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. */
34 #include <sys/types.h>
37 #endif /* HAVE_UNISTD_H */
43 #endif /* HAVE_STRING_H */
63 /* Total size of downloaded files. Used to enforce quota. */
64 LARGE_INT total_downloaded_bytes;
66 /* If non-NULL, the stream to which output should be written. This
67 stream is initialized when `-O' is used. */
70 /* Whether output_document is a regular file we can manipulate,
71 i.e. not `-' or a device file. */
72 int output_stream_regular;
81 limit_bandwidth_reset (void)
83 limit_data.chunk_bytes = 0;
84 limit_data.chunk_start = 0;
87 /* Limit the bandwidth by pausing the download for an amount of time.
88 BYTES is the number of bytes received from the network, and TIMER
89 is the timer that started at the beginning of download. */
92 limit_bandwidth (wgint bytes, struct ptimer *timer)
94 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
97 limit_data.chunk_bytes += bytes;
99 /* Calculate the amount of time we expect downloading the chunk
100 should take. If in reality it took less time, sleep to
101 compensate for the difference. */
102 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
104 if (expected > delta_t)
106 double slp = expected - delta_t + limit_data.sleep_adjust;
110 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
111 slp, number_to_static_string (limit_data.chunk_bytes),
115 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
116 slp, number_to_static_string (limit_data.chunk_bytes),
117 limit_data.sleep_adjust));
119 t0 = ptimer_read (timer);
121 t1 = ptimer_measure (timer);
123 /* Due to scheduling, we probably slept slightly longer (or
124 shorter) than desired. Calculate the difference between the
125 desired and the actual sleep, and adjust the next sleep by
127 limit_data.sleep_adjust = slp - (t1 - t0);
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 (in milliseconds) 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 int progress_interactive = 0;
213 int 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) / 1000;
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 terminator
338 is whatever the TERMINATOR function determines it to be; for
339 example, it can be a line of data, or the head of an HTTP response.
340 The function returns the data read allocated with malloc.
342 In case of error, NULL is returned. In case of EOF and no data
343 read, NULL is returned and errno set to 0. In case of EOF with
344 data having been read, the data is returned, but it will
345 (obviously) not contain the terminator.
347 The idea is to be able to read a line of input, or otherwise a hunk
348 of text, such as the head of an HTTP request, without crossing the
349 boundary, so that the next call to fd_read etc. reads the data
350 after the hunk. To achieve that, this function does the following:
352 1. Peek at available data.
354 2. Determine whether the peeked data, along with the previously
355 read data, includes the terminator.
357 2a. If yes, read the data until the end of the terminator, and
360 2b. If no, read the peeked data and goto 1.
362 The function is careful to assume as little as possible about the
363 implementation of peeking. For example, every peek is followed by
364 a read. If the read returns a different amount of data, the
365 process is retried until all data arrives safely.
367 SIZEHINT is the buffer size sufficient to hold all the data in the
368 typical case (it is used as the initial buffer size). MAXSIZE is
369 the maximum amount of memory this function is allowed to allocate,
370 or 0 if no upper limit is to be enforced.
372 This function should be used as a building block for other
373 functions -- see fd_read_line as a simple example. */
376 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
378 long bufsize = sizehint;
379 char *hunk = xmalloc (bufsize);
380 int tail = 0; /* tail position in HUNK */
382 assert (maxsize >= bufsize);
387 int pklen, rdlen, remain;
389 /* First, peek at the available data. */
391 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1.0);
397 end = terminator (hunk, tail, pklen);
400 /* The data contains the terminator: we'll drain the data up
401 to the end of the terminator. */
402 remain = end - (hunk + tail);
405 /* No more data needs to be read. */
409 if (bufsize - 1 < tail + remain)
411 bufsize = tail + remain + 1;
412 hunk = xrealloc (hunk, bufsize);
416 /* No terminator: simply read the data we know is (or should
420 /* Now, read the data. Note that we make no assumptions about
421 how much data we'll get. (Some TCP stacks are notorious for
422 read returning less data than the previous MSG_PEEK.) */
424 rdlen = fd_read (fd, hunk + tail, remain, 0.0);
437 /* EOF without anything having been read */
443 /* EOF seen: return the data we've read. */
446 if (end && rdlen == remain)
447 /* The terminator was seen and the remaining data drained --
448 we got what we came for. */
451 /* Keep looping until all the data arrives. */
453 if (tail == bufsize - 1)
455 /* Double the buffer size, but refuse to allocate more than
457 if (maxsize && bufsize >= maxsize)
464 if (maxsize && bufsize > maxsize)
466 hunk = xrealloc (hunk, bufsize);
472 line_terminator (const char *hunk, int oldlen, int peeklen)
474 const char *p = memchr (hunk + oldlen, '\n', peeklen);
476 /* p+1 because we want the line to include '\n' */
481 /* The maximum size of the single line we agree to accept. This is
482 not meant to impose an arbitrary limit, but to protect the user
483 from Wget slurping up available memory upon encountering malicious
484 or buggy server output. Define it to 0 to remove the limit. */
485 #define FD_READ_LINE_MAX 4096
487 /* Read one line from FD and return it. The line is allocated using
488 malloc, but is never larger than FD_READ_LINE_MAX.
490 If an error occurs, or if no data can be read, NULL is returned.
491 In the former case errno indicates the error condition, and in the
492 latter case, errno is NULL. */
495 fd_read_line (int fd)
497 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
500 /* Return a printed representation of the download rate, as
501 appropriate for the speed. If PAD is non-zero, strings will be
502 padded to the width of 7 characters (xxxx.xx). */
504 retr_rate (wgint bytes, double msecs, int pad)
507 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
510 double dlrate = calc_rate (bytes, msecs, &units);
511 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
516 /* Calculate the download rate and trim it as appropriate for the
517 speed. Appropriate means that if rate is greater than 1K/s,
518 kilobytes are used, and if rate is greater than 1MB/s, megabytes
521 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
524 calc_rate (wgint bytes, double msecs, int *units)
532 /* If elapsed time is exactly zero, it means we're under the
533 resolution of the timer. This can easily happen on systems
534 that use time() for the timer. Since the interval lies between
535 0 and the timer's resolution, assume half the resolution. */
536 msecs = ptimer_resolution () / 2.0;
538 dlrate = 1000.0 * bytes / msecs;
541 else if (dlrate < 1024.0 * 1024.0)
542 *units = 1, dlrate /= 1024.0;
543 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
544 *units = 2, dlrate /= (1024.0 * 1024.0);
546 /* Maybe someone will need this, one day. */
547 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
552 /* Maximum number of allowed redirections. 20 was chosen as a
553 "reasonable" value, which is low enough to not cause havoc, yet
554 high enough to guarantee that normal retrievals will not be hurt by
557 #define MAX_REDIRECTIONS 20
559 #define SUSPEND_POST_DATA do { \
560 post_data_suspended = 1; \
561 saved_post_data = opt.post_data; \
562 saved_post_file_name = opt.post_file_name; \
563 opt.post_data = NULL; \
564 opt.post_file_name = NULL; \
567 #define RESTORE_POST_DATA do { \
568 if (post_data_suspended) \
570 opt.post_data = saved_post_data; \
571 opt.post_file_name = saved_post_file_name; \
572 post_data_suspended = 0; \
576 static char *getproxy PARAMS ((struct url *));
578 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
581 /* #### This function should be rewritten so it doesn't return from
585 retrieve_url (const char *origurl, char **file, char **newloc,
586 const char *refurl, int *dt)
590 int location_changed, dummy;
591 char *mynewloc, *proxy;
592 struct url *u, *proxy_url;
593 int up_error_code; /* url parse error code */
595 int redirection_count = 0;
597 int post_data_suspended = 0;
598 char *saved_post_data = NULL;
599 char *saved_post_file_name = NULL;
601 /* If dt is NULL, use local storage. */
607 url = xstrdup (origurl);
613 u = url_parse (url, &up_error_code);
616 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
622 refurl = opt.referer;
631 proxy = getproxy (u);
634 /* Parse the proxy URL. */
635 proxy_url = url_parse (proxy, &up_error_code);
638 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
639 proxy, url_error (up_error_code));
644 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
646 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
647 url_free (proxy_url);
654 if (u->scheme == SCHEME_HTTP
656 || u->scheme == SCHEME_HTTPS
658 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
660 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
662 else if (u->scheme == SCHEME_FTP)
664 /* If this is a redirection, we must not allow recursive FTP
665 retrieval, so we save recursion to oldrec, and restore it
667 int oldrec = opt.recursive;
668 if (redirection_count)
670 result = ftp_loop (u, dt, proxy_url);
671 opt.recursive = oldrec;
673 /* There is a possibility of having HTTP being redirected to
674 FTP. In these cases we must decide whether the text is HTML
675 according to the suffix. The HTML suffixes are `.html',
676 `.htm' and a few others, case-insensitive. */
677 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
679 if (has_html_suffix_p (local_file))
686 url_free (proxy_url);
690 location_changed = (result == NEWLOCATION);
691 if (location_changed)
693 char *construced_newloc;
694 struct url *newloc_parsed;
696 assert (mynewloc != NULL);
701 /* The HTTP specs only allow absolute URLs to appear in
702 redirects, but a ton of boneheaded webservers and CGIs out
703 there break the rules and use relative URLs, and popular
704 browsers are lenient about this, so wget should be too. */
705 construced_newloc = uri_merge (url, mynewloc);
707 mynewloc = construced_newloc;
709 /* Now, see if this new location makes sense. */
710 newloc_parsed = url_parse (mynewloc, &up_error_code);
713 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
714 url_error (up_error_code));
722 /* Now mynewloc will become newloc_parsed->url, because if the
723 Location contained relative paths like .././something, we
724 don't want that propagating as url. */
726 mynewloc = xstrdup (newloc_parsed->url);
728 /* Check for max. number of redirections. */
729 if (++redirection_count > MAX_REDIRECTIONS)
731 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
733 url_free (newloc_parsed);
746 /* If we're being redirected from POST, we don't want to POST
747 again. Many requests answer POST with a redirection to an
748 index page; that redirection is clearly a GET. We "suspend"
749 POST data for the duration of the redirections, and restore
750 it when we're done. */
751 if (!post_data_suspended)
761 register_download (u->url, local_file);
762 if (redirection_count && 0 != strcmp (origurl, u->url))
763 register_redirection (origurl, u->url);
765 register_html (u->url, local_file);
770 *file = local_file ? local_file : NULL;
772 xfree_null (local_file);
776 if (redirection_count)
795 /* Find the URLs in the file and call retrieve_url() for each of
796 them. If HTML is non-zero, treat the file as HTML, and construct
797 the URLs accordingly.
799 If opt.recursive is set, call retrieve_tree() for each file. */
802 retrieve_from_file (const char *file, int html, int *count)
805 struct urlpos *url_list, *cur_url;
807 url_list = (html ? get_urls_html (file, NULL, NULL)
808 : get_urls_file (file));
809 status = RETROK; /* Suppose everything is OK. */
810 *count = 0; /* Reset the URL count. */
812 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
814 char *filename = NULL, *new_file = NULL;
817 if (cur_url->ignore_when_downloading)
820 if (opt.quota && total_downloaded_bytes > opt.quota)
825 if ((opt.recursive || opt.page_requisites)
826 && cur_url->url->scheme != SCHEME_FTP)
827 status = retrieve_tree (cur_url->url->url);
829 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
831 if (filename && opt.delete_after && file_exists_p (filename))
834 Removing file due to --delete-after in retrieve_from_file():\n"));
835 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
836 if (unlink (filename))
837 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
841 xfree_null (new_file);
842 xfree_null (filename);
845 /* Free the linked list of URL-s. */
846 free_urlpos (url_list);
851 /* Print `giving up', or `retrying', depending on the impending
852 action. N1 and N2 are the attempt number and the attempt limit. */
854 printwhat (int n1, int n2)
856 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
859 /* If opt.wait or opt.waitretry are specified, and if certain
860 conditions are met, sleep the appropriate number of seconds. See
861 the documentation of --wait and --waitretry for more information.
863 COUNT is the count of current retrieval, beginning with 1. */
866 sleep_between_retrievals (int count)
868 static int first_retrieval = 1;
872 /* Don't sleep before the very first retrieval. */
877 if (opt.waitretry && count > 1)
879 /* If opt.waitretry is specified and this is a retry, wait for
880 COUNT-1 number of seconds, or for opt.waitretry seconds. */
881 if (count <= opt.waitretry)
882 xsleep (count - 1.0);
884 xsleep (opt.waitretry);
888 if (!opt.random_wait || count > 1)
889 /* If random-wait is not specified, or if we are sleeping
890 between retries of the same download, sleep the fixed
895 /* Sleep a random amount of time averaging in opt.wait
896 seconds. The sleeping amount ranges from 0 to
897 opt.wait*2, inclusive. */
898 double waitsecs = 2 * opt.wait * random_float ();
899 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
900 opt.wait, waitsecs));
906 /* Free the linked list of urlpos. */
908 free_urlpos (struct urlpos *l)
912 struct urlpos *next = l->next;
915 xfree_null (l->local_name);
921 /* Rotate FNAME opt.backups times */
923 rotate_backups(const char *fname)
925 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
926 char *from = (char *)alloca (maxlen);
927 char *to = (char *)alloca (maxlen);
931 if (stat (fname, &sb) == 0)
932 if (S_ISREG (sb.st_mode) == 0)
935 for (i = opt.backups; i > 1; i--)
937 sprintf (from, "%s.%d", fname, i - 1);
938 sprintf (to, "%s.%d", fname, i);
942 sprintf (to, "%s.%d", fname, 1);
946 static int no_proxy_match PARAMS ((const char *, const char **));
948 /* Return the URL of the proxy appropriate for url U. */
951 getproxy (struct url *u)
955 static char rewritten_storage[1024];
959 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
965 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
969 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
973 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
978 if (!proxy || !*proxy)
981 /* Handle shorthands. `rewritten_storage' is a kludge to allow
982 getproxy() to return static storage. */
983 rewritten_url = rewrite_shorthand_url (proxy);
986 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
987 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
988 proxy = rewritten_storage;
994 /* Should a host be accessed through proxy, concerning no_proxy? */
996 no_proxy_match (const char *host, const char **no_proxy)
1001 return !sufmatch (no_proxy, host);