2 Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
36 #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 milliseconds. */
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)
78 limit_data.chunk_bytes = 0;
79 limit_data.chunk_start = 0;
80 limit_data.sleep_adjust = 0;
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 = 1000.0 * 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, number_to_static_string (limit_data.chunk_bytes),
111 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
112 slp, 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 > 500)
127 limit_data.sleep_adjust = 500;
128 else if (limit_data.sleep_adjust < -500)
129 limit_data.sleep_adjust = -500;
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 (in milliseconds) 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) / 1000;
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 terminator
340 is whatever the TERMINATOR function determines it to be; for
341 example, it can be a line of data, or the head of an HTTP response.
342 The function returns the data read allocated with malloc.
344 In case of error, NULL is returned. In case of EOF and no data
345 read, NULL is returned and errno set to 0. In case of EOF with
346 data having been read, the data is returned, but it will
347 (obviously) not contain the terminator.
349 The idea is to be able to read a line of input, or otherwise a hunk
350 of text, such as the head of an HTTP request, without crossing the
351 boundary, so that the next call to fd_read etc. reads the data
352 after the hunk. To achieve that, this function does the following:
354 1. Peek at available data.
356 2. Determine whether the peeked data, along with the previously
357 read data, includes the terminator.
359 2a. If yes, read the data until the end of the terminator, and
362 2b. If no, read the peeked data and goto 1.
364 The function is careful to assume as little as possible about the
365 implementation of peeking. For example, every peek is followed by
366 a read. If the read returns a different amount of data, the
367 process is retried until all data arrives safely.
369 SIZEHINT is the buffer size sufficient to hold all the data in the
370 typical case (it is used as the initial buffer size). MAXSIZE is
371 the maximum amount of memory this function is allowed to allocate,
372 or 0 if no upper limit is to be enforced.
374 This function should be used as a building block for other
375 functions -- see fd_read_line as a simple example. */
378 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
380 long bufsize = sizehint;
381 char *hunk = xmalloc (bufsize);
382 int tail = 0; /* tail position in HUNK */
384 assert (maxsize >= bufsize);
389 int pklen, rdlen, remain;
391 /* First, peek at the available data. */
393 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
399 end = terminator (hunk, tail, pklen);
402 /* The data contains the terminator: we'll drain the data up
403 to the end of the terminator. */
404 remain = end - (hunk + tail);
407 /* No more data needs to be read. */
411 if (bufsize - 1 < tail + remain)
413 bufsize = tail + remain + 1;
414 hunk = xrealloc (hunk, bufsize);
418 /* No terminator: simply read the data we know is (or should
422 /* Now, read the data. Note that we make no assumptions about
423 how much data we'll get. (Some TCP stacks are notorious for
424 read returning less data than the previous MSG_PEEK.) */
426 rdlen = fd_read (fd, hunk + tail, remain, 0);
439 /* EOF without anything having been read */
445 /* EOF seen: return the data we've read. */
448 if (end && rdlen == remain)
449 /* The terminator was seen and the remaining data drained --
450 we got what we came for. */
453 /* Keep looping until all the data arrives. */
455 if (tail == bufsize - 1)
457 /* Double the buffer size, but refuse to allocate more than
459 if (maxsize && bufsize >= maxsize)
466 if (maxsize && bufsize > maxsize)
468 hunk = xrealloc (hunk, bufsize);
474 line_terminator (const char *hunk, int oldlen, int peeklen)
476 const char *p = memchr (hunk + oldlen, '\n', peeklen);
478 /* p+1 because we want the line to include '\n' */
483 /* The maximum size of the single line we agree to accept. This is
484 not meant to impose an arbitrary limit, but to protect the user
485 from Wget slurping up available memory upon encountering malicious
486 or buggy server output. Define it to 0 to remove the limit. */
487 #define FD_READ_LINE_MAX 4096
489 /* Read one line from FD and return it. The line is allocated using
490 malloc, but is never larger than FD_READ_LINE_MAX.
492 If an error occurs, or if no data can be read, NULL is returned.
493 In the former case errno indicates the error condition, and in the
494 latter case, errno is NULL. */
497 fd_read_line (int fd)
499 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
502 /* Return a printed representation of the download rate, along with
503 the units appropriate for the download speed. */
506 retr_rate (wgint bytes, double msecs)
509 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
512 double dlrate = calc_rate (bytes, msecs, &units);
513 sprintf (res, "%.2f %s", dlrate, rate_names[units]);
518 /* Calculate the download rate and trim it as appropriate for the
519 speed. Appropriate means that if rate is greater than 1K/s,
520 kilobytes are used, and if rate is greater than 1MB/s, megabytes
523 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
526 calc_rate (wgint bytes, double msecs, int *units)
534 /* If elapsed time is exactly zero, it means we're under the
535 resolution of the timer. This can easily happen on systems
536 that use time() for the timer. Since the interval lies between
537 0 and the timer's resolution, assume half the resolution. */
538 msecs = ptimer_resolution () / 2.0;
540 dlrate = 1000.0 * bytes / msecs;
543 else if (dlrate < 1024.0 * 1024.0)
544 *units = 1, dlrate /= 1024.0;
545 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
546 *units = 2, dlrate /= (1024.0 * 1024.0);
548 /* Maybe someone will need this, one day. */
549 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
554 /* Maximum number of allowed redirections. 20 was chosen as a
555 "reasonable" value, which is low enough to not cause havoc, yet
556 high enough to guarantee that normal retrievals will not be hurt by
559 #define MAX_REDIRECTIONS 20
561 #define SUSPEND_POST_DATA do { \
562 post_data_suspended = true; \
563 saved_post_data = opt.post_data; \
564 saved_post_file_name = opt.post_file_name; \
565 opt.post_data = NULL; \
566 opt.post_file_name = NULL; \
569 #define RESTORE_POST_DATA do { \
570 if (post_data_suspended) \
572 opt.post_data = saved_post_data; \
573 opt.post_file_name = saved_post_file_name; \
574 post_data_suspended = false; \
578 static char *getproxy (struct url *);
580 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
583 /* #### This function should be rewritten so it doesn't return from
587 retrieve_url (const char *origurl, char **file, char **newloc,
588 const char *refurl, int *dt)
592 bool location_changed;
594 char *mynewloc, *proxy;
595 struct url *u, *proxy_url;
596 int up_error_code; /* url parse error code */
598 int redirection_count = 0;
600 bool post_data_suspended = false;
601 char *saved_post_data = NULL;
602 char *saved_post_file_name = NULL;
604 /* If dt is NULL, use local storage. */
610 url = xstrdup (origurl);
616 u = url_parse (url, &up_error_code);
619 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
625 refurl = opt.referer;
634 proxy = getproxy (u);
637 /* Parse the proxy URL. */
638 proxy_url = url_parse (proxy, &up_error_code);
641 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
642 proxy, url_error (up_error_code));
647 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
649 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
650 url_free (proxy_url);
657 if (u->scheme == SCHEME_HTTP
659 || u->scheme == SCHEME_HTTPS
661 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
663 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
665 else if (u->scheme == SCHEME_FTP)
667 /* If this is a redirection, temporarily turn off opt.ftp_glob
668 and opt.recursive, both being undesirable when following
670 bool oldrec = opt.recursive, oldglob = opt.ftp_glob;
671 if (redirection_count)
672 opt.recursive = opt.ftp_glob = false;
674 result = ftp_loop (u, dt, proxy_url);
675 opt.recursive = oldrec;
676 opt.ftp_glob = oldglob;
678 /* There is a possibility of having HTTP being redirected to
679 FTP. In these cases we must decide whether the text is HTML
680 according to the suffix. The HTML suffixes are `.html',
681 `.htm' and a few others, case-insensitive. */
682 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
684 if (has_html_suffix_p (local_file))
691 url_free (proxy_url);
695 location_changed = (result == NEWLOCATION);
696 if (location_changed)
698 char *construced_newloc;
699 struct url *newloc_parsed;
701 assert (mynewloc != NULL);
706 /* The HTTP specs only allow absolute URLs to appear in
707 redirects, but a ton of boneheaded webservers and CGIs out
708 there break the rules and use relative URLs, and popular
709 browsers are lenient about this, so wget should be too. */
710 construced_newloc = uri_merge (url, mynewloc);
712 mynewloc = construced_newloc;
714 /* Now, see if this new location makes sense. */
715 newloc_parsed = url_parse (mynewloc, &up_error_code);
718 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
719 url_error (up_error_code));
727 /* Now mynewloc will become newloc_parsed->url, because if the
728 Location contained relative paths like .././something, we
729 don't want that propagating as url. */
731 mynewloc = xstrdup (newloc_parsed->url);
733 /* Check for max. number of redirections. */
734 if (++redirection_count > MAX_REDIRECTIONS)
736 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
738 url_free (newloc_parsed);
751 /* If we're being redirected from POST, we don't want to POST
752 again. Many requests answer POST with a redirection to an
753 index page; that redirection is clearly a GET. We "suspend"
754 POST data for the duration of the redirections, and restore
755 it when we're done. */
756 if (!post_data_suspended)
766 register_download (u->url, local_file);
767 if (redirection_count && 0 != strcmp (origurl, u->url))
768 register_redirection (origurl, u->url);
770 register_html (u->url, local_file);
775 *file = local_file ? local_file : NULL;
777 xfree_null (local_file);
781 if (redirection_count)
800 /* Find the URLs in the file and call retrieve_url() for each of them.
801 If HTML is true, treat the file as HTML, and construct the URLs
804 If opt.recursive is set, call retrieve_tree() for each file. */
807 retrieve_from_file (const char *file, bool html, int *count)
810 struct urlpos *url_list, *cur_url;
812 url_list = (html ? get_urls_html (file, NULL, NULL)
813 : get_urls_file (file));
814 status = RETROK; /* Suppose everything is OK. */
815 *count = 0; /* Reset the URL count. */
817 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
819 char *filename = NULL, *new_file = NULL;
822 if (cur_url->ignore_when_downloading)
825 if (opt.quota && total_downloaded_bytes > opt.quota)
830 if ((opt.recursive || opt.page_requisites)
831 && cur_url->url->scheme != SCHEME_FTP)
832 status = retrieve_tree (cur_url->url->url);
834 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
836 if (filename && opt.delete_after && file_exists_p (filename))
839 Removing file due to --delete-after in retrieve_from_file():\n"));
840 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
841 if (unlink (filename))
842 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
846 xfree_null (new_file);
847 xfree_null (filename);
850 /* Free the linked list of URL-s. */
851 free_urlpos (url_list);
856 /* Print `giving up', or `retrying', depending on the impending
857 action. N1 and N2 are the attempt number and the attempt limit. */
859 printwhat (int n1, int n2)
861 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
864 /* If opt.wait or opt.waitretry are specified, and if certain
865 conditions are met, sleep the appropriate number of seconds. See
866 the documentation of --wait and --waitretry for more information.
868 COUNT is the count of current retrieval, beginning with 1. */
871 sleep_between_retrievals (int count)
873 static bool first_retrieval = true;
877 /* Don't sleep before the very first retrieval. */
878 first_retrieval = false;
882 if (opt.waitretry && count > 1)
884 /* If opt.waitretry is specified and this is a retry, wait for
885 COUNT-1 number of seconds, or for opt.waitretry seconds. */
886 if (count <= opt.waitretry)
889 xsleep (opt.waitretry);
893 if (!opt.random_wait || count > 1)
894 /* If random-wait is not specified, or if we are sleeping
895 between retries of the same download, sleep the fixed
900 /* Sleep a random amount of time averaging in opt.wait
901 seconds. The sleeping amount ranges from 0 to
902 opt.wait*2, inclusive. */
903 double waitsecs = 2 * opt.wait * random_float ();
904 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
905 opt.wait, waitsecs));
911 /* Free the linked list of urlpos. */
913 free_urlpos (struct urlpos *l)
917 struct urlpos *next = l->next;
920 xfree_null (l->local_name);
926 /* Rotate FNAME opt.backups times */
928 rotate_backups(const char *fname)
930 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
931 char *from = (char *)alloca (maxlen);
932 char *to = (char *)alloca (maxlen);
936 if (stat (fname, &sb) == 0)
937 if (S_ISREG (sb.st_mode) == 0)
940 for (i = opt.backups; i > 1; i--)
942 sprintf (from, "%s.%d", fname, i - 1);
943 sprintf (to, "%s.%d", fname, i);
947 sprintf (to, "%s.%d", fname, 1);
951 static bool no_proxy_match (const char *, const char **);
953 /* Return the URL of the proxy appropriate for url U. */
956 getproxy (struct url *u)
960 static char rewritten_storage[1024];
964 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
970 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
974 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
978 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
983 if (!proxy || !*proxy)
986 /* Handle shorthands. `rewritten_storage' is a kludge to allow
987 getproxy() to return static storage. */
988 rewritten_url = rewrite_shorthand_url (proxy);
991 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
992 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
993 proxy = rewritten_storage;
999 /* Should a host be accessed through proxy, concerning no_proxy? */
1001 no_proxy_match (const char *host, const char **no_proxy)
1006 return !sufmatch (no_proxy, host);