, local_file);
787 *file = local_file ? local_file : NULL;
789 xfree_null (local_file);
793 if (redirection_count)
812 /* Find the URLs in the file and call retrieve_url() for each of them.
813 If HTML is true, treat the file as HTML, and construct the URLs
816 If opt.recursive is set, call retrieve_tree() for each file. */
819 retrieve_from_file (const char *file, bool html, int *count)
822 struct urlpos *url_list, *cur_url;
824 url_list = (html ? get_urls_html (file, NULL, NULL)
825 : get_urls_file (file));
826 status = RETROK; /* Suppose everything is OK. */
827 *count = 0; /* Reset the URL count. */
829 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
831 char *filename = NULL, *new_file = NULL;
834 if (cur_url->ignore_when_downloading)
837 if (opt.quota && total_downloaded_bytes > opt.quota)
842 if ((opt.recursive || opt.page_requisites)
843 && (cur_url->url->scheme != SCHEME_FTP || getproxy (cur_url->url)))
845 int old_follow_ftp = opt.follow_ftp;
847 /* Turn opt.follow_ftp on in case of recursive FTP retrieval */
848 if (cur_url->url->scheme == SCHEME_FTP)
851 status = retrieve_tree (cur_url->url->url);
853 opt.follow_ftp = old_follow_ftp;
856 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt, opt.recursive);
858 if (filename && opt.delete_after && file_exists_p (filename))
861 Removing file due to --delete-after in retrieve_from_file():\n"));
862 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
863 if (unlink (filename))
864 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
868 xfree_null (new_file);
869 xfree_null (filename);
872 /* Free the linked list of URL-s. */
873 free_urlpos (url_list);
878 /* Print `giving up', or `retrying', depending on the impending
879 action. N1 and N2 are the attempt number and the attempt limit. */
881 printwhat (int n1, int n2)
883 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
886 /* If opt.wait or opt.waitretry are specified, and if certain
887 conditions are met, sleep the appropriate number of seconds. See
888 the documentation of --wait and --waitretry for more information.
890 COUNT is the count of current retrieval, beginning with 1. */
893 sleep_between_retrievals (int count)
895 static bool first_retrieval = true;
899 /* Don't sleep before the very first retrieval. */
900 first_retrieval = false;
904 if (opt.waitretry && count > 1)
906 /* If opt.waitretry is specified and this is a retry, wait for
907 COUNT-1 number of seconds, or for opt.waitretry seconds. */
908 if (count <= opt.waitretry)
911 xsleep (opt.waitretry);
915 if (!opt.random_wait || count > 1)
916 /* If random-wait is not specified, or if we are sleeping
917 between retries of the same download, sleep the fixed
922 /* Sleep a random amount of time averaging in opt.wait
923 seconds. The sleeping amount ranges from 0.5*opt.wait to
925 double waitsecs = (0.5 + random_float ()) * opt.wait;
926 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
927 opt.wait, waitsecs));
933 /* Free the linked list of urlpos. */
935 free_urlpos (struct urlpos *l)
939 struct urlpos *next = l->next;
942 xfree_null (l->local_name);
948 /* Rotate FNAME opt.backups times */
950 rotate_backups(const char *fname)
952 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
953 char *from = (char *)alloca (maxlen);
954 char *to = (char *)alloca (maxlen);
958 if (stat (fname, &sb) == 0)
959 if (S_ISREG (sb.st_mode) == 0)
962 for (i = opt.backups; i > 1; i--)
964 sprintf (from, "%s.%d", fname, i - 1);
965 sprintf (to, "%s.%d", fname, i);
969 sprintf (to, "%s.%d", fname, 1);
973 static bool no_proxy_match (const char *, const char **);
975 /* Return the URL of the proxy appropriate for url U. */
978 getproxy (struct url *u)
982 static char rewritten_storage[1024];
986 if (no_proxy_match (u->host, (const char **)opt.no_proxy))
992 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
996 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1000 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1002 case SCHEME_INVALID:
1005 if (!proxy || !*proxy)
1008 /* Handle shorthands. `rewritten_storage' is a kludge to allow
1009 getproxy() to return static storage. */
1010 rewritten_url = rewrite_shorthand_url (proxy);
1013 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
1014 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
1015 proxy = rewritten_storage;
1021 /* Returns true if URL would be downloaded through a proxy. */
1024 url_uses_proxy (const char *url)
1027 struct url *u = url_parse (url, NULL);
1030 ret = getproxy (u) != NULL;
1035 /* Should a host be accessed through proxy, concerning no_proxy? */
1037 no_proxy_match (const char *host, const char **no_proxy)
1042 return sufmatch (no_proxy, host);