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 */
59 # include "gen_sslfunc.h" /* for ssl_iread */
66 /* Total size of downloaded files. Used to enforce quota. */
67 LARGE_INT total_downloaded_bytes;
69 /* If non-NULL, the stream to which output should be written. This
70 stream is initialized when `-O' is used. */
73 /* Whether output_document is a regular file we can manipulate,
74 i.e. not `-' or a device file. */
75 int output_stream_regular;
84 limit_bandwidth_reset (void)
86 limit_data.chunk_bytes = 0;
87 limit_data.chunk_start = 0;
90 /* Limit the bandwidth by pausing the download for an amount of time.
91 BYTES is the number of bytes received from the network, and TIMER
92 is the timer that started at the beginning of download. */
95 limit_bandwidth (long bytes, struct wget_timer *timer)
97 double delta_t = wtimer_read (timer) - limit_data.chunk_start;
100 limit_data.chunk_bytes += bytes;
102 /* Calculate the amount of time we expect downloading the chunk
103 should take. If in reality it took less time, sleep to
104 compensate for the difference. */
105 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
107 if (expected > delta_t)
109 double slp = expected - delta_t + limit_data.sleep_adjust;
113 DEBUGP (("deferring a %.2f ms sleep (%ld/%.2f).\n",
114 slp, limit_data.chunk_bytes, delta_t));
117 DEBUGP (("\nsleeping %.2f ms for %ld bytes, adjust %.2f ms\n",
118 slp, limit_data.chunk_bytes, limit_data.sleep_adjust));
120 t0 = wtimer_read (timer);
122 wtimer_update (timer);
123 t1 = wtimer_read (timer);
125 /* Due to scheduling, we probably slept slightly longer (or
126 shorter) than desired. Calculate the difference between the
127 desired and the actual sleep, and adjust the next sleep by
129 limit_data.sleep_adjust = slp - (t1 - t0);
132 limit_data.chunk_bytes = 0;
133 limit_data.chunk_start = wtimer_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, long *skip,
163 *transferred += bufsize;
164 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. The amount of data read gets stored to
184 *TRANSFERRED. The time it took to download the data (in
185 milliseconds) 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, long toread, long startpos,
193 long *transferred, double *elapsed, int flags)
197 static char dlbuf[16384];
198 int dlbufsize = sizeof (dlbuf);
200 struct wget_timer *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 int progress_interactive = 0;
212 int exact = flags & rb_read_exactly;
215 /* How much data we've read. This is used internally and is
216 unaffected by skipping STARTPOS. */
220 if (flags & rb_skip_startpos)
225 /* If we're skipping STARTPOS bytes, hide it from
226 progress_create because the indicator can't deal with it. */
227 progress = progress_create (skip ? 0 : 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 = wtimer_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 || (total_read < toread))
256 int rdsize = exact ? MIN (toread - total_read, dlbufsize) : dlbufsize;
257 double tmout = opt.read_timeout;
258 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 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
266 if (waittm + tmout > opt.read_timeout)
268 /* Don't allow waiting time to exceed read timeout. */
269 tmout = opt.read_timeout - waittm;
272 /* We've already exceeded the timeout. */
273 ret = -1, errno = ETIMEDOUT;
278 ret = fd_read (fd, dlbuf, rdsize, tmout);
280 if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
283 ret = 0; /* timeout */
285 if (progress || opt.limit_rate)
287 wtimer_update (timer);
289 last_successful_read_tm = wtimer_read (timer);
295 if (!write_data (out, dlbuf, ret, &skip, transferred))
303 limit_bandwidth (ret, timer);
306 progress_update (progress, ret, wtimer_read (timer));
309 ws_percenttitle (100.0 *
310 (startpos + total_read) / (startpos + toread));
318 progress_finish (progress, wtimer_read (timer));
320 *elapsed = wtimer_read (timer);
322 wtimer_delete (timer);
327 /* Read a hunk of data from FD, up until a terminator. The terminator
328 is whatever the TERMINATOR function determines it to be; for
329 example, it can be a line of data, or the head of an HTTP response.
330 The function returns the data read allocated with malloc.
332 In case of error, NULL is returned. In case of EOF and no data
333 read, NULL is returned and errno set to 0. In case of EOF with
334 data having been read, the data is returned, but it will
335 (obviously) not contain the terminator.
337 The idea is to be able to read a line of input, or otherwise a hunk
338 of text, such as the head of an HTTP request, without crossing the
339 boundary, so that the next call to fd_read etc. reads the data
340 after the hunk. To achieve that, this function does the following:
342 1. Peek at available data.
344 2. Determine whether the peeked data, along with the previously
345 read data, includes the terminator.
347 2a. If yes, read the data until the end of the terminator, and
350 2b. If no, read the peeked data and goto 1.
352 The function is careful to assume as little as possible about the
353 implementation of peeking. For example, every peek is followed by
354 a read. If the read returns a different amount of data, the
355 process is retried until all data arrives safely.
357 BUFSIZE is the size of the initial buffer expected to read all the
358 data in the typical case.
360 This function should be used as a building block for other
361 functions -- see fd_read_line as a simple example. */
364 fd_read_hunk (int fd, hunk_terminator_t hunk_terminator, int bufsize)
366 char *hunk = xmalloc (bufsize);
367 int tail = 0; /* tail position in HUNK */
372 int pklen, rdlen, remain;
374 /* First, peek at the available data. */
376 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
382 end = hunk_terminator (hunk, tail, pklen);
385 /* The data contains the terminator: we'll drain the data up
386 to the end of the terminator. */
387 remain = end - (hunk + tail);
390 /* No more data needs to be read. */
394 if (bufsize - 1 < tail + remain)
396 bufsize = tail + remain + 1;
397 hunk = xrealloc (hunk, bufsize);
401 /* No terminator: simply read the data we know is (or should
405 /* Now, read the data. Note that we make no assumptions about
406 how much data we'll get. (Some TCP stacks are notorious for
407 read returning less data than the previous MSG_PEEK.) */
409 rdlen = fd_read (fd, hunk + tail, remain, 0);
422 /* EOF without anything having been read */
428 /* EOF seen: return the data we've read. */
431 if (end && rdlen == remain)
432 /* The terminator was seen and the remaining data drained --
433 we got what we came for. */
436 /* Keep looping until all the data arrives. */
438 if (tail == bufsize - 1)
441 hunk = xrealloc (hunk, bufsize);
447 line_terminator (const char *hunk, int oldlen, int peeklen)
449 const char *p = memchr (hunk + oldlen, '\n', peeklen);
451 /* p+1 because we want the line to include '\n' */
456 /* Read one line from FD and return it. The line is allocated using
459 If an error occurs, or if no data can be read, NULL is returned.
460 In the former case errno indicates the error condition, and in the
461 latter case, errno is NULL. */
464 fd_read_line (int fd)
466 return fd_read_hunk (fd, line_terminator, 128);
469 /* Return a printed representation of the download rate, as
470 appropriate for the speed. If PAD is non-zero, strings will be
471 padded to the width of 7 characters (xxxx.xx). */
473 retr_rate (long bytes, double msecs, int pad)
476 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
479 double dlrate = calc_rate (bytes, msecs, &units);
480 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
485 /* Calculate the download rate and trim it as appropriate for the
486 speed. Appropriate means that if rate is greater than 1K/s,
487 kilobytes are used, and if rate is greater than 1MB/s, megabytes
490 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
493 calc_rate (long bytes, double msecs, int *units)
501 /* If elapsed time is exactly zero, it means we're under the
502 granularity of the timer. This often happens on systems that
503 use time() for the timer. */
504 msecs = wtimer_granularity ();
506 dlrate = (double)1000 * bytes / msecs;
509 else if (dlrate < 1024.0 * 1024.0)
510 *units = 1, dlrate /= 1024.0;
511 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
512 *units = 2, dlrate /= (1024.0 * 1024.0);
514 /* Maybe someone will need this, one day. */
515 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
520 /* Maximum number of allowed redirections. 20 was chosen as a
521 "reasonable" value, which is low enough to not cause havoc, yet
522 high enough to guarantee that normal retrievals will not be hurt by
525 #define MAX_REDIRECTIONS 20
527 #define SUSPEND_POST_DATA do { \
528 post_data_suspended = 1; \
529 saved_post_data = opt.post_data; \
530 saved_post_file_name = opt.post_file_name; \
531 opt.post_data = NULL; \
532 opt.post_file_name = NULL; \
535 #define RESTORE_POST_DATA do { \
536 if (post_data_suspended) \
538 opt.post_data = saved_post_data; \
539 opt.post_file_name = saved_post_file_name; \
540 post_data_suspended = 0; \
544 static char *getproxy PARAMS ((struct url *));
546 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
549 /* #### This function should be rewritten so it doesn't return from
553 retrieve_url (const char *origurl, char **file, char **newloc,
554 const char *refurl, int *dt)
558 int location_changed, dummy;
559 char *mynewloc, *proxy;
560 struct url *u, *proxy_url;
561 int up_error_code; /* url parse error code */
563 int redirection_count = 0;
565 int post_data_suspended = 0;
566 char *saved_post_data = NULL;
567 char *saved_post_file_name = NULL;
569 /* If dt is NULL, use local storage. */
575 url = xstrdup (origurl);
581 u = url_parse (url, &up_error_code);
584 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
590 refurl = opt.referer;
599 proxy = getproxy (u);
602 /* Parse the proxy URL. */
603 proxy_url = url_parse (proxy, &up_error_code);
606 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
607 proxy, url_error (up_error_code));
612 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
614 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
615 url_free (proxy_url);
622 if (u->scheme == SCHEME_HTTP
624 || u->scheme == SCHEME_HTTPS
626 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
628 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
630 else if (u->scheme == SCHEME_FTP)
632 /* If this is a redirection, we must not allow recursive FTP
633 retrieval, so we save recursion to oldrec, and restore it
635 int oldrec = opt.recursive;
636 if (redirection_count)
638 result = ftp_loop (u, dt, proxy_url);
639 opt.recursive = oldrec;
641 /* There is a possibility of having HTTP being redirected to
642 FTP. In these cases we must decide whether the text is HTML
643 according to the suffix. The HTML suffixes are `.html',
644 `.htm' and a few others, case-insensitive. */
645 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
647 if (has_html_suffix_p (local_file))
654 url_free (proxy_url);
658 location_changed = (result == NEWLOCATION);
659 if (location_changed)
661 char *construced_newloc;
662 struct url *newloc_parsed;
664 assert (mynewloc != NULL);
669 /* The HTTP specs only allow absolute URLs to appear in
670 redirects, but a ton of boneheaded webservers and CGIs out
671 there break the rules and use relative URLs, and popular
672 browsers are lenient about this, so wget should be too. */
673 construced_newloc = uri_merge (url, mynewloc);
675 mynewloc = construced_newloc;
677 /* Now, see if this new location makes sense. */
678 newloc_parsed = url_parse (mynewloc, &up_error_code);
681 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
682 url_error (up_error_code));
690 /* Now mynewloc will become newloc_parsed->url, because if the
691 Location contained relative paths like .././something, we
692 don't want that propagating as url. */
694 mynewloc = xstrdup (newloc_parsed->url);
696 /* Check for max. number of redirections. */
697 if (++redirection_count > MAX_REDIRECTIONS)
699 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
701 url_free (newloc_parsed);
714 /* If we're being redirected from POST, we don't want to POST
715 again. Many requests answer POST with a redirection to an
716 index page; that redirection is clearly a GET. We "suspend"
717 POST data for the duration of the redirections, and restore
718 it when we're done. */
719 if (!post_data_suspended)
729 register_download (u->url, local_file);
730 if (redirection_count && 0 != strcmp (origurl, u->url))
731 register_redirection (origurl, u->url);
733 register_html (u->url, local_file);
738 *file = local_file ? local_file : NULL;
740 xfree_null (local_file);
744 if (redirection_count)
763 /* Find the URLs in the file and call retrieve_url() for each of
764 them. If HTML is non-zero, treat the file as HTML, and construct
765 the URLs accordingly.
767 If opt.recursive is set, call retrieve_tree() for each file. */
770 retrieve_from_file (const char *file, int html, int *count)
773 struct urlpos *url_list, *cur_url;
775 url_list = (html ? get_urls_html (file, NULL, NULL)
776 : get_urls_file (file));
777 status = RETROK; /* Suppose everything is OK. */
778 *count = 0; /* Reset the URL count. */
780 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
782 char *filename = NULL, *new_file = NULL;
785 if (cur_url->ignore_when_downloading)
788 if (opt.quota && total_downloaded_bytes > opt.quota)
793 if ((opt.recursive || opt.page_requisites)
794 && cur_url->url->scheme != SCHEME_FTP)
795 status = retrieve_tree (cur_url->url->url);
797 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
799 if (filename && opt.delete_after && file_exists_p (filename))
801 DEBUGP (("Removing file due to --delete-after in"
802 " retrieve_from_file():\n"));
803 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
804 if (unlink (filename))
805 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
809 xfree_null (new_file);
810 xfree_null (filename);
813 /* Free the linked list of URL-s. */
814 free_urlpos (url_list);
819 /* Print `giving up', or `retrying', depending on the impending
820 action. N1 and N2 are the attempt number and the attempt limit. */
822 printwhat (int n1, int n2)
824 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
827 /* If opt.wait or opt.waitretry are specified, and if certain
828 conditions are met, sleep the appropriate number of seconds. See
829 the documentation of --wait and --waitretry for more information.
831 COUNT is the count of current retrieval, beginning with 1. */
834 sleep_between_retrievals (int count)
836 static int first_retrieval = 1;
840 /* Don't sleep before the very first retrieval. */
845 if (opt.waitretry && count > 1)
847 /* If opt.waitretry is specified and this is a retry, wait for
848 COUNT-1 number of seconds, or for opt.waitretry seconds. */
849 if (count <= opt.waitretry)
852 xsleep (opt.waitretry);
856 if (!opt.random_wait || count > 1)
857 /* If random-wait is not specified, or if we are sleeping
858 between retries of the same download, sleep the fixed
863 /* Sleep a random amount of time averaging in opt.wait
864 seconds. The sleeping amount ranges from 0 to
865 opt.wait*2, inclusive. */
866 double waitsecs = 2 * opt.wait * random_float ();
867 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
868 opt.wait, waitsecs));
874 /* Free the linked list of urlpos. */
876 free_urlpos (struct urlpos *l)
880 struct urlpos *next = l->next;
883 xfree_null (l->local_name);
889 /* Rotate FNAME opt.backups times */
891 rotate_backups(const char *fname)
893 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
894 char *from = (char *)alloca (maxlen);
895 char *to = (char *)alloca (maxlen);
899 if (stat (fname, &sb) == 0)
900 if (S_ISREG (sb.st_mode) == 0)
903 for (i = opt.backups; i > 1; i--)
905 sprintf (from, "%s.%d", fname, i - 1);
906 sprintf (to, "%s.%d", fname, i);
910 sprintf (to, "%s.%d", fname, 1);
914 static int no_proxy_match PARAMS ((const char *, const char **));
916 /* Return the URL of the proxy appropriate for url U. */
919 getproxy (struct url *u)
923 static char rewritten_storage[1024];
927 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
933 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
937 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
941 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
946 if (!proxy || !*proxy)
949 /* Handle shorthands. `rewritten_storage' is a kludge to allow
950 getproxy() to return static storage. */
951 rewritten_url = rewrite_shorthand_url (proxy);
954 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
955 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
956 proxy = rewritten_storage;
962 /* Should a host be accessed through proxy, concerning no_proxy? */
964 no_proxy_match (const char *host, const char **no_proxy)
969 return !sufmatch (no_proxy, host);