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,
164 fwrite (buf, 1, bufsize, out);
167 /* Immediately flush the downloaded data. This should not hinder
168 performance: fast downloads will arrive in large 16K chunks
169 (which stdio would write out immediately anyway), and slow
170 downloads wouldn't be limited by disk speed. */
172 return !ferror (out);
175 /* Read the contents of file descriptor FD until it the connection
176 terminates or a read error occurs. The data is read in portions of
177 up to 16K and written to OUT as it arrives. If opt.verbose is set,
178 the progress is shown.
180 TOREAD is the amount of data expected to arrive, normally only used
181 by the progress gauge.
183 STARTPOS is the position from which the download starts, used by
184 the progress gauge. If QTYREAD is non-NULL, the value it points to
185 is incremented by the amount of data read from the network. If
186 QTYWRITTEN is non-NULL, the value it points to is incremented by
187 the amount of data written to disk. The time it took to download
188 the data (in milliseconds) is stored to ELAPSED.
190 The function exits and returns the amount of data read. In case of
191 error while reading data, -1 is returned. In case of error while
192 writing data, -2 is returned. */
195 fd_read_body (int fd, FILE *out, long toread, long startpos,
196 long *qtyread, long *qtywritten, double *elapsed, int flags)
200 static char dlbuf[16384];
201 int dlbufsize = sizeof (dlbuf);
203 struct wget_timer *timer = NULL;
204 double last_successful_read_tm = 0;
206 /* The progress gauge, set according to the user preferences. */
207 void *progress = NULL;
209 /* Non-zero if the progress gauge is interactive, i.e. if it can
210 continually update the display. When true, smaller timeout
211 values are used so that the gauge can update the display when
212 data arrives slowly. */
213 int progress_interactive = 0;
215 int exact = flags & rb_read_exactly;
218 /* How much data we've read/written. */
220 long sum_written = 0;
222 if (flags & rb_skip_startpos)
227 /* If we're skipping STARTPOS bytes, hide it from
228 progress_create because the indicator can't deal with it. */
229 progress = progress_create (skip ? 0 : 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 = wtimer_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)
263 /* For interactive progress gauges, always specify a ~1s
264 timeout, so that the gauge can be updated regularly even
265 when the data arrives very slowly or stalls. */
267 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
268 if (waittm + tmout > opt.read_timeout)
270 /* Don't let total idle time exceed read timeout. */
271 tmout = opt.read_timeout - waittm;
274 /* We've already exceeded the timeout. */
275 ret = -1, errno = ETIMEDOUT;
280 ret = fd_read (fd, dlbuf, rdsize, tmout);
282 if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
283 break; /* read error */
285 ret = 0; /* read timeout */
287 if (progress || opt.limit_rate)
289 wtimer_update (timer);
291 last_successful_read_tm = wtimer_read (timer);
297 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
305 limit_bandwidth (ret, timer);
308 progress_update (progress, ret, wtimer_read (timer));
311 ws_percenttitle (100.0 *
312 (startpos + sum_read) / (startpos + toread));
320 progress_finish (progress, wtimer_read (timer));
323 *elapsed = wtimer_read (timer);
325 wtimer_delete (timer);
328 *qtyread += sum_read;
330 *qtywritten += sum_written;
335 /* Read a hunk of data from FD, up until a terminator. The terminator
336 is whatever the TERMINATOR function determines it to be; for
337 example, it can be a line of data, or the head of an HTTP response.
338 The function returns the data read allocated with malloc.
340 In case of error, NULL is returned. In case of EOF and no data
341 read, NULL is returned and errno set to 0. In case of EOF with
342 data having been read, the data is returned, but it will
343 (obviously) not contain the terminator.
345 The idea is to be able to read a line of input, or otherwise a hunk
346 of text, such as the head of an HTTP request, without crossing the
347 boundary, so that the next call to fd_read etc. reads the data
348 after the hunk. To achieve that, this function does the following:
350 1. Peek at available data.
352 2. Determine whether the peeked data, along with the previously
353 read data, includes the terminator.
355 2a. If yes, read the data until the end of the terminator, and
358 2b. If no, read the peeked data and goto 1.
360 The function is careful to assume as little as possible about the
361 implementation of peeking. For example, every peek is followed by
362 a read. If the read returns a different amount of data, the
363 process is retried until all data arrives safely.
365 BUFSIZE is the size of the initial buffer expected to read all the
366 data in the typical case.
368 This function should be used as a building block for other
369 functions -- see fd_read_line as a simple example. */
372 fd_read_hunk (int fd, hunk_terminator_t hunk_terminator, int bufsize)
374 char *hunk = xmalloc (bufsize);
375 int tail = 0; /* tail position in HUNK */
380 int pklen, rdlen, remain;
382 /* First, peek at the available data. */
384 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
390 end = hunk_terminator (hunk, tail, pklen);
393 /* The data contains the terminator: we'll drain the data up
394 to the end of the terminator. */
395 remain = end - (hunk + tail);
398 /* No more data needs to be read. */
402 if (bufsize - 1 < tail + remain)
404 bufsize = tail + remain + 1;
405 hunk = xrealloc (hunk, bufsize);
409 /* No terminator: simply read the data we know is (or should
413 /* Now, read the data. Note that we make no assumptions about
414 how much data we'll get. (Some TCP stacks are notorious for
415 read returning less data than the previous MSG_PEEK.) */
417 rdlen = fd_read (fd, hunk + tail, remain, 0);
430 /* EOF without anything having been read */
436 /* EOF seen: return the data we've read. */
439 if (end && rdlen == remain)
440 /* The terminator was seen and the remaining data drained --
441 we got what we came for. */
444 /* Keep looping until all the data arrives. */
446 if (tail == bufsize - 1)
449 hunk = xrealloc (hunk, bufsize);
455 line_terminator (const char *hunk, int oldlen, int peeklen)
457 const char *p = memchr (hunk + oldlen, '\n', peeklen);
459 /* p+1 because we want the line to include '\n' */
464 /* Read one line from FD and return it. The line is allocated using
467 If an error occurs, or if no data can be read, NULL is returned.
468 In the former case errno indicates the error condition, and in the
469 latter case, errno is NULL. */
472 fd_read_line (int fd)
474 return fd_read_hunk (fd, line_terminator, 128);
477 /* Return a printed representation of the download rate, as
478 appropriate for the speed. If PAD is non-zero, strings will be
479 padded to the width of 7 characters (xxxx.xx). */
481 retr_rate (long bytes, double msecs, int pad)
484 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
487 double dlrate = calc_rate (bytes, msecs, &units);
488 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
493 /* Calculate the download rate and trim it as appropriate for the
494 speed. Appropriate means that if rate is greater than 1K/s,
495 kilobytes are used, and if rate is greater than 1MB/s, megabytes
498 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
501 calc_rate (long bytes, double msecs, int *units)
509 /* If elapsed time is exactly zero, it means we're under the
510 granularity of the timer. This often happens on systems that
511 use time() for the timer. */
512 msecs = wtimer_granularity ();
514 dlrate = (double)1000 * bytes / msecs;
517 else if (dlrate < 1024.0 * 1024.0)
518 *units = 1, dlrate /= 1024.0;
519 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
520 *units = 2, dlrate /= (1024.0 * 1024.0);
522 /* Maybe someone will need this, one day. */
523 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
528 /* Maximum number of allowed redirections. 20 was chosen as a
529 "reasonable" value, which is low enough to not cause havoc, yet
530 high enough to guarantee that normal retrievals will not be hurt by
533 #define MAX_REDIRECTIONS 20
535 #define SUSPEND_POST_DATA do { \
536 post_data_suspended = 1; \
537 saved_post_data = opt.post_data; \
538 saved_post_file_name = opt.post_file_name; \
539 opt.post_data = NULL; \
540 opt.post_file_name = NULL; \
543 #define RESTORE_POST_DATA do { \
544 if (post_data_suspended) \
546 opt.post_data = saved_post_data; \
547 opt.post_file_name = saved_post_file_name; \
548 post_data_suspended = 0; \
552 static char *getproxy PARAMS ((struct url *));
554 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
557 /* #### This function should be rewritten so it doesn't return from
561 retrieve_url (const char *origurl, char **file, char **newloc,
562 const char *refurl, int *dt)
566 int location_changed, dummy;
567 char *mynewloc, *proxy;
568 struct url *u, *proxy_url;
569 int up_error_code; /* url parse error code */
571 int redirection_count = 0;
573 int post_data_suspended = 0;
574 char *saved_post_data = NULL;
575 char *saved_post_file_name = NULL;
577 /* If dt is NULL, use local storage. */
583 url = xstrdup (origurl);
589 u = url_parse (url, &up_error_code);
592 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
598 refurl = opt.referer;
607 proxy = getproxy (u);
610 /* Parse the proxy URL. */
611 proxy_url = url_parse (proxy, &up_error_code);
614 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
615 proxy, url_error (up_error_code));
620 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
622 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
623 url_free (proxy_url);
630 if (u->scheme == SCHEME_HTTP
632 || u->scheme == SCHEME_HTTPS
634 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
636 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
638 else if (u->scheme == SCHEME_FTP)
640 /* If this is a redirection, we must not allow recursive FTP
641 retrieval, so we save recursion to oldrec, and restore it
643 int oldrec = opt.recursive;
644 if (redirection_count)
646 result = ftp_loop (u, dt, proxy_url);
647 opt.recursive = oldrec;
649 /* There is a possibility of having HTTP being redirected to
650 FTP. In these cases we must decide whether the text is HTML
651 according to the suffix. The HTML suffixes are `.html',
652 `.htm' and a few others, case-insensitive. */
653 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
655 if (has_html_suffix_p (local_file))
662 url_free (proxy_url);
666 location_changed = (result == NEWLOCATION);
667 if (location_changed)
669 char *construced_newloc;
670 struct url *newloc_parsed;
672 assert (mynewloc != NULL);
677 /* The HTTP specs only allow absolute URLs to appear in
678 redirects, but a ton of boneheaded webservers and CGIs out
679 there break the rules and use relative URLs, and popular
680 browsers are lenient about this, so wget should be too. */
681 construced_newloc = uri_merge (url, mynewloc);
683 mynewloc = construced_newloc;
685 /* Now, see if this new location makes sense. */
686 newloc_parsed = url_parse (mynewloc, &up_error_code);
689 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
690 url_error (up_error_code));
698 /* Now mynewloc will become newloc_parsed->url, because if the
699 Location contained relative paths like .././something, we
700 don't want that propagating as url. */
702 mynewloc = xstrdup (newloc_parsed->url);
704 /* Check for max. number of redirections. */
705 if (++redirection_count > MAX_REDIRECTIONS)
707 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
709 url_free (newloc_parsed);
722 /* If we're being redirected from POST, we don't want to POST
723 again. Many requests answer POST with a redirection to an
724 index page; that redirection is clearly a GET. We "suspend"
725 POST data for the duration of the redirections, and restore
726 it when we're done. */
727 if (!post_data_suspended)
737 register_download (u->url, local_file);
738 if (redirection_count && 0 != strcmp (origurl, u->url))
739 register_redirection (origurl, u->url);
741 register_html (u->url, local_file);
746 *file = local_file ? local_file : NULL;
748 xfree_null (local_file);
752 if (redirection_count)
771 /* Find the URLs in the file and call retrieve_url() for each of
772 them. If HTML is non-zero, treat the file as HTML, and construct
773 the URLs accordingly.
775 If opt.recursive is set, call retrieve_tree() for each file. */
778 retrieve_from_file (const char *file, int html, int *count)
781 struct urlpos *url_list, *cur_url;
783 url_list = (html ? get_urls_html (file, NULL, NULL)
784 : get_urls_file (file));
785 status = RETROK; /* Suppose everything is OK. */
786 *count = 0; /* Reset the URL count. */
788 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
790 char *filename = NULL, *new_file = NULL;
793 if (cur_url->ignore_when_downloading)
796 if (opt.quota && total_downloaded_bytes > opt.quota)
801 if ((opt.recursive || opt.page_requisites)
802 && cur_url->url->scheme != SCHEME_FTP)
803 status = retrieve_tree (cur_url->url->url);
805 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
807 if (filename && opt.delete_after && file_exists_p (filename))
809 DEBUGP (("Removing file due to --delete-after in"
810 " retrieve_from_file():\n"));
811 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
812 if (unlink (filename))
813 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
817 xfree_null (new_file);
818 xfree_null (filename);
821 /* Free the linked list of URL-s. */
822 free_urlpos (url_list);
827 /* Print `giving up', or `retrying', depending on the impending
828 action. N1 and N2 are the attempt number and the attempt limit. */
830 printwhat (int n1, int n2)
832 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
835 /* If opt.wait or opt.waitretry are specified, and if certain
836 conditions are met, sleep the appropriate number of seconds. See
837 the documentation of --wait and --waitretry for more information.
839 COUNT is the count of current retrieval, beginning with 1. */
842 sleep_between_retrievals (int count)
844 static int first_retrieval = 1;
848 /* Don't sleep before the very first retrieval. */
853 if (opt.waitretry && count > 1)
855 /* If opt.waitretry is specified and this is a retry, wait for
856 COUNT-1 number of seconds, or for opt.waitretry seconds. */
857 if (count <= opt.waitretry)
860 xsleep (opt.waitretry);
864 if (!opt.random_wait || count > 1)
865 /* If random-wait is not specified, or if we are sleeping
866 between retries of the same download, sleep the fixed
871 /* Sleep a random amount of time averaging in opt.wait
872 seconds. The sleeping amount ranges from 0 to
873 opt.wait*2, inclusive. */
874 double waitsecs = 2 * opt.wait * random_float ();
875 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
876 opt.wait, waitsecs));
882 /* Free the linked list of urlpos. */
884 free_urlpos (struct urlpos *l)
888 struct urlpos *next = l->next;
891 xfree_null (l->local_name);
897 /* Rotate FNAME opt.backups times */
899 rotate_backups(const char *fname)
901 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
902 char *from = (char *)alloca (maxlen);
903 char *to = (char *)alloca (maxlen);
907 if (stat (fname, &sb) == 0)
908 if (S_ISREG (sb.st_mode) == 0)
911 for (i = opt.backups; i > 1; i--)
913 sprintf (from, "%s.%d", fname, i - 1);
914 sprintf (to, "%s.%d", fname, i);
918 sprintf (to, "%s.%d", fname, 1);
922 static int no_proxy_match PARAMS ((const char *, const char **));
924 /* Return the URL of the proxy appropriate for url U. */
927 getproxy (struct url *u)
931 static char rewritten_storage[1024];
935 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
941 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
945 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
949 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
954 if (!proxy || !*proxy)
957 /* Handle shorthands. `rewritten_storage' is a kludge to allow
958 getproxy() to return static storage. */
959 rewritten_url = rewrite_shorthand_url (proxy);
962 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
963 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
964 proxy = rewritten_storage;
970 /* Should a host be accessed through proxy, concerning no_proxy? */
972 no_proxy_match (const char *host, const char **no_proxy)
977 return !sufmatch (no_proxy, host);