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;
85 limit_data.sleep_adjust = 0;
88 /* Limit the bandwidth by pausing the download for an amount of time.
89 BYTES is the number of bytes received from the network, and TIMER
90 is the timer that started at the beginning of download. */
93 limit_bandwidth (wgint bytes, struct ptimer *timer)
95 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
98 limit_data.chunk_bytes += bytes;
100 /* Calculate the amount of time we expect downloading the chunk
101 should take. If in reality it took less time, sleep to
102 compensate for the difference. */
103 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
105 if (expected > delta_t)
107 double slp = expected - delta_t + limit_data.sleep_adjust;
111 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
112 slp, number_to_static_string (limit_data.chunk_bytes),
116 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
117 slp, number_to_static_string (limit_data.chunk_bytes),
118 limit_data.sleep_adjust));
120 t0 = ptimer_read (timer);
122 t1 = ptimer_measure (timer);
124 /* Due to scheduling, we probably slept slightly longer (or
125 shorter) than desired. Calculate the difference between the
126 desired and the actual sleep, and adjust the next sleep by
128 limit_data.sleep_adjust = slp - (t1 - t0);
129 /* If sleep_adjust is very large, it's likely due to suspension
130 and not clock inaccuracy. Don't enforce those. */
131 if (limit_data.sleep_adjust > 500)
132 limit_data.sleep_adjust = 500;
133 else if (limit_data.sleep_adjust < -500)
134 limit_data.sleep_adjust = -500;
137 limit_data.chunk_bytes = 0;
138 limit_data.chunk_start = ptimer_read (timer);
142 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
145 /* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
146 amount of data and decrease SKIP. Increment *TOTAL by the amount
150 write_data (FILE *out, const char *buf, int bufsize, wgint *skip,
169 fwrite (buf, 1, bufsize, out);
172 /* Immediately flush the downloaded data. This should not hinder
173 performance: fast downloads will arrive in large 16K chunks
174 (which stdio would write out immediately anyway), and slow
175 downloads wouldn't be limited by disk speed. */
177 return !ferror (out);
180 /* Read the contents of file descriptor FD until it the connection
181 terminates or a read error occurs. The data is read in portions of
182 up to 16K and written to OUT as it arrives. If opt.verbose is set,
183 the progress is shown.
185 TOREAD is the amount of data expected to arrive, normally only used
186 by the progress gauge.
188 STARTPOS is the position from which the download starts, used by
189 the progress gauge. If QTYREAD is non-NULL, the value it points to
190 is incremented by the amount of data read from the network. If
191 QTYWRITTEN is non-NULL, the value it points to is incremented by
192 the amount of data written to disk. The time it took to download
193 the data (in milliseconds) is stored to ELAPSED.
195 The function exits and returns the amount of data read. In case of
196 error while reading data, -1 is returned. In case of error while
197 writing data, -2 is returned. */
200 fd_read_body (int fd, FILE *out, wgint toread, wgint startpos,
201 wgint *qtyread, wgint *qtywritten, double *elapsed, int flags)
205 static char dlbuf[16384];
206 int dlbufsize = sizeof (dlbuf);
208 struct ptimer *timer = NULL;
209 double last_successful_read_tm = 0;
211 /* The progress gauge, set according to the user preferences. */
212 void *progress = NULL;
214 /* Non-zero if the progress gauge is interactive, i.e. if it can
215 continually update the display. When true, smaller timeout
216 values are used so that the gauge can update the display when
217 data arrives slowly. */
218 int progress_interactive = 0;
220 int exact = flags & rb_read_exactly;
223 /* How much data we've read/written. */
225 wgint sum_written = 0;
227 if (flags & rb_skip_startpos)
232 /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
233 argument to progress_create because the indicator doesn't
234 (yet) know about "skipping" data. */
235 progress = progress_create (skip ? 0 : startpos, startpos + toread);
236 progress_interactive = progress_interactive_p (progress);
240 limit_bandwidth_reset ();
242 /* A timer is needed for tracking progress, for throttling, and for
243 tracking elapsed time. If either of these are requested, start
245 if (progress || opt.limit_rate || elapsed)
247 timer = ptimer_new ();
248 last_successful_read_tm = 0;
251 /* Use a smaller buffer for low requested bandwidths. For example,
252 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
253 data and then sleep for 8s. With buffer size equal to the limit,
254 we never have to sleep for more than one second. */
255 if (opt.limit_rate && opt.limit_rate < dlbufsize)
256 dlbufsize = opt.limit_rate;
258 /* Read from FD while there is data to read. Normally toread==0
259 means that it is unknown how much data is to arrive. However, if
260 EXACT is set, then toread==0 means what it says: that no data
262 while (!exact || (sum_read < toread))
264 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
265 double tmout = opt.read_timeout;
266 if (progress_interactive)
268 /* For interactive progress gauges, always specify a ~1s
269 timeout, so that the gauge can be updated regularly even
270 when the data arrives very slowly or stalls. */
272 if (opt.read_timeout)
275 waittm = (ptimer_read (timer) - last_successful_read_tm) / 1000;
276 if (waittm + tmout > opt.read_timeout)
278 /* Don't let total idle time exceed read timeout. */
279 tmout = opt.read_timeout - waittm;
282 /* We've already exceeded the timeout. */
283 ret = -1, errno = ETIMEDOUT;
289 ret = fd_read (fd, dlbuf, rdsize, tmout);
291 if (progress_interactive && ret < 0 && errno == ETIMEDOUT)
292 ret = 0; /* interactive timeout, handled above */
294 break; /* EOF or read error */
296 if (progress || opt.limit_rate)
298 ptimer_measure (timer);
300 last_successful_read_tm = ptimer_read (timer);
306 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
314 limit_bandwidth (ret, timer);
317 progress_update (progress, ret, ptimer_read (timer));
319 if (toread > 0 && !opt.quiet)
320 ws_percenttitle (100.0 *
321 (startpos + sum_read) / (startpos + toread));
329 progress_finish (progress, ptimer_read (timer));
332 *elapsed = ptimer_read (timer);
334 ptimer_destroy (timer);
337 *qtyread += sum_read;
339 *qtywritten += sum_written;
344 /* Read a hunk of data from FD, up until a terminator. The terminator
345 is whatever the TERMINATOR function determines it to be; for
346 example, it can be a line of data, or the head of an HTTP response.
347 The function returns the data read allocated with malloc.
349 In case of error, NULL is returned. In case of EOF and no data
350 read, NULL is returned and errno set to 0. In case of EOF with
351 data having been read, the data is returned, but it will
352 (obviously) not contain the terminator.
354 The idea is to be able to read a line of input, or otherwise a hunk
355 of text, such as the head of an HTTP request, without crossing the
356 boundary, so that the next call to fd_read etc. reads the data
357 after the hunk. To achieve that, this function does the following:
359 1. Peek at available data.
361 2. Determine whether the peeked data, along with the previously
362 read data, includes the terminator.
364 2a. If yes, read the data until the end of the terminator, and
367 2b. If no, read the peeked data and goto 1.
369 The function is careful to assume as little as possible about the
370 implementation of peeking. For example, every peek is followed by
371 a read. If the read returns a different amount of data, the
372 process is retried until all data arrives safely.
374 SIZEHINT is the buffer size sufficient to hold all the data in the
375 typical case (it is used as the initial buffer size). MAXSIZE is
376 the maximum amount of memory this function is allowed to allocate,
377 or 0 if no upper limit is to be enforced.
379 This function should be used as a building block for other
380 functions -- see fd_read_line as a simple example. */
383 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
385 long bufsize = sizehint;
386 char *hunk = xmalloc (bufsize);
387 int tail = 0; /* tail position in HUNK */
389 assert (maxsize >= bufsize);
394 int pklen, rdlen, remain;
396 /* First, peek at the available data. */
398 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1.0);
404 end = terminator (hunk, tail, pklen);
407 /* The data contains the terminator: we'll drain the data up
408 to the end of the terminator. */
409 remain = end - (hunk + tail);
412 /* No more data needs to be read. */
416 if (bufsize - 1 < tail + remain)
418 bufsize = tail + remain + 1;
419 hunk = xrealloc (hunk, bufsize);
423 /* No terminator: simply read the data we know is (or should
427 /* Now, read the data. Note that we make no assumptions about
428 how much data we'll get. (Some TCP stacks are notorious for
429 read returning less data than the previous MSG_PEEK.) */
431 rdlen = fd_read (fd, hunk + tail, remain, 0.0);
444 /* EOF without anything having been read */
450 /* EOF seen: return the data we've read. */
453 if (end && rdlen == remain)
454 /* The terminator was seen and the remaining data drained --
455 we got what we came for. */
458 /* Keep looping until all the data arrives. */
460 if (tail == bufsize - 1)
462 /* Double the buffer size, but refuse to allocate more than
464 if (maxsize && bufsize >= maxsize)
471 if (maxsize && bufsize > maxsize)
473 hunk = xrealloc (hunk, bufsize);
479 line_terminator (const char *hunk, int oldlen, int peeklen)
481 const char *p = memchr (hunk + oldlen, '\n', peeklen);
483 /* p+1 because we want the line to include '\n' */
488 /* The maximum size of the single line we agree to accept. This is
489 not meant to impose an arbitrary limit, but to protect the user
490 from Wget slurping up available memory upon encountering malicious
491 or buggy server output. Define it to 0 to remove the limit. */
492 #define FD_READ_LINE_MAX 4096
494 /* Read one line from FD and return it. The line is allocated using
495 malloc, but is never larger than FD_READ_LINE_MAX.
497 If an error occurs, or if no data can be read, NULL is returned.
498 In the former case errno indicates the error condition, and in the
499 latter case, errno is NULL. */
502 fd_read_line (int fd)
504 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
507 /* Return a printed representation of the download rate, as
508 appropriate for the speed. If PAD is non-zero, strings will be
509 padded to the width of 7 characters (xxxx.xx). */
511 retr_rate (wgint bytes, double msecs, int pad)
514 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
517 double dlrate = calc_rate (bytes, msecs, &units);
518 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
523 /* Calculate the download rate and trim it as appropriate for the
524 speed. Appropriate means that if rate is greater than 1K/s,
525 kilobytes are used, and if rate is greater than 1MB/s, megabytes
528 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
531 calc_rate (wgint bytes, double msecs, int *units)
539 /* If elapsed time is exactly zero, it means we're under the
540 resolution of the timer. This can easily happen on systems
541 that use time() for the timer. Since the interval lies between
542 0 and the timer's resolution, assume half the resolution. */
543 msecs = ptimer_resolution () / 2.0;
545 dlrate = 1000.0 * bytes / msecs;
548 else if (dlrate < 1024.0 * 1024.0)
549 *units = 1, dlrate /= 1024.0;
550 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
551 *units = 2, dlrate /= (1024.0 * 1024.0);
553 /* Maybe someone will need this, one day. */
554 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
559 /* Maximum number of allowed redirections. 20 was chosen as a
560 "reasonable" value, which is low enough to not cause havoc, yet
561 high enough to guarantee that normal retrievals will not be hurt by
564 #define MAX_REDIRECTIONS 20
566 #define SUSPEND_POST_DATA do { \
567 post_data_suspended = 1; \
568 saved_post_data = opt.post_data; \
569 saved_post_file_name = opt.post_file_name; \
570 opt.post_data = NULL; \
571 opt.post_file_name = NULL; \
574 #define RESTORE_POST_DATA do { \
575 if (post_data_suspended) \
577 opt.post_data = saved_post_data; \
578 opt.post_file_name = saved_post_file_name; \
579 post_data_suspended = 0; \
583 static char *getproxy PARAMS ((struct url *));
585 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
588 /* #### This function should be rewritten so it doesn't return from
592 retrieve_url (const char *origurl, char **file, char **newloc,
593 const char *refurl, int *dt)
597 int location_changed, dummy;
598 char *mynewloc, *proxy;
599 struct url *u, *proxy_url;
600 int up_error_code; /* url parse error code */
602 int redirection_count = 0;
604 int post_data_suspended = 0;
605 char *saved_post_data = NULL;
606 char *saved_post_file_name = NULL;
608 /* If dt is NULL, use local storage. */
614 url = xstrdup (origurl);
620 u = url_parse (url, &up_error_code);
623 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
629 refurl = opt.referer;
638 proxy = getproxy (u);
641 /* Parse the proxy URL. */
642 proxy_url = url_parse (proxy, &up_error_code);
645 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
646 proxy, url_error (up_error_code));
651 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
653 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
654 url_free (proxy_url);
661 if (u->scheme == SCHEME_HTTP
663 || u->scheme == SCHEME_HTTPS
665 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
667 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
669 else if (u->scheme == SCHEME_FTP)
671 /* If this is a redirection, we must not allow recursive FTP
672 retrieval, so we save recursion to oldrec, and restore it
674 int oldrec = opt.recursive;
675 if (redirection_count)
677 result = ftp_loop (u, dt, proxy_url);
678 opt.recursive = oldrec;
680 /* There is a possibility of having HTTP being redirected to
681 FTP. In these cases we must decide whether the text is HTML
682 according to the suffix. The HTML suffixes are `.html',
683 `.htm' and a few others, case-insensitive. */
684 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
686 if (has_html_suffix_p (local_file))
693 url_free (proxy_url);
697 location_changed = (result == NEWLOCATION);
698 if (location_changed)
700 char *construced_newloc;
701 struct url *newloc_parsed;
703 assert (mynewloc != NULL);
708 /* The HTTP specs only allow absolute URLs to appear in
709 redirects, but a ton of boneheaded webservers and CGIs out
710 there break the rules and use relative URLs, and popular
711 browsers are lenient about this, so wget should be too. */
712 construced_newloc = uri_merge (url, mynewloc);
714 mynewloc = construced_newloc;
716 /* Now, see if this new location makes sense. */
717 newloc_parsed = url_parse (mynewloc, &up_error_code);
720 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
721 url_error (up_error_code));
729 /* Now mynewloc will become newloc_parsed->url, because if the
730 Location contained relative paths like .././something, we
731 don't want that propagating as url. */
733 mynewloc = xstrdup (newloc_parsed->url);
735 /* Check for max. number of redirections. */
736 if (++redirection_count > MAX_REDIRECTIONS)
738 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
740 url_free (newloc_parsed);
753 /* If we're being redirected from POST, we don't want to POST
754 again. Many requests answer POST with a redirection to an
755 index page; that redirection is clearly a GET. We "suspend"
756 POST data for the duration of the redirections, and restore
757 it when we're done. */
758 if (!post_data_suspended)
768 register_download (u->url, local_file);
769 if (redirection_count && 0 != strcmp (origurl, u->url))
770 register_redirection (origurl, u->url);
772 register_html (u->url, local_file);
777 *file = local_file ? local_file : NULL;
779 xfree_null (local_file);
783 if (redirection_count)
802 /* Find the URLs in the file and call retrieve_url() for each of
803 them. If HTML is non-zero, treat the file as HTML, and construct
804 the URLs accordingly.
806 If opt.recursive is set, call retrieve_tree() for each file. */
809 retrieve_from_file (const char *file, int html, int *count)
812 struct urlpos *url_list, *cur_url;
814 url_list = (html ? get_urls_html (file, NULL, NULL)
815 : get_urls_file (file));
816 status = RETROK; /* Suppose everything is OK. */
817 *count = 0; /* Reset the URL count. */
819 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
821 char *filename = NULL, *new_file = NULL;
824 if (cur_url->ignore_when_downloading)
827 if (opt.quota && total_downloaded_bytes > opt.quota)
832 if ((opt.recursive || opt.page_requisites)
833 && cur_url->url->scheme != SCHEME_FTP)
834 status = retrieve_tree (cur_url->url->url);
836 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
838 if (filename && opt.delete_after && file_exists_p (filename))
841 Removing file due to --delete-after in retrieve_from_file():\n"));
842 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
843 if (unlink (filename))
844 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
848 xfree_null (new_file);
849 xfree_null (filename);
852 /* Free the linked list of URL-s. */
853 free_urlpos (url_list);
858 /* Print `giving up', or `retrying', depending on the impending
859 action. N1 and N2 are the attempt number and the attempt limit. */
861 printwhat (int n1, int n2)
863 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
866 /* If opt.wait or opt.waitretry are specified, and if certain
867 conditions are met, sleep the appropriate number of seconds. See
868 the documentation of --wait and --waitretry for more information.
870 COUNT is the count of current retrieval, beginning with 1. */
873 sleep_between_retrievals (int count)
875 static int first_retrieval = 1;
879 /* Don't sleep before the very first retrieval. */
884 if (opt.waitretry && count > 1)
886 /* If opt.waitretry is specified and this is a retry, wait for
887 COUNT-1 number of seconds, or for opt.waitretry seconds. */
888 if (count <= opt.waitretry)
889 xsleep (count - 1.0);
891 xsleep (opt.waitretry);
895 if (!opt.random_wait || count > 1)
896 /* If random-wait is not specified, or if we are sleeping
897 between retries of the same download, sleep the fixed
902 /* Sleep a random amount of time averaging in opt.wait
903 seconds. The sleeping amount ranges from 0 to
904 opt.wait*2, inclusive. */
905 double waitsecs = 2 * opt.wait * random_float ();
906 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
907 opt.wait, waitsecs));
913 /* Free the linked list of urlpos. */
915 free_urlpos (struct urlpos *l)
919 struct urlpos *next = l->next;
922 xfree_null (l->local_name);
928 /* Rotate FNAME opt.backups times */
930 rotate_backups(const char *fname)
932 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
933 char *from = (char *)alloca (maxlen);
934 char *to = (char *)alloca (maxlen);
938 if (stat (fname, &sb) == 0)
939 if (S_ISREG (sb.st_mode) == 0)
942 for (i = opt.backups; i > 1; i--)
944 sprintf (from, "%s.%d", fname, i - 1);
945 sprintf (to, "%s.%d", fname, i);
949 sprintf (to, "%s.%d", fname, 1);
953 static int no_proxy_match PARAMS ((const char *, const char **));
955 /* Return the URL of the proxy appropriate for url U. */
958 getproxy (struct url *u)
962 static char rewritten_storage[1024];
966 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
972 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
976 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
980 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
985 if (!proxy || !*proxy)
988 /* Handle shorthands. `rewritten_storage' is a kludge to allow
989 getproxy() to return static storage. */
990 rewritten_url = rewrite_shorthand_url (proxy);
993 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
994 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
995 proxy = rewritten_storage;
1001 /* Should a host be accessed through proxy, concerning no_proxy? */
1003 no_proxy_match (const char *host, const char **no_proxy)
1008 return !sufmatch (no_proxy, host);