2 Copyright (C) 1996-2006 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 3 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, see <http://www.gnu.org/licenses/>.
19 In addition, as a special exception, the Free Software Foundation
20 gives permission to link the code of its release of Wget with the
21 OpenSSL project's "OpenSSL" library (or with modified versions of it
22 that use the same license as the "OpenSSL" library), and distribute
23 the linked executables. You must obey the GNU General Public License
24 in all respects for all of the code used other than "OpenSSL". If you
25 modify this file, you may extend this exception to your version of the
26 file, but you are not obligated to do so. If you do not wish to do
27 so, delete this exception statement from your version. */
35 #endif /* HAVE_UNISTD_H */
54 /* Total size of downloaded files. Used to enforce quota. */
55 SUM_SIZE_INT total_downloaded_bytes;
57 /* Total download time in seconds. */
58 double total_download_time;
60 /* If non-NULL, the stream to which output should be written. This
61 stream is initialized when `-O' is used. */
64 /* Whether output_document is a regular file we can manipulate,
65 i.e. not `-' or a device file. */
66 bool output_stream_regular;
75 limit_bandwidth_reset (void)
80 /* Limit the bandwidth by pausing the download for an amount of time.
81 BYTES is the number of bytes received from the network, and TIMER
82 is the timer that started at the beginning of download. */
85 limit_bandwidth (wgint bytes, struct ptimer *timer)
87 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
90 limit_data.chunk_bytes += bytes;
92 /* Calculate the amount of time we expect downloading the chunk
93 should take. If in reality it took less time, sleep to
94 compensate for the difference. */
95 expected = (double) limit_data.chunk_bytes / opt.limit_rate;
97 if (expected > delta_t)
99 double slp = expected - delta_t + limit_data.sleep_adjust;
103 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
104 slp * 1000, number_to_static_string (limit_data.chunk_bytes),
108 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
109 slp * 1000, number_to_static_string (limit_data.chunk_bytes),
110 limit_data.sleep_adjust));
112 t0 = ptimer_read (timer);
114 t1 = ptimer_measure (timer);
116 /* Due to scheduling, we probably slept slightly longer (or
117 shorter) than desired. Calculate the difference between the
118 desired and the actual sleep, and adjust the next sleep by
120 limit_data.sleep_adjust = slp - (t1 - t0);
121 /* If sleep_adjust is very large, it's likely due to suspension
122 and not clock inaccuracy. Don't enforce those. */
123 if (limit_data.sleep_adjust > 0.5)
124 limit_data.sleep_adjust = 0.5;
125 else if (limit_data.sleep_adjust < -0.5)
126 limit_data.sleep_adjust = -0.5;
129 limit_data.chunk_bytes = 0;
130 limit_data.chunk_start = ptimer_read (timer);
134 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
137 /* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
138 amount of data and decrease SKIP. Increment *TOTAL by the amount
142 write_data (FILE *out, const char *buf, int bufsize, wgint *skip,
161 fwrite (buf, 1, bufsize, out);
164 /* Immediately flush the downloaded data. This should not hinder
165 performance: fast downloads will arrive in large 16K chunks
166 (which stdio would write out immediately anyway), and slow
167 downloads wouldn't be limited by disk speed. */
169 return !ferror (out);
172 /* Read the contents of file descriptor FD until it the connection
173 terminates or a read error occurs. The data is read in portions of
174 up to 16K and written to OUT as it arrives. If opt.verbose is set,
175 the progress is shown.
177 TOREAD is the amount of data expected to arrive, normally only used
178 by the progress gauge.
180 STARTPOS is the position from which the download starts, used by
181 the progress gauge. If QTYREAD is non-NULL, the value it points to
182 is incremented by the amount of data read from the network. If
183 QTYWRITTEN is non-NULL, the value it points to is incremented by
184 the amount of data written to disk. The time it took to download
185 the data is stored to ELAPSED.
187 The function exits and returns the amount of data read. In case of
188 error while reading data, -1 is returned. In case of error while
189 writing data, -2 is returned. */
192 fd_read_body (int fd, FILE *out, wgint toread, wgint startpos,
193 wgint *qtyread, wgint *qtywritten, double *elapsed, int flags)
197 static char dlbuf[16384];
198 int dlbufsize = sizeof (dlbuf);
200 struct ptimer *timer = NULL;
201 double last_successful_read_tm = 0;
203 /* The progress gauge, set according to the user preferences. */
204 void *progress = NULL;
206 /* Non-zero if the progress gauge is interactive, i.e. if it can
207 continually update the display. When true, smaller timeout
208 values are used so that the gauge can update the display when
209 data arrives slowly. */
210 bool progress_interactive = false;
212 bool exact = !!(flags & rb_read_exactly);
215 /* How much data we've read/written. */
217 wgint sum_written = 0;
219 if (flags & rb_skip_startpos)
224 /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
225 argument to progress_create because the indicator doesn't
226 (yet) know about "skipping" data. */
227 progress = progress_create (skip ? 0 : startpos, startpos + toread);
228 progress_interactive = progress_interactive_p (progress);
232 limit_bandwidth_reset ();
234 /* A timer is needed for tracking progress, for throttling, and for
235 tracking elapsed time. If either of these are requested, start
237 if (progress || opt.limit_rate || elapsed)
239 timer = ptimer_new ();
240 last_successful_read_tm = 0;
243 /* Use a smaller buffer for low requested bandwidths. For example,
244 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
245 data and then sleep for 8s. With buffer size equal to the limit,
246 we never have to sleep for more than one second. */
247 if (opt.limit_rate && opt.limit_rate < dlbufsize)
248 dlbufsize = opt.limit_rate;
250 /* Read from FD while there is data to read. Normally toread==0
251 means that it is unknown how much data is to arrive. However, if
252 EXACT is set, then toread==0 means what it says: that no data
254 while (!exact || (sum_read < toread))
256 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
257 double tmout = opt.read_timeout;
258 if (progress_interactive)
260 /* For interactive progress gauges, always specify a ~1s
261 timeout, so that the gauge can be updated regularly even
262 when the data arrives very slowly or stalls. */
264 if (opt.read_timeout)
267 waittm = ptimer_read (timer) - last_successful_read_tm;
268 if (waittm + tmout > opt.read_timeout)
270 /* Don't let total idle time exceed read timeout. */
271 tmout = opt.read_timeout - waittm;
274 /* We've already exceeded the timeout. */
275 ret = -1, errno = ETIMEDOUT;
281 ret = fd_read (fd, dlbuf, rdsize, tmout);
283 if (progress_interactive && ret < 0 && errno == ETIMEDOUT)
284 ret = 0; /* interactive timeout, handled above */
286 break; /* EOF or read error */
288 if (progress || opt.limit_rate)
290 ptimer_measure (timer);
292 last_successful_read_tm = ptimer_read (timer);
298 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
306 limit_bandwidth (ret, timer);
309 progress_update (progress, ret, ptimer_read (timer));
311 if (toread > 0 && !opt.quiet)
312 ws_percenttitle (100.0 *
313 (startpos + sum_read) / (startpos + toread));
321 progress_finish (progress, ptimer_read (timer));
324 *elapsed = ptimer_read (timer);
326 ptimer_destroy (timer);
329 *qtyread += sum_read;
331 *qtywritten += sum_written;
336 /* Read a hunk of data from FD, up until a terminator. The hunk is
337 limited by whatever the TERMINATOR callback chooses as its
338 terminator. For example, if terminator stops at newline, the hunk
339 will consist of a line of data; if terminator stops at two
340 newlines, it can be used to read the head of an HTTP response.
341 Upon determining the boundary, the function returns the data (up to
342 the terminator) in malloc-allocated storage.
344 In case of read error, NULL is returned. In case of EOF and no
345 data read, NULL is returned and errno set to 0. In case of having
346 read some data, but encountering EOF before seeing the terminator,
347 the data that has been read is returned, but it will (obviously)
348 not contain the terminator.
350 The TERMINATOR function is called with three arguments: the
351 beginning of the data read so far, the beginning of the current
352 block of peeked-at data, and the length of the current block.
353 Depending on its needs, the function is free to choose whether to
354 analyze all data or just the newly arrived data. If TERMINATOR
355 returns NULL, it means that the terminator has not been seen.
356 Otherwise it should return a pointer to the charactre immediately
357 following the terminator.
359 The idea is to be able to read a line of input, or otherwise a hunk
360 of text, such as the head of an HTTP request, without crossing the
361 boundary, so that the next call to fd_read etc. reads the data
362 after the hunk. To achieve that, this function does the following:
364 1. Peek at incoming data.
366 2. Determine whether the peeked data, along with the previously
367 read data, includes the terminator.
369 2a. If yes, read the data until the end of the terminator, and
372 2b. If no, read the peeked data and goto 1.
374 The function is careful to assume as little as possible about the
375 implementation of peeking. For example, every peek is followed by
376 a read. If the read returns a different amount of data, the
377 process is retried until all data arrives safely.
379 SIZEHINT is the buffer size sufficient to hold all the data in the
380 typical case (it is used as the initial buffer size). MAXSIZE is
381 the maximum amount of memory this function is allowed to allocate,
382 or 0 if no upper limit is to be enforced.
384 This function should be used as a building block for other
385 functions -- see fd_read_line as a simple example. */
388 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
390 long bufsize = sizehint;
391 char *hunk = xmalloc (bufsize);
392 int tail = 0; /* tail position in HUNK */
394 assert (maxsize >= bufsize);
399 int pklen, rdlen, remain;
401 /* First, peek at the available data. */
403 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
409 end = terminator (hunk, hunk + tail, pklen);
412 /* The data contains the terminator: we'll drain the data up
413 to the end of the terminator. */
414 remain = end - (hunk + tail);
415 assert (remain >= 0);
418 /* No more data needs to be read. */
422 if (bufsize - 1 < tail + remain)
424 bufsize = tail + remain + 1;
425 hunk = xrealloc (hunk, bufsize);
429 /* No terminator: simply read the data we know is (or should
433 /* Now, read the data. Note that we make no assumptions about
434 how much data we'll get. (Some TCP stacks are notorious for
435 read returning less data than the previous MSG_PEEK.) */
437 rdlen = fd_read (fd, hunk + tail, remain, 0);
450 /* EOF without anything having been read */
456 /* EOF seen: return the data we've read. */
459 if (end && rdlen == remain)
460 /* The terminator was seen and the remaining data drained --
461 we got what we came for. */
464 /* Keep looping until all the data arrives. */
466 if (tail == bufsize - 1)
468 /* Double the buffer size, but refuse to allocate more than
470 if (maxsize && bufsize >= maxsize)
477 if (maxsize && bufsize > maxsize)
479 hunk = xrealloc (hunk, bufsize);
485 line_terminator (const char *start, const char *peeked, int peeklen)
487 const char *p = memchr (peeked, '\n', peeklen);
489 /* p+1 because the line must include '\n' */
494 /* The maximum size of the single line we agree to accept. This is
495 not meant to impose an arbitrary limit, but to protect the user
496 from Wget slurping up available memory upon encountering malicious
497 or buggy server output. Define it to 0 to remove the limit. */
498 #define FD_READ_LINE_MAX 4096
500 /* Read one line from FD and return it. The line is allocated using
501 malloc, but is never larger than FD_READ_LINE_MAX.
503 If an error occurs, or if no data can be read, NULL is returned.
504 In the former case errno indicates the error condition, and in the
505 latter case, errno is NULL. */
508 fd_read_line (int fd)
510 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
513 /* Return a printed representation of the download rate, along with
514 the units appropriate for the download speed. */
517 retr_rate (wgint bytes, double secs)
520 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
523 double dlrate = calc_rate (bytes, secs, &units);
524 /* Use more digits for smaller numbers (regardless of unit used),
525 e.g. "1022", "247", "12.5", "2.38". */
526 sprintf (res, "%.*f %s",
527 dlrate >= 99.95 ? 0 : dlrate >= 9.995 ? 1 : 2,
528 dlrate, rate_names[units]);
533 /* Calculate the download rate and trim it as appropriate for the
534 speed. Appropriate means that if rate is greater than 1K/s,
535 kilobytes are used, and if rate is greater than 1MB/s, megabytes
538 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
542 calc_rate (wgint bytes, double secs, int *units)
550 /* If elapsed time is exactly zero, it means we're under the
551 resolution of the timer. This can easily happen on systems
552 that use time() for the timer. Since the interval lies between
553 0 and the timer's resolution, assume half the resolution. */
554 secs = ptimer_resolution () / 2.0;
556 dlrate = bytes / secs;
559 else if (dlrate < 1024.0 * 1024.0)
560 *units = 1, dlrate /= 1024.0;
561 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
562 *units = 2, dlrate /= (1024.0 * 1024.0);
564 /* Maybe someone will need this, one day. */
565 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
570 /* Maximum number of allowed redirections. 20 was chosen as a
571 "reasonable" value, which is low enough to not cause havoc, yet
572 high enough to guarantee that normal retrievals will not be hurt by
575 #define MAX_REDIRECTIONS 20
577 #define SUSPEND_POST_DATA do { \
578 post_data_suspended = true; \
579 saved_post_data = opt.post_data; \
580 saved_post_file_name = opt.post_file_name; \
581 opt.post_data = NULL; \
582 opt.post_file_name = NULL; \
585 #define RESTORE_POST_DATA do { \
586 if (post_data_suspended) \
588 opt.post_data = saved_post_data; \
589 opt.post_file_name = saved_post_file_name; \
590 post_data_suspended = false; \
594 static char *getproxy (struct url *);
596 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
599 /* #### This function should be rewritten so it doesn't return from
603 retrieve_url (const char *origurl, char **file, char **newloc,
604 const char *refurl, int *dt, bool recursive)
608 bool location_changed;
610 char *mynewloc, *proxy;
611 struct url *u, *proxy_url;
612 int up_error_code; /* url parse error code */
614 int redirection_count = 0;
616 bool post_data_suspended = false;
617 char *saved_post_data = NULL;
618 char *saved_post_file_name = NULL;
620 /* If dt is NULL, use local storage. */
626 url = xstrdup (origurl);
632 u = url_parse (url, &up_error_code);
635 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
641 refurl = opt.referer;
650 proxy = getproxy (u);
653 /* Parse the proxy URL. */
654 proxy_url = url_parse (proxy, &up_error_code);
657 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
658 proxy, url_error (up_error_code));
663 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
665 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
666 url_free (proxy_url);
673 if (u->scheme == SCHEME_HTTP
675 || u->scheme == SCHEME_HTTPS
677 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
679 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
681 else if (u->scheme == SCHEME_FTP)
683 /* If this is a redirection, temporarily turn off opt.ftp_glob
684 and opt.recursive, both being undesirable when following
686 bool oldrec = recursive, glob = opt.ftp_glob;
687 if (redirection_count)
688 oldrec = glob = false;
690 result = ftp_loop (u, dt, proxy_url, recursive, glob);
693 /* There is a possibility of having HTTP being redirected to
694 FTP. In these cases we must decide whether the text is HTML
695 according to the suffix. The HTML suffixes are `.html',
696 `.htm' and a few others, case-insensitive. */
697 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
699 if (has_html_suffix_p (local_file))
706 url_free (proxy_url);
710 location_changed = (result == NEWLOCATION);
711 if (location_changed)
713 char *construced_newloc;
714 struct url *newloc_parsed;
716 assert (mynewloc != NULL);
721 /* The HTTP specs only allow absolute URLs to appear in
722 redirects, but a ton of boneheaded webservers and CGIs out
723 there break the rules and use relative URLs, and popular
724 browsers are lenient about this, so wget should be too. */
725 construced_newloc = uri_merge (url, mynewloc);
727 mynewloc = construced_newloc;
729 /* Now, see if this new location makes sense. */
730 newloc_parsed = url_parse (mynewloc, &up_error_code);
733 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
734 url_error (up_error_code));
742 /* Now mynewloc will become newloc_parsed->url, because if the
743 Location contained relative paths like .././something, we
744 don't want that propagating as url. */
746 mynewloc = xstrdup (newloc_parsed->url);
748 /* Check for max. number of redirections. */
749 if (++redirection_count > MAX_REDIRECTIONS)
751 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
753 url_free (newloc_parsed);
766 /* If we're being redirected from POST, we don't want to POST
767 again. Many requests answer POST with a redirection to an
768 index page; that redirection is clearly a GET. We "suspend"
769 POST data for the duration of the redirections, and restore
770 it when we're done. */
771 if (!post_data_suspended)
781 register_download (u->url, local_file);
782 if (redirection_count && 0 != strcmp (origurl, u->url))
783 register_redirection (origurl, u->url);
785 register_html (u->url, local_file);
790 *file = local_file ? local_file : NULL;
792 xfree_null (local_file);
796 if (redirection_count)
815 /* Find the URLs in the file and call retrieve_url() for each of them.
816 If HTML is true, treat the file as HTML, and construct the URLs
819 If opt.recursive is set, call retrieve_tree() for each file. */
822 retrieve_from_file (const char *file, bool html, int *count)
825 struct urlpos *url_list, *cur_url;
827 url_list = (html ? get_urls_html (file, NULL, NULL)
828 : get_urls_file (file));
829 status = RETROK; /* Suppose everything is OK. */
830 *count = 0; /* Reset the URL count. */
832 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
834 char *filename = NULL, *new_file = NULL;
837 if (cur_url->ignore_when_downloading)
840 if (opt.quota && total_downloaded_bytes > opt.quota)
845 if ((opt.recursive || opt.page_requisites)
846 && (cur_url->url->scheme != SCHEME_FTP || getproxy (cur_url->url)))
848 int old_follow_ftp = opt.follow_ftp;
850 /* Turn opt.follow_ftp on in case of recursive FTP retrieval */
851 if (cur_url->url->scheme == SCHEME_FTP)
854 status = retrieve_tree (cur_url->url->url);
856 opt.follow_ftp = old_follow_ftp;
859 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt, opt.recursive);
861 if (filename && opt.delete_after && file_exists_p (filename))
864 Removing file due to --delete-after in retrieve_from_file():\n"));
865 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
866 if (unlink (filename))
867 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
871 xfree_null (new_file);
872 xfree_null (filename);
875 /* Free the linked list of URL-s. */
876 free_urlpos (url_list);
881 /* Print `giving up', or `retrying', depending on the impending
882 action. N1 and N2 are the attempt number and the attempt limit. */
884 printwhat (int n1, int n2)
886 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
889 /* If opt.wait or opt.waitretry are specified, and if certain
890 conditions are met, sleep the appropriate number of seconds. See
891 the documentation of --wait and --waitretry for more information.
893 COUNT is the count of current retrieval, beginning with 1. */
896 sleep_between_retrievals (int count)
898 static bool first_retrieval = true;
902 /* Don't sleep before the very first retrieval. */
903 first_retrieval = false;
907 if (opt.waitretry && count > 1)
909 /* If opt.waitretry is specified and this is a retry, wait for
910 COUNT-1 number of seconds, or for opt.waitretry seconds. */
911 if (count <= opt.waitretry)
914 xsleep (opt.waitretry);
918 if (!opt.random_wait || count > 1)
919 /* If random-wait is not specified, or if we are sleeping
920 between retries of the same download, sleep the fixed
925 /* Sleep a random amount of time averaging in opt.wait
926 seconds. The sleeping amount ranges from 0.5*opt.wait to
928 double waitsecs = (0.5 + random_float ()) * opt.wait;
929 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
930 opt.wait, waitsecs));
936 /* Free the linked list of urlpos. */
938 free_urlpos (struct urlpos *l)
942 struct urlpos *next = l->next;
945 xfree_null (l->local_name);
951 /* Rotate FNAME opt.backups times */
953 rotate_backups(const char *fname)
955 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
956 char *from = (char *)alloca (maxlen);
957 char *to = (char *)alloca (maxlen);
961 if (stat (fname, &sb) == 0)
962 if (S_ISREG (sb.st_mode) == 0)
965 for (i = opt.backups; i > 1; i--)
967 sprintf (from, "%s.%d", fname, i - 1);
968 sprintf (to, "%s.%d", fname, i);
972 sprintf (to, "%s.%d", fname, 1);
976 static bool no_proxy_match (const char *, const char **);
978 /* Return the URL of the proxy appropriate for url U. */
981 getproxy (struct url *u)
985 static char rewritten_storage[1024];
989 if (no_proxy_match (u->host, (const char **)opt.no_proxy))
995 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
999 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1003 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1005 case SCHEME_INVALID:
1008 if (!proxy || !*proxy)
1011 /* Handle shorthands. `rewritten_storage' is a kludge to allow
1012 getproxy() to return static storage. */
1013 rewritten_url = rewrite_shorthand_url (proxy);
1016 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
1017 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
1018 proxy = rewritten_storage;
1024 /* Returns true if URL would be downloaded through a proxy. */
1027 url_uses_proxy (const char *url)
1030 struct url *u = url_parse (url, NULL);
1033 ret = getproxy (u) != NULL;
1038 /* Should a host be accessed through proxy, concerning no_proxy? */
1040 no_proxy_match (const char *host, const char **no_proxy)
1045 return sufmatch (no_proxy, host);