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 /* See the comment in gethttp() why this is needed. */
67 int global_download_count;
69 /* Total size of downloaded files. Used to enforce quota. */
70 LARGE_INT total_downloaded_bytes;
80 limit_bandwidth_reset (void)
82 limit_data.chunk_bytes = 0;
83 limit_data.chunk_start = 0;
86 /* Limit the bandwidth by pausing the download for an amount of time.
87 BYTES is the number of bytes received from the network, and TIMER
88 is the timer that started at the beginning of download. */
91 limit_bandwidth (long bytes, struct wget_timer *timer)
93 double delta_t = wtimer_read (timer) - limit_data.chunk_start;
96 limit_data.chunk_bytes += bytes;
98 /* Calculate the amount of time we expect downloading the chunk
99 should take. If in reality it took less time, sleep to
100 compensate for the difference. */
101 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
103 if (expected > delta_t)
105 double slp = expected - delta_t + limit_data.sleep_adjust;
109 DEBUGP (("deferring a %.2f ms sleep (%ld/%.2f).\n",
110 slp, limit_data.chunk_bytes, delta_t));
113 DEBUGP (("\nsleeping %.2f ms for %ld bytes, adjust %.2f ms\n",
114 slp, limit_data.chunk_bytes, limit_data.sleep_adjust));
116 t0 = wtimer_read (timer);
118 wtimer_update (timer);
119 t1 = wtimer_read (timer);
121 /* Due to scheduling, we probably slept slightly longer (or
122 shorter) than desired. Calculate the difference between the
123 desired and the actual sleep, and adjust the next sleep by
125 limit_data.sleep_adjust = slp - (t1 - t0);
128 limit_data.chunk_bytes = 0;
129 limit_data.chunk_start = wtimer_read (timer);
133 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
136 /* Reads the contents of file descriptor FD, until it is closed, or a
137 read error occurs. The data is read in 8K chunks, and stored to
138 stream fp, which should have been open for writing.
140 The EXPECTED argument is passed to show_progress() unchanged, but
143 If opt.verbose is set, the progress is also shown. RESTVAL
144 represents a value from which to start downloading (which will be
145 shown accordingly). If RESTVAL is non-zero, the stream should have
146 been open for appending.
148 The function exits and returns codes of 0, -1 and -2 if the
149 connection was closed, there was a read error, or if it could not
150 write to the output stream, respectively. */
153 fd_read_body (int fd, FILE *out, long *len, long restval, long expected,
154 int use_expected, double *elapsed)
158 static char dlbuf[16384];
159 int dlbufsize = sizeof (dlbuf);
161 struct wget_timer *timer = wtimer_allocate ();
162 double last_successful_read_tm;
164 /* The progress gauge, set according to the user preferences. */
165 void *progress = NULL;
167 /* Non-zero if the progress gauge is interactive, i.e. if it can
168 continually update the display. When true, smaller timeout
169 values are used so that the gauge can update the display when
170 data arrives slowly. */
171 int progress_interactive = 0;
177 progress = progress_create (restval, expected);
178 progress_interactive = progress_interactive_p (progress);
182 limit_bandwidth_reset ();
183 wtimer_reset (timer);
184 last_successful_read_tm = 0;
186 /* Use a smaller buffer for low requested bandwidths. For example,
187 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
188 data and then sleep for 8s. With buffer size equal to the limit,
189 we never have to sleep for more than one second. */
190 if (opt.limit_rate && opt.limit_rate < dlbufsize)
191 dlbufsize = opt.limit_rate;
193 /* Read from fd while there is available data.
195 Normally, if expected is 0, it means that it is not known how
196 much data is expected. However, if use_expected is specified,
197 then expected being zero means exactly that. */
198 while (!use_expected || (*len < expected))
200 int amount_to_read = (use_expected
201 ? MIN (expected - *len, dlbufsize) : dlbufsize);
202 double tmout = opt.read_timeout;
203 if (progress_interactive)
206 /* For interactive progress gauges, always specify a ~1s
207 timeout, so that the gauge can be updated regularly even
208 when the data arrives very slowly or stalls. */
210 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
211 if (waittm + tmout > opt.read_timeout)
213 /* Don't allow waiting time to exceed read timeout. */
214 tmout = opt.read_timeout - waittm;
217 /* We've already exceeded the timeout. */
218 res = -1, errno = ETIMEDOUT;
223 res = fd_read (fd, dlbuf, amount_to_read, tmout);
225 if (res == 0 || (res < 0 && errno != ETIMEDOUT))
228 res = 0; /* timeout */
230 wtimer_update (timer);
233 fwrite (dlbuf, 1, res, out);
234 /* Always flush the contents of the network packet. This
235 should not hinder performance: fast downloads will be
236 received in 16K chunks (which stdio would write out
237 anyway), and slow downloads won't be limited by disk
245 last_successful_read_tm = wtimer_read (timer);
249 limit_bandwidth (res, timer);
253 progress_update (progress, res, wtimer_read (timer));
255 if (use_expected && expected > 0)
256 ws_percenttitle (100.0 * (double)(*len) / (double)expected);
264 progress_finish (progress, wtimer_read (timer));
266 *elapsed = wtimer_read (timer);
267 wtimer_delete (timer);
272 /* Read a hunk of data from FD, up until a terminator. The terminator
273 is whatever the TERMINATOR function determines it to be; for
274 example, it can be a line of data, or the head of an HTTP response.
275 The function returns the data read allocated with malloc.
277 In case of error, NULL is returned. In case of EOF and no data
278 read, NULL is returned and errno set to 0. In case of EOF with
279 data having been read, the data is returned, but it will
280 (obviously) not contain the terminator.
282 The idea is to be able to read a line of input, or otherwise a hunk
283 of text, such as the head of an HTTP request, without crossing the
284 boundary, so that the next call to fd_read etc. reads the data
285 after the hunk. To achieve that, this function does the following:
287 1. Peek at available data.
289 2. Determine whether the peeked data, along with the previously
290 read data, includes the terminator.
292 2a. If yes, read the data until the end of the terminator, and
295 2b. If no, read the peeked data and goto 1.
297 The function is careful to assume as little as possible about the
298 implementation of peeking. For example, every peek is followed by
299 a read. If the read returns a different amount of data, the
300 process is retried until all data arrives safely.
302 BUFSIZE is the size of the initial buffer expected to read all the
303 data in the typical case.
305 This function should be used as a building block for other
306 functions -- see fd_read_line as a simple example. */
309 fd_read_hunk (int fd, hunk_terminator_t hunk_terminator, int bufsize)
311 char *hunk = xmalloc (bufsize);
312 int tail = 0; /* tail position in HUNK */
317 int pklen, rdlen, remain;
319 /* First, peek at the available data. */
321 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
327 end = hunk_terminator (hunk, tail, pklen);
330 /* The data contains the terminator: we'll drain the data up
331 to the end of the terminator. */
332 remain = end - (hunk + tail);
335 /* No more data needs to be read. */
339 if (bufsize - 1 < tail + remain)
341 bufsize = tail + remain + 1;
342 hunk = xrealloc (hunk, bufsize);
346 /* No terminator: simply read the data we know is (or should
350 /* Now, read the data. Note that we make no assumptions about
351 how much data we'll get. (Some TCP stacks are notorious for
352 read returning less data than the previous MSG_PEEK.) */
354 rdlen = fd_read (fd, hunk + tail, remain, 0);
367 /* EOF without anything having been read */
373 /* EOF seen: return the data we've read. */
376 if (end && rdlen == remain)
377 /* The terminator was seen and the remaining data drained --
378 we got what we came for. */
381 /* Keep looping until all the data arrives. */
383 if (tail == bufsize - 1)
386 hunk = xrealloc (hunk, bufsize);
392 line_terminator (const char *hunk, int oldlen, int peeklen)
394 const char *p = memchr (hunk + oldlen, '\n', peeklen);
396 /* p+1 because we want the line to include '\n' */
401 /* Read one line from FD and return it. The line is allocated using
404 If an error occurs, or if no data can be read, NULL is returned.
405 In the former case errno indicates the error condition, and in the
406 latter case, errno is NULL. */
409 fd_read_line (int fd)
411 return fd_read_hunk (fd, line_terminator, 128);
414 /* Return a printed representation of the download rate, as
415 appropriate for the speed. If PAD is non-zero, strings will be
416 padded to the width of 7 characters (xxxx.xx). */
418 retr_rate (long bytes, double msecs, int pad)
421 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
424 double dlrate = calc_rate (bytes, msecs, &units);
425 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
430 /* Calculate the download rate and trim it as appropriate for the
431 speed. Appropriate means that if rate is greater than 1K/s,
432 kilobytes are used, and if rate is greater than 1MB/s, megabytes
435 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
438 calc_rate (long bytes, double msecs, int *units)
446 /* If elapsed time is exactly zero, it means we're under the
447 granularity of the timer. This often happens on systems that
448 use time() for the timer. */
449 msecs = wtimer_granularity ();
451 dlrate = (double)1000 * bytes / msecs;
454 else if (dlrate < 1024.0 * 1024.0)
455 *units = 1, dlrate /= 1024.0;
456 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
457 *units = 2, dlrate /= (1024.0 * 1024.0);
459 /* Maybe someone will need this, one day. */
460 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
465 /* Maximum number of allowed redirections. 20 was chosen as a
466 "reasonable" value, which is low enough to not cause havoc, yet
467 high enough to guarantee that normal retrievals will not be hurt by
470 #define MAX_REDIRECTIONS 20
472 #define SUSPEND_POST_DATA do { \
473 post_data_suspended = 1; \
474 saved_post_data = opt.post_data; \
475 saved_post_file_name = opt.post_file_name; \
476 opt.post_data = NULL; \
477 opt.post_file_name = NULL; \
480 #define RESTORE_POST_DATA do { \
481 if (post_data_suspended) \
483 opt.post_data = saved_post_data; \
484 opt.post_file_name = saved_post_file_name; \
485 post_data_suspended = 0; \
489 static char *getproxy PARAMS ((struct url *));
491 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
494 /* #### This function should be rewritten so it doesn't return from
498 retrieve_url (const char *origurl, char **file, char **newloc,
499 const char *refurl, int *dt)
503 int location_changed, dummy;
504 char *mynewloc, *proxy;
505 struct url *u, *proxy_url;
506 int up_error_code; /* url parse error code */
508 int redirection_count = 0;
510 int post_data_suspended = 0;
511 char *saved_post_data = NULL;
512 char *saved_post_file_name = NULL;
514 /* If dt is NULL, use local storage. */
520 url = xstrdup (origurl);
526 u = url_parse (url, &up_error_code);
529 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
535 refurl = opt.referer;
544 proxy = getproxy (u);
547 /* Parse the proxy URL. */
548 proxy_url = url_parse (proxy, &up_error_code);
551 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
552 proxy, url_error (up_error_code));
557 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
559 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
560 url_free (proxy_url);
567 if (u->scheme == SCHEME_HTTP
569 || u->scheme == SCHEME_HTTPS
571 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
573 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
575 else if (u->scheme == SCHEME_FTP)
577 /* If this is a redirection, we must not allow recursive FTP
578 retrieval, so we save recursion to oldrec, and restore it
580 int oldrec = opt.recursive;
581 if (redirection_count)
583 result = ftp_loop (u, dt, proxy_url);
584 opt.recursive = oldrec;
586 /* There is a possibility of having HTTP being redirected to
587 FTP. In these cases we must decide whether the text is HTML
588 according to the suffix. The HTML suffixes are `.html',
589 `.htm' and a few others, case-insensitive. */
590 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
592 if (has_html_suffix_p (local_file))
599 url_free (proxy_url);
603 location_changed = (result == NEWLOCATION);
604 if (location_changed)
606 char *construced_newloc;
607 struct url *newloc_parsed;
609 assert (mynewloc != NULL);
614 /* The HTTP specs only allow absolute URLs to appear in
615 redirects, but a ton of boneheaded webservers and CGIs out
616 there break the rules and use relative URLs, and popular
617 browsers are lenient about this, so wget should be too. */
618 construced_newloc = uri_merge (url, mynewloc);
620 mynewloc = construced_newloc;
622 /* Now, see if this new location makes sense. */
623 newloc_parsed = url_parse (mynewloc, &up_error_code);
626 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
627 url_error (up_error_code));
635 /* Now mynewloc will become newloc_parsed->url, because if the
636 Location contained relative paths like .././something, we
637 don't want that propagating as url. */
639 mynewloc = xstrdup (newloc_parsed->url);
641 /* Check for max. number of redirections. */
642 if (++redirection_count > MAX_REDIRECTIONS)
644 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
646 url_free (newloc_parsed);
659 /* If we're being redirected from POST, we don't want to POST
660 again. Many requests answer POST with a redirection to an
661 index page; that redirection is clearly a GET. We "suspend"
662 POST data for the duration of the redirections, and restore
663 it when we're done. */
664 if (!post_data_suspended)
674 register_download (u->url, local_file);
675 if (redirection_count && 0 != strcmp (origurl, u->url))
676 register_redirection (origurl, u->url);
678 register_html (u->url, local_file);
683 *file = local_file ? local_file : NULL;
685 xfree_null (local_file);
689 if (redirection_count)
703 ++global_download_count;
709 /* Find the URLs in the file and call retrieve_url() for each of
710 them. If HTML is non-zero, treat the file as HTML, and construct
711 the URLs accordingly.
713 If opt.recursive is set, call retrieve_tree() for each file. */
716 retrieve_from_file (const char *file, int html, int *count)
719 struct urlpos *url_list, *cur_url;
721 url_list = (html ? get_urls_html (file, NULL, NULL)
722 : get_urls_file (file));
723 status = RETROK; /* Suppose everything is OK. */
724 *count = 0; /* Reset the URL count. */
726 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
728 char *filename = NULL, *new_file = NULL;
731 if (cur_url->ignore_when_downloading)
734 if (opt.quota && total_downloaded_bytes > opt.quota)
739 if ((opt.recursive || opt.page_requisites)
740 && cur_url->url->scheme != SCHEME_FTP)
741 status = retrieve_tree (cur_url->url->url);
743 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
745 if (filename && opt.delete_after && file_exists_p (filename))
747 DEBUGP (("Removing file due to --delete-after in"
748 " retrieve_from_file():\n"));
749 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
750 if (unlink (filename))
751 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
755 xfree_null (new_file);
756 xfree_null (filename);
759 /* Free the linked list of URL-s. */
760 free_urlpos (url_list);
765 /* Print `giving up', or `retrying', depending on the impending
766 action. N1 and N2 are the attempt number and the attempt limit. */
768 printwhat (int n1, int n2)
770 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
773 /* If opt.wait or opt.waitretry are specified, and if certain
774 conditions are met, sleep the appropriate number of seconds. See
775 the documentation of --wait and --waitretry for more information.
777 COUNT is the count of current retrieval, beginning with 1. */
780 sleep_between_retrievals (int count)
782 static int first_retrieval = 1;
786 /* Don't sleep before the very first retrieval. */
791 if (opt.waitretry && count > 1)
793 /* If opt.waitretry is specified and this is a retry, wait for
794 COUNT-1 number of seconds, or for opt.waitretry seconds. */
795 if (count <= opt.waitretry)
798 xsleep (opt.waitretry);
802 if (!opt.random_wait || count > 1)
803 /* If random-wait is not specified, or if we are sleeping
804 between retries of the same download, sleep the fixed
809 /* Sleep a random amount of time averaging in opt.wait
810 seconds. The sleeping amount ranges from 0 to
811 opt.wait*2, inclusive. */
812 double waitsecs = 2 * opt.wait * random_float ();
813 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
814 opt.wait, waitsecs));
820 /* Free the linked list of urlpos. */
822 free_urlpos (struct urlpos *l)
826 struct urlpos *next = l->next;
829 xfree_null (l->local_name);
835 /* Rotate FNAME opt.backups times */
837 rotate_backups(const char *fname)
839 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
840 char *from = (char *)alloca (maxlen);
841 char *to = (char *)alloca (maxlen);
845 if (stat (fname, &sb) == 0)
846 if (S_ISREG (sb.st_mode) == 0)
849 for (i = opt.backups; i > 1; i--)
851 sprintf (from, "%s.%d", fname, i - 1);
852 sprintf (to, "%s.%d", fname, i);
856 sprintf (to, "%s.%d", fname, 1);
860 static int no_proxy_match PARAMS ((const char *, const char **));
862 /* Return the URL of the proxy appropriate for url U. */
865 getproxy (struct url *u)
869 static char rewritten_storage[1024];
873 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
879 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
883 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
887 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
892 if (!proxy || !*proxy)
895 /* Handle shorthands. `rewritten_storage' is a kludge to allow
896 getproxy() to return static storage. */
897 rewritten_url = rewrite_shorthand_url (proxy);
900 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
901 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
902 proxy = rewritten_storage;
908 /* Should a host be accessed through proxy, concerning no_proxy? */
910 no_proxy_match (const char *host, const char **no_proxy)
915 return !sufmatch (no_proxy, host);