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 /* Read the contents of file descriptor FD until it the connection
137 terminates or a read error occurs. The data is read in portions of
138 up to 16K and written to OUT as it arrives. If opt.verbose is set,
139 the progress is shown.
141 TOREAD is the amount of data expected to arrive, normally only used
142 by the progress gauge. However, if EXACT is set, no more than
143 TOREAD octets will be read.
145 STARTPOS is the position from which the download starts, used by
146 the progress gauge. The amount of data read gets stored to
147 *AMOUNT_READ. The time it took to download the data (in
148 milliseconds) is stored to *ELAPSED.
150 The function exits and returns the amount of data read. In case of
151 error while reading data, -1 is returned. In case of error while
152 writing data, -2 is returned. */
155 fd_read_body (int fd, FILE *out, long toread, int exact, long startpos,
156 long *amount_read, double *elapsed)
160 static char dlbuf[16384];
161 int dlbufsize = sizeof (dlbuf);
163 struct wget_timer *timer = NULL;
164 double last_successful_read_tm = 0;
166 /* The progress gauge, set according to the user preferences. */
167 void *progress = NULL;
169 /* Non-zero if the progress gauge is interactive, i.e. if it can
170 continually update the display. When true, smaller timeout
171 values are used so that the gauge can update the display when
172 data arrives slowly. */
173 int progress_interactive = 0;
179 progress = progress_create (startpos, toread);
180 progress_interactive = progress_interactive_p (progress);
184 limit_bandwidth_reset ();
186 /* A timer is needed for tracking progress, for throttling, and for
187 tracking elapsed time. If either of these are requested, start
189 if (progress || opt.limit_rate || elapsed)
191 timer = wtimer_new ();
192 last_successful_read_tm = 0;
195 /* Use a smaller buffer for low requested bandwidths. For example,
196 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
197 data and then sleep for 8s. With buffer size equal to the limit,
198 we never have to sleep for more than one second. */
199 if (opt.limit_rate && opt.limit_rate < dlbufsize)
200 dlbufsize = opt.limit_rate;
202 /* Read from FD while there is data to read. Normally toread==0
203 means that it is unknown how much data is to arrive. However, if
204 EXACT is set, then toread==0 means what it says: that no data
206 while (!exact || (*amount_read < toread))
208 int rdsize = exact ? MIN (toread - *amount_read, dlbufsize) : dlbufsize;
209 double tmout = opt.read_timeout;
210 if (progress_interactive)
213 /* For interactive progress gauges, always specify a ~1s
214 timeout, so that the gauge can be updated regularly even
215 when the data arrives very slowly or stalls. */
217 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
218 if (waittm + tmout > opt.read_timeout)
220 /* Don't allow waiting time to exceed read timeout. */
221 tmout = opt.read_timeout - waittm;
224 /* We've already exceeded the timeout. */
225 ret = -1, errno = ETIMEDOUT;
230 ret = fd_read (fd, dlbuf, rdsize, tmout);
232 if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
235 ret = 0; /* timeout */
237 if (progress || opt.limit_rate)
239 wtimer_update (timer);
241 last_successful_read_tm = wtimer_read (timer);
244 if (ret > 0 && out != NULL)
246 fwrite (dlbuf, 1, ret, out);
247 /* Immediately flush the downloaded data. This should not
248 hinder performance: fast downloads will arrive in large
249 16K chunks (which stdio would write out anyway), and slow
250 downloads wouldn't be limited by disk speed. */
260 limit_bandwidth (ret, timer);
264 progress_update (progress, ret, wtimer_read (timer));
267 ws_percenttitle (100.0 *
268 (startpos + *amount_read) / (startpos + toread));
276 progress_finish (progress, wtimer_read (timer));
278 *elapsed = wtimer_read (timer);
280 wtimer_delete (timer);
285 /* Read a hunk of data from FD, up until a terminator. The terminator
286 is whatever the TERMINATOR function determines it to be; for
287 example, it can be a line of data, or the head of an HTTP response.
288 The function returns the data read allocated with malloc.
290 In case of error, NULL is returned. In case of EOF and no data
291 read, NULL is returned and errno set to 0. In case of EOF with
292 data having been read, the data is returned, but it will
293 (obviously) not contain the terminator.
295 The idea is to be able to read a line of input, or otherwise a hunk
296 of text, such as the head of an HTTP request, without crossing the
297 boundary, so that the next call to fd_read etc. reads the data
298 after the hunk. To achieve that, this function does the following:
300 1. Peek at available data.
302 2. Determine whether the peeked data, along with the previously
303 read data, includes the terminator.
305 2a. If yes, read the data until the end of the terminator, and
308 2b. If no, read the peeked data and goto 1.
310 The function is careful to assume as little as possible about the
311 implementation of peeking. For example, every peek is followed by
312 a read. If the read returns a different amount of data, the
313 process is retried until all data arrives safely.
315 BUFSIZE is the size of the initial buffer expected to read all the
316 data in the typical case.
318 This function should be used as a building block for other
319 functions -- see fd_read_line as a simple example. */
322 fd_read_hunk (int fd, hunk_terminator_t hunk_terminator, int bufsize)
324 char *hunk = xmalloc (bufsize);
325 int tail = 0; /* tail position in HUNK */
330 int pklen, rdlen, remain;
332 /* First, peek at the available data. */
334 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
340 end = hunk_terminator (hunk, tail, pklen);
343 /* The data contains the terminator: we'll drain the data up
344 to the end of the terminator. */
345 remain = end - (hunk + tail);
348 /* No more data needs to be read. */
352 if (bufsize - 1 < tail + remain)
354 bufsize = tail + remain + 1;
355 hunk = xrealloc (hunk, bufsize);
359 /* No terminator: simply read the data we know is (or should
363 /* Now, read the data. Note that we make no assumptions about
364 how much data we'll get. (Some TCP stacks are notorious for
365 read returning less data than the previous MSG_PEEK.) */
367 rdlen = fd_read (fd, hunk + tail, remain, 0);
380 /* EOF without anything having been read */
386 /* EOF seen: return the data we've read. */
389 if (end && rdlen == remain)
390 /* The terminator was seen and the remaining data drained --
391 we got what we came for. */
394 /* Keep looping until all the data arrives. */
396 if (tail == bufsize - 1)
399 hunk = xrealloc (hunk, bufsize);
405 line_terminator (const char *hunk, int oldlen, int peeklen)
407 const char *p = memchr (hunk + oldlen, '\n', peeklen);
409 /* p+1 because we want the line to include '\n' */
414 /* Read one line from FD and return it. The line is allocated using
417 If an error occurs, or if no data can be read, NULL is returned.
418 In the former case errno indicates the error condition, and in the
419 latter case, errno is NULL. */
422 fd_read_line (int fd)
424 return fd_read_hunk (fd, line_terminator, 128);
427 /* Return a printed representation of the download rate, as
428 appropriate for the speed. If PAD is non-zero, strings will be
429 padded to the width of 7 characters (xxxx.xx). */
431 retr_rate (long bytes, double msecs, int pad)
434 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
437 double dlrate = calc_rate (bytes, msecs, &units);
438 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
443 /* Calculate the download rate and trim it as appropriate for the
444 speed. Appropriate means that if rate is greater than 1K/s,
445 kilobytes are used, and if rate is greater than 1MB/s, megabytes
448 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
451 calc_rate (long bytes, double msecs, int *units)
459 /* If elapsed time is exactly zero, it means we're under the
460 granularity of the timer. This often happens on systems that
461 use time() for the timer. */
462 msecs = wtimer_granularity ();
464 dlrate = (double)1000 * bytes / msecs;
467 else if (dlrate < 1024.0 * 1024.0)
468 *units = 1, dlrate /= 1024.0;
469 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
470 *units = 2, dlrate /= (1024.0 * 1024.0);
472 /* Maybe someone will need this, one day. */
473 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
478 /* Maximum number of allowed redirections. 20 was chosen as a
479 "reasonable" value, which is low enough to not cause havoc, yet
480 high enough to guarantee that normal retrievals will not be hurt by
483 #define MAX_REDIRECTIONS 20
485 #define SUSPEND_POST_DATA do { \
486 post_data_suspended = 1; \
487 saved_post_data = opt.post_data; \
488 saved_post_file_name = opt.post_file_name; \
489 opt.post_data = NULL; \
490 opt.post_file_name = NULL; \
493 #define RESTORE_POST_DATA do { \
494 if (post_data_suspended) \
496 opt.post_data = saved_post_data; \
497 opt.post_file_name = saved_post_file_name; \
498 post_data_suspended = 0; \
502 static char *getproxy PARAMS ((struct url *));
504 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
507 /* #### This function should be rewritten so it doesn't return from
511 retrieve_url (const char *origurl, char **file, char **newloc,
512 const char *refurl, int *dt)
516 int location_changed, dummy;
517 char *mynewloc, *proxy;
518 struct url *u, *proxy_url;
519 int up_error_code; /* url parse error code */
521 int redirection_count = 0;
523 int post_data_suspended = 0;
524 char *saved_post_data = NULL;
525 char *saved_post_file_name = NULL;
527 /* If dt is NULL, use local storage. */
533 url = xstrdup (origurl);
539 u = url_parse (url, &up_error_code);
542 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
548 refurl = opt.referer;
557 proxy = getproxy (u);
560 /* Parse the proxy URL. */
561 proxy_url = url_parse (proxy, &up_error_code);
564 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
565 proxy, url_error (up_error_code));
570 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
572 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
573 url_free (proxy_url);
580 if (u->scheme == SCHEME_HTTP
582 || u->scheme == SCHEME_HTTPS
584 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
586 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
588 else if (u->scheme == SCHEME_FTP)
590 /* If this is a redirection, we must not allow recursive FTP
591 retrieval, so we save recursion to oldrec, and restore it
593 int oldrec = opt.recursive;
594 if (redirection_count)
596 result = ftp_loop (u, dt, proxy_url);
597 opt.recursive = oldrec;
599 /* There is a possibility of having HTTP being redirected to
600 FTP. In these cases we must decide whether the text is HTML
601 according to the suffix. The HTML suffixes are `.html',
602 `.htm' and a few others, case-insensitive. */
603 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
605 if (has_html_suffix_p (local_file))
612 url_free (proxy_url);
616 location_changed = (result == NEWLOCATION);
617 if (location_changed)
619 char *construced_newloc;
620 struct url *newloc_parsed;
622 assert (mynewloc != NULL);
627 /* The HTTP specs only allow absolute URLs to appear in
628 redirects, but a ton of boneheaded webservers and CGIs out
629 there break the rules and use relative URLs, and popular
630 browsers are lenient about this, so wget should be too. */
631 construced_newloc = uri_merge (url, mynewloc);
633 mynewloc = construced_newloc;
635 /* Now, see if this new location makes sense. */
636 newloc_parsed = url_parse (mynewloc, &up_error_code);
639 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
640 url_error (up_error_code));
648 /* Now mynewloc will become newloc_parsed->url, because if the
649 Location contained relative paths like .././something, we
650 don't want that propagating as url. */
652 mynewloc = xstrdup (newloc_parsed->url);
654 /* Check for max. number of redirections. */
655 if (++redirection_count > MAX_REDIRECTIONS)
657 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
659 url_free (newloc_parsed);
672 /* If we're being redirected from POST, we don't want to POST
673 again. Many requests answer POST with a redirection to an
674 index page; that redirection is clearly a GET. We "suspend"
675 POST data for the duration of the redirections, and restore
676 it when we're done. */
677 if (!post_data_suspended)
687 register_download (u->url, local_file);
688 if (redirection_count && 0 != strcmp (origurl, u->url))
689 register_redirection (origurl, u->url);
691 register_html (u->url, local_file);
696 *file = local_file ? local_file : NULL;
698 xfree_null (local_file);
702 if (redirection_count)
716 ++global_download_count;
722 /* Find the URLs in the file and call retrieve_url() for each of
723 them. If HTML is non-zero, treat the file as HTML, and construct
724 the URLs accordingly.
726 If opt.recursive is set, call retrieve_tree() for each file. */
729 retrieve_from_file (const char *file, int html, int *count)
732 struct urlpos *url_list, *cur_url;
734 url_list = (html ? get_urls_html (file, NULL, NULL)
735 : get_urls_file (file));
736 status = RETROK; /* Suppose everything is OK. */
737 *count = 0; /* Reset the URL count. */
739 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
741 char *filename = NULL, *new_file = NULL;
744 if (cur_url->ignore_when_downloading)
747 if (opt.quota && total_downloaded_bytes > opt.quota)
752 if ((opt.recursive || opt.page_requisites)
753 && cur_url->url->scheme != SCHEME_FTP)
754 status = retrieve_tree (cur_url->url->url);
756 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
758 if (filename && opt.delete_after && file_exists_p (filename))
760 DEBUGP (("Removing file due to --delete-after in"
761 " retrieve_from_file():\n"));
762 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
763 if (unlink (filename))
764 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
768 xfree_null (new_file);
769 xfree_null (filename);
772 /* Free the linked list of URL-s. */
773 free_urlpos (url_list);
778 /* Print `giving up', or `retrying', depending on the impending
779 action. N1 and N2 are the attempt number and the attempt limit. */
781 printwhat (int n1, int n2)
783 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
786 /* If opt.wait or opt.waitretry are specified, and if certain
787 conditions are met, sleep the appropriate number of seconds. See
788 the documentation of --wait and --waitretry for more information.
790 COUNT is the count of current retrieval, beginning with 1. */
793 sleep_between_retrievals (int count)
795 static int first_retrieval = 1;
799 /* Don't sleep before the very first retrieval. */
804 if (opt.waitretry && count > 1)
806 /* If opt.waitretry is specified and this is a retry, wait for
807 COUNT-1 number of seconds, or for opt.waitretry seconds. */
808 if (count <= opt.waitretry)
811 xsleep (opt.waitretry);
815 if (!opt.random_wait || count > 1)
816 /* If random-wait is not specified, or if we are sleeping
817 between retries of the same download, sleep the fixed
822 /* Sleep a random amount of time averaging in opt.wait
823 seconds. The sleeping amount ranges from 0 to
824 opt.wait*2, inclusive. */
825 double waitsecs = 2 * opt.wait * random_float ();
826 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
827 opt.wait, waitsecs));
833 /* Free the linked list of urlpos. */
835 free_urlpos (struct urlpos *l)
839 struct urlpos *next = l->next;
842 xfree_null (l->local_name);
848 /* Rotate FNAME opt.backups times */
850 rotate_backups(const char *fname)
852 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
853 char *from = (char *)alloca (maxlen);
854 char *to = (char *)alloca (maxlen);
858 if (stat (fname, &sb) == 0)
859 if (S_ISREG (sb.st_mode) == 0)
862 for (i = opt.backups; i > 1; i--)
864 sprintf (from, "%s.%d", fname, i - 1);
865 sprintf (to, "%s.%d", fname, i);
869 sprintf (to, "%s.%d", fname, 1);
873 static int no_proxy_match PARAMS ((const char *, const char **));
875 /* Return the URL of the proxy appropriate for url U. */
878 getproxy (struct url *u)
882 static char rewritten_storage[1024];
886 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
892 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
896 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
900 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
905 if (!proxy || !*proxy)
908 /* Handle shorthands. `rewritten_storage' is a kludge to allow
909 getproxy() to return static storage. */
910 rewritten_url = rewrite_shorthand_url (proxy);
913 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
914 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
915 proxy = rewritten_storage;
921 /* Should a host be accessed through proxy, concerning no_proxy? */
923 no_proxy_match (const char *host, const char **no_proxy)
928 return !sufmatch (no_proxy, host);