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 */
54 /* Total size of downloaded files. Used to enforce quota. */
55 SUM_SIZE_INT total_downloaded_bytes;
57 /* Total download time in milliseconds. */
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)
77 limit_data.chunk_bytes = 0;
78 limit_data.chunk_start = 0;
79 limit_data.sleep_adjust = 0;
82 /* Limit the bandwidth by pausing the download for an amount of time.
83 BYTES is the number of bytes received from the network, and TIMER
84 is the timer that started at the beginning of download. */
87 limit_bandwidth (wgint bytes, struct ptimer *timer)
89 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
92 limit_data.chunk_bytes += bytes;
94 /* Calculate the amount of time we expect downloading the chunk
95 should take. If in reality it took less time, sleep to
96 compensate for the difference. */
97 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
99 if (expected > delta_t)
101 double slp = expected - delta_t + limit_data.sleep_adjust;
105 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
106 slp, number_to_static_string (limit_data.chunk_bytes),
110 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
111 slp, number_to_static_string (limit_data.chunk_bytes),
112 limit_data.sleep_adjust));
114 t0 = ptimer_read (timer);
116 t1 = ptimer_measure (timer);
118 /* Due to scheduling, we probably slept slightly longer (or
119 shorter) than desired. Calculate the difference between the
120 desired and the actual sleep, and adjust the next sleep by
122 limit_data.sleep_adjust = slp - (t1 - t0);
123 /* If sleep_adjust is very large, it's likely due to suspension
124 and not clock inaccuracy. Don't enforce those. */
125 if (limit_data.sleep_adjust > 500)
126 limit_data.sleep_adjust = 500;
127 else if (limit_data.sleep_adjust < -500)
128 limit_data.sleep_adjust = -500;
131 limit_data.chunk_bytes = 0;
132 limit_data.chunk_start = ptimer_read (timer);
136 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
139 /* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
140 amount of data and decrease SKIP. Increment *TOTAL by the amount
144 write_data (FILE *out, const char *buf, int bufsize, wgint *skip,
163 fwrite (buf, 1, bufsize, out);
166 /* Immediately flush the downloaded data. This should not hinder
167 performance: fast downloads will arrive in large 16K chunks
168 (which stdio would write out immediately anyway), and slow
169 downloads wouldn't be limited by disk speed. */
171 return !ferror (out);
174 /* Read the contents of file descriptor FD until it the connection
175 terminates or a read error occurs. The data is read in portions of
176 up to 16K and written to OUT as it arrives. If opt.verbose is set,
177 the progress is shown.
179 TOREAD is the amount of data expected to arrive, normally only used
180 by the progress gauge.
182 STARTPOS is the position from which the download starts, used by
183 the progress gauge. If QTYREAD is non-NULL, the value it points to
184 is incremented by the amount of data read from the network. If
185 QTYWRITTEN is non-NULL, the value it points to is incremented by
186 the amount of data written to disk. The time it took to download
187 the data (in milliseconds) is stored to ELAPSED.
189 The function exits and returns the amount of data read. In case of
190 error while reading data, -1 is returned. In case of error while
191 writing data, -2 is returned. */
194 fd_read_body (int fd, FILE *out, wgint toread, wgint startpos,
195 wgint *qtyread, wgint *qtywritten, double *elapsed, int flags)
199 static char dlbuf[16384];
200 int dlbufsize = sizeof (dlbuf);
202 struct ptimer *timer = NULL;
203 double last_successful_read_tm = 0;
205 /* The progress gauge, set according to the user preferences. */
206 void *progress = NULL;
208 /* Non-zero if the progress gauge is interactive, i.e. if it can
209 continually update the display. When true, smaller timeout
210 values are used so that the gauge can update the display when
211 data arrives slowly. */
212 bool progress_interactive = false;
214 bool exact = !!(flags & rb_read_exactly);
217 /* How much data we've read/written. */
219 wgint sum_written = 0;
221 if (flags & rb_skip_startpos)
226 /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
227 argument to progress_create because the indicator doesn't
228 (yet) know about "skipping" data. */
229 progress = progress_create (skip ? 0 : startpos, startpos + toread);
230 progress_interactive = progress_interactive_p (progress);
234 limit_bandwidth_reset ();
236 /* A timer is needed for tracking progress, for throttling, and for
237 tracking elapsed time. If either of these are requested, start
239 if (progress || opt.limit_rate || elapsed)
241 timer = ptimer_new ();
242 last_successful_read_tm = 0;
245 /* Use a smaller buffer for low requested bandwidths. For example,
246 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
247 data and then sleep for 8s. With buffer size equal to the limit,
248 we never have to sleep for more than one second. */
249 if (opt.limit_rate && opt.limit_rate < dlbufsize)
250 dlbufsize = opt.limit_rate;
252 /* Read from FD while there is data to read. Normally toread==0
253 means that it is unknown how much data is to arrive. However, if
254 EXACT is set, then toread==0 means what it says: that no data
256 while (!exact || (sum_read < toread))
258 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
259 double tmout = opt.read_timeout;
260 if (progress_interactive)
262 /* For interactive progress gauges, always specify a ~1s
263 timeout, so that the gauge can be updated regularly even
264 when the data arrives very slowly or stalls. */
266 if (opt.read_timeout)
269 waittm = (ptimer_read (timer) - last_successful_read_tm) / 1000;
270 if (waittm + tmout > opt.read_timeout)
272 /* Don't let total idle time exceed read timeout. */
273 tmout = opt.read_timeout - waittm;
276 /* We've already exceeded the timeout. */
277 ret = -1, errno = ETIMEDOUT;
283 ret = fd_read (fd, dlbuf, rdsize, tmout);
285 if (progress_interactive && ret < 0 && errno == ETIMEDOUT)
286 ret = 0; /* interactive timeout, handled above */
288 break; /* EOF or read error */
290 if (progress || opt.limit_rate)
292 ptimer_measure (timer);
294 last_successful_read_tm = ptimer_read (timer);
300 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
308 limit_bandwidth (ret, timer);
311 progress_update (progress, ret, ptimer_read (timer));
313 if (toread > 0 && !opt.quiet)
314 ws_percenttitle (100.0 *
315 (startpos + sum_read) / (startpos + toread));
323 progress_finish (progress, ptimer_read (timer));
326 *elapsed = ptimer_read (timer);
328 ptimer_destroy (timer);
331 *qtyread += sum_read;
333 *qtywritten += sum_written;
338 /* Read a hunk of data from FD, up until a terminator. The terminator
339 is whatever the TERMINATOR function determines it to be; for
340 example, it can be a line of data, or the head of an HTTP response.
341 The function returns the data read allocated with malloc.
343 In case of error, NULL is returned. In case of EOF and no data
344 read, NULL is returned and errno set to 0. In case of EOF with
345 data having been read, the data is returned, but it will
346 (obviously) not contain the terminator.
348 The idea is to be able to read a line of input, or otherwise a hunk
349 of text, such as the head of an HTTP request, without crossing the
350 boundary, so that the next call to fd_read etc. reads the data
351 after the hunk. To achieve that, this function does the following:
353 1. Peek at available data.
355 2. Determine whether the peeked data, along with the previously
356 read data, includes the terminator.
358 2a. If yes, read the data until the end of the terminator, and
361 2b. If no, read the peeked data and goto 1.
363 The function is careful to assume as little as possible about the
364 implementation of peeking. For example, every peek is followed by
365 a read. If the read returns a different amount of data, the
366 process is retried until all data arrives safely.
368 SIZEHINT is the buffer size sufficient to hold all the data in the
369 typical case (it is used as the initial buffer size). MAXSIZE is
370 the maximum amount of memory this function is allowed to allocate,
371 or 0 if no upper limit is to be enforced.
373 This function should be used as a building block for other
374 functions -- see fd_read_line as a simple example. */
377 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
379 long bufsize = sizehint;
380 char *hunk = xmalloc (bufsize);
381 int tail = 0; /* tail position in HUNK */
383 assert (maxsize >= bufsize);
388 int pklen, rdlen, remain;
390 /* First, peek at the available data. */
392 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
398 end = terminator (hunk, tail, pklen);
401 /* The data contains the terminator: we'll drain the data up
402 to the end of the terminator. */
403 remain = end - (hunk + tail);
406 /* No more data needs to be read. */
410 if (bufsize - 1 < tail + remain)
412 bufsize = tail + remain + 1;
413 hunk = xrealloc (hunk, bufsize);
417 /* No terminator: simply read the data we know is (or should
421 /* Now, read the data. Note that we make no assumptions about
422 how much data we'll get. (Some TCP stacks are notorious for
423 read returning less data than the previous MSG_PEEK.) */
425 rdlen = fd_read (fd, hunk + tail, remain, 0);
438 /* EOF without anything having been read */
444 /* EOF seen: return the data we've read. */
447 if (end && rdlen == remain)
448 /* The terminator was seen and the remaining data drained --
449 we got what we came for. */
452 /* Keep looping until all the data arrives. */
454 if (tail == bufsize - 1)
456 /* Double the buffer size, but refuse to allocate more than
458 if (maxsize && bufsize >= maxsize)
465 if (maxsize && bufsize > maxsize)
467 hunk = xrealloc (hunk, bufsize);
473 line_terminator (const char *hunk, int oldlen, int peeklen)
475 const char *p = memchr (hunk + oldlen, '\n', peeklen);
477 /* p+1 because we want the line to include '\n' */
482 /* The maximum size of the single line we agree to accept. This is
483 not meant to impose an arbitrary limit, but to protect the user
484 from Wget slurping up available memory upon encountering malicious
485 or buggy server output. Define it to 0 to remove the limit. */
486 #define FD_READ_LINE_MAX 4096
488 /* Read one line from FD and return it. The line is allocated using
489 malloc, but is never larger than FD_READ_LINE_MAX.
491 If an error occurs, or if no data can be read, NULL is returned.
492 In the former case errno indicates the error condition, and in the
493 latter case, errno is NULL. */
496 fd_read_line (int fd)
498 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
501 /* Return a printed representation of the download rate, along with
502 the units appropriate for the download speed. */
505 retr_rate (wgint bytes, double msecs)
508 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
511 double dlrate = calc_rate (bytes, msecs, &units);
512 sprintf (res, "%.2f %s", dlrate, rate_names[units]);
517 /* Calculate the download rate and trim it as appropriate for the
518 speed. Appropriate means that if rate is greater than 1K/s,
519 kilobytes are used, and if rate is greater than 1MB/s, megabytes
522 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
525 calc_rate (wgint bytes, double msecs, int *units)
533 /* If elapsed time is exactly zero, it means we're under the
534 resolution of the timer. This can easily happen on systems
535 that use time() for the timer. Since the interval lies between
536 0 and the timer's resolution, assume half the resolution. */
537 msecs = ptimer_resolution () / 2.0;
539 dlrate = 1000.0 * bytes / msecs;
542 else if (dlrate < 1024.0 * 1024.0)
543 *units = 1, dlrate /= 1024.0;
544 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
545 *units = 2, dlrate /= (1024.0 * 1024.0);
547 /* Maybe someone will need this, one day. */
548 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
553 /* Maximum number of allowed redirections. 20 was chosen as a
554 "reasonable" value, which is low enough to not cause havoc, yet
555 high enough to guarantee that normal retrievals will not be hurt by
558 #define MAX_REDIRECTIONS 20
560 #define SUSPEND_POST_DATA do { \
561 post_data_suspended = true; \
562 saved_post_data = opt.post_data; \
563 saved_post_file_name = opt.post_file_name; \
564 opt.post_data = NULL; \
565 opt.post_file_name = NULL; \
568 #define RESTORE_POST_DATA do { \
569 if (post_data_suspended) \
571 opt.post_data = saved_post_data; \
572 opt.post_file_name = saved_post_file_name; \
573 post_data_suspended = false; \
577 static char *getproxy (struct url *);
579 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
582 /* #### This function should be rewritten so it doesn't return from
586 retrieve_url (const char *origurl, char **file, char **newloc,
587 const char *refurl, int *dt)
591 bool location_changed;
593 char *mynewloc, *proxy;
594 struct url *u, *proxy_url;
595 int up_error_code; /* url parse error code */
597 int redirection_count = 0;
599 bool post_data_suspended = false;
600 char *saved_post_data = NULL;
601 char *saved_post_file_name = NULL;
603 /* If dt is NULL, use local storage. */
609 url = xstrdup (origurl);
615 u = url_parse (url, &up_error_code);
618 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
624 refurl = opt.referer;
633 proxy = getproxy (u);
636 /* Parse the proxy URL. */
637 proxy_url = url_parse (proxy, &up_error_code);
640 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
641 proxy, url_error (up_error_code));
646 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
648 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
649 url_free (proxy_url);
656 if (u->scheme == SCHEME_HTTP
658 || u->scheme == SCHEME_HTTPS
660 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
662 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
664 else if (u->scheme == SCHEME_FTP)
666 /* If this is a redirection, temporarily turn off opt.ftp_glob
667 and opt.recursive, both being undesirable when following
669 bool oldrec = opt.recursive, oldglob = opt.ftp_glob;
670 if (redirection_count)
671 opt.recursive = opt.ftp_glob = false;
673 result = ftp_loop (u, dt, proxy_url);
674 opt.recursive = oldrec;
675 opt.ftp_glob = oldglob;
677 /* There is a possibility of having HTTP being redirected to
678 FTP. In these cases we must decide whether the text is HTML
679 according to the suffix. The HTML suffixes are `.html',
680 `.htm' and a few others, case-insensitive. */
681 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
683 if (has_html_suffix_p (local_file))
690 url_free (proxy_url);
694 location_changed = (result == NEWLOCATION);
695 if (location_changed)
697 char *construced_newloc;
698 struct url *newloc_parsed;
700 assert (mynewloc != NULL);
705 /* The HTTP specs only allow absolute URLs to appear in
706 redirects, but a ton of boneheaded webservers and CGIs out
707 there break the rules and use relative URLs, and popular
708 browsers are lenient about this, so wget should be too. */
709 construced_newloc = uri_merge (url, mynewloc);
711 mynewloc = construced_newloc;
713 /* Now, see if this new location makes sense. */
714 newloc_parsed = url_parse (mynewloc, &up_error_code);
717 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
718 url_error (up_error_code));
726 /* Now mynewloc will become newloc_parsed->url, because if the
727 Location contained relative paths like .././something, we
728 don't want that propagating as url. */
730 mynewloc = xstrdup (newloc_parsed->url);
732 /* Check for max. number of redirections. */
733 if (++redirection_count > MAX_REDIRECTIONS)
735 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
737 url_free (newloc_parsed);
750 /* If we're being redirected from POST, we don't want to POST
751 again. Many requests answer POST with a redirection to an
752 index page; that redirection is clearly a GET. We "suspend"
753 POST data for the duration of the redirections, and restore
754 it when we're done. */
755 if (!post_data_suspended)
765 register_download (u->url, local_file);
766 if (redirection_count && 0 != strcmp (origurl, u->url))
767 register_redirection (origurl, u->url);
769 register_html (u->url, local_file);
774 *file = local_file ? local_file : NULL;
776 xfree_null (local_file);
780 if (redirection_count)
799 /* Find the URLs in the file and call retrieve_url() for each of them.
800 If HTML is true, treat the file as HTML, and construct the URLs
803 If opt.recursive is set, call retrieve_tree() for each file. */
806 retrieve_from_file (const char *file, bool html, int *count)
809 struct urlpos *url_list, *cur_url;
811 url_list = (html ? get_urls_html (file, NULL, NULL)
812 : get_urls_file (file));
813 status = RETROK; /* Suppose everything is OK. */
814 *count = 0; /* Reset the URL count. */
816 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
818 char *filename = NULL, *new_file = NULL;
821 if (cur_url->ignore_when_downloading)
824 if (opt.quota && total_downloaded_bytes > opt.quota)
829 if ((opt.recursive || opt.page_requisites)
830 && cur_url->url->scheme != SCHEME_FTP)
831 status = retrieve_tree (cur_url->url->url);
833 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
835 if (filename && opt.delete_after && file_exists_p (filename))
838 Removing file due to --delete-after in retrieve_from_file():\n"));
839 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
840 if (unlink (filename))
841 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
845 xfree_null (new_file);
846 xfree_null (filename);
849 /* Free the linked list of URL-s. */
850 free_urlpos (url_list);
855 /* Print `giving up', or `retrying', depending on the impending
856 action. N1 and N2 are the attempt number and the attempt limit. */
858 printwhat (int n1, int n2)
860 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
863 /* If opt.wait or opt.waitretry are specified, and if certain
864 conditions are met, sleep the appropriate number of seconds. See
865 the documentation of --wait and --waitretry for more information.
867 COUNT is the count of current retrieval, beginning with 1. */
870 sleep_between_retrievals (int count)
872 static bool first_retrieval = true;
876 /* Don't sleep before the very first retrieval. */
877 first_retrieval = false;
881 if (opt.waitretry && count > 1)
883 /* If opt.waitretry is specified and this is a retry, wait for
884 COUNT-1 number of seconds, or for opt.waitretry seconds. */
885 if (count <= opt.waitretry)
888 xsleep (opt.waitretry);
892 if (!opt.random_wait || count > 1)
893 /* If random-wait is not specified, or if we are sleeping
894 between retries of the same download, sleep the fixed
899 /* Sleep a random amount of time averaging in opt.wait
900 seconds. The sleeping amount ranges from 0 to
901 opt.wait*2, inclusive. */
902 double waitsecs = 2 * opt.wait * random_float ();
903 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
904 opt.wait, waitsecs));
910 /* Free the linked list of urlpos. */
912 free_urlpos (struct urlpos *l)
916 struct urlpos *next = l->next;
919 xfree_null (l->local_name);
925 /* Rotate FNAME opt.backups times */
927 rotate_backups(const char *fname)
929 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
930 char *from = (char *)alloca (maxlen);
931 char *to = (char *)alloca (maxlen);
935 if (stat (fname, &sb) == 0)
936 if (S_ISREG (sb.st_mode) == 0)
939 for (i = opt.backups; i > 1; i--)
941 sprintf (from, "%s.%d", fname, i - 1);
942 sprintf (to, "%s.%d", fname, i);
946 sprintf (to, "%s.%d", fname, 1);
950 static bool no_proxy_match (const char *, const char **);
952 /* Return the URL of the proxy appropriate for url U. */
955 getproxy (struct url *u)
959 static char rewritten_storage[1024];
963 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
969 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
973 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
977 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
982 if (!proxy || !*proxy)
985 /* Handle shorthands. `rewritten_storage' is a kludge to allow
986 getproxy() to return static storage. */
987 rewritten_url = rewrite_shorthand_url (proxy);
990 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
991 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
992 proxy = rewritten_storage;
998 /* Should a host be accessed through proxy, concerning no_proxy? */
1000 no_proxy_match (const char *host, const char **no_proxy)
1005 return !sufmatch (no_proxy, host);