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 (wgint 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 (%s/%.2f).\n",
114 slp, number_to_static_string (limit_data.chunk_bytes),
118 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
119 slp, number_to_static_string (limit_data.chunk_bytes),
120 limit_data.sleep_adjust));
122 t0 = wtimer_read (timer);
124 wtimer_update (timer);
125 t1 = wtimer_read (timer);
127 /* Due to scheduling, we probably slept slightly longer (or
128 shorter) than desired. Calculate the difference between the
129 desired and the actual sleep, and adjust the next sleep by
131 limit_data.sleep_adjust = slp - (t1 - t0);
134 limit_data.chunk_bytes = 0;
135 limit_data.chunk_start = wtimer_read (timer);
139 # define MIN(i, j) ((i) <= (j) ? (i) : (j))
142 /* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
143 amount of data and decrease SKIP. Increment *TOTAL by the amount
147 write_data (FILE *out, const char *buf, int bufsize, wgint *skip,
166 fwrite (buf, 1, bufsize, out);
169 /* Immediately flush the downloaded data. This should not hinder
170 performance: fast downloads will arrive in large 16K chunks
171 (which stdio would write out immediately anyway), and slow
172 downloads wouldn't be limited by disk speed. */
174 return !ferror (out);
177 /* Read the contents of file descriptor FD until it the connection
178 terminates or a read error occurs. The data is read in portions of
179 up to 16K and written to OUT as it arrives. If opt.verbose is set,
180 the progress is shown.
182 TOREAD is the amount of data expected to arrive, normally only used
183 by the progress gauge.
185 STARTPOS is the position from which the download starts, used by
186 the progress gauge. If QTYREAD is non-NULL, the value it points to
187 is incremented by the amount of data read from the network. If
188 QTYWRITTEN is non-NULL, the value it points to is incremented by
189 the amount of data written to disk. The time it took to download
190 the data (in milliseconds) is stored to ELAPSED.
192 The function exits and returns the amount of data read. In case of
193 error while reading data, -1 is returned. In case of error while
194 writing data, -2 is returned. */
197 fd_read_body (int fd, FILE *out, wgint toread, wgint startpos,
198 wgint *qtyread, wgint *qtywritten, double *elapsed, int flags)
202 static char dlbuf[16384];
203 int dlbufsize = sizeof (dlbuf);
205 struct wget_timer *timer = NULL;
206 double last_successful_read_tm = 0;
208 /* The progress gauge, set according to the user preferences. */
209 void *progress = NULL;
211 /* Non-zero if the progress gauge is interactive, i.e. if it can
212 continually update the display. When true, smaller timeout
213 values are used so that the gauge can update the display when
214 data arrives slowly. */
215 int progress_interactive = 0;
217 int exact = flags & rb_read_exactly;
220 /* How much data we've read/written. */
222 wgint sum_written = 0;
224 if (flags & rb_skip_startpos)
229 /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
230 argument to progress_create because the indicator doesn't
231 (yet) know about "skipping" data. */
232 progress = progress_create (skip ? 0 : startpos, startpos + toread);
233 progress_interactive = progress_interactive_p (progress);
237 limit_bandwidth_reset ();
239 /* A timer is needed for tracking progress, for throttling, and for
240 tracking elapsed time. If either of these are requested, start
242 if (progress || opt.limit_rate || elapsed)
244 timer = wtimer_new ();
245 last_successful_read_tm = 0;
248 /* Use a smaller buffer for low requested bandwidths. For example,
249 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
250 data and then sleep for 8s. With buffer size equal to the limit,
251 we never have to sleep for more than one second. */
252 if (opt.limit_rate && opt.limit_rate < dlbufsize)
253 dlbufsize = opt.limit_rate;
255 /* Read from FD while there is data to read. Normally toread==0
256 means that it is unknown how much data is to arrive. However, if
257 EXACT is set, then toread==0 means what it says: that no data
259 while (!exact || (sum_read < toread))
261 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
262 double tmout = opt.read_timeout;
263 if (progress_interactive)
265 /* For interactive progress gauges, always specify a ~1s
266 timeout, so that the gauge can be updated regularly even
267 when the data arrives very slowly or stalls. */
269 if (opt.read_timeout)
272 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
273 if (waittm + tmout > opt.read_timeout)
275 /* Don't let total idle time exceed read timeout. */
276 tmout = opt.read_timeout - waittm;
279 /* We've already exceeded the timeout. */
280 ret = -1, errno = ETIMEDOUT;
286 ret = fd_read (fd, dlbuf, rdsize, tmout);
288 /* when retrieving from http-proxy wget sometimes does not trust the
289 * file length reported by server.
290 * this check is to tell wget not to stubbornly try to read again and
291 * again until another errno code was received. */
292 if ( ret == -1 && errno == ETIMEDOUT && sum_read == toread && toread > 0 )
295 if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
296 break; /* read error */
298 ret = 0; /* read timeout */
300 if (progress || opt.limit_rate)
302 wtimer_update (timer);
304 last_successful_read_tm = wtimer_read (timer);
310 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
318 limit_bandwidth (ret, timer);
321 progress_update (progress, ret, wtimer_read (timer));
323 if (toread > 0 && !opt.quiet)
324 ws_percenttitle (100.0 *
325 (startpos + sum_read) / (startpos + toread));
333 progress_finish (progress, wtimer_read (timer));
336 *elapsed = wtimer_read (timer);
338 wtimer_delete (timer);
341 *qtyread += sum_read;
343 *qtywritten += sum_written;
348 /* Read a hunk of data from FD, up until a terminator. The terminator
349 is whatever the TERMINATOR function determines it to be; for
350 example, it can be a line of data, or the head of an HTTP response.
351 The function returns the data read allocated with malloc.
353 In case of error, NULL is returned. In case of EOF and no data
354 read, NULL is returned and errno set to 0. In case of EOF with
355 data having been read, the data is returned, but it will
356 (obviously) not contain the terminator.
358 The idea is to be able to read a line of input, or otherwise a hunk
359 of text, such as the head of an HTTP request, without crossing the
360 boundary, so that the next call to fd_read etc. reads the data
361 after the hunk. To achieve that, this function does the following:
363 1. Peek at available data.
365 2. Determine whether the peeked data, along with the previously
366 read data, includes the terminator.
368 2a. If yes, read the data until the end of the terminator, and
371 2b. If no, read the peeked data and goto 1.
373 The function is careful to assume as little as possible about the
374 implementation of peeking. For example, every peek is followed by
375 a read. If the read returns a different amount of data, the
376 process is retried until all data arrives safely.
378 SIZEHINT is the buffer size sufficient to hold all the data in the
379 typical case (it is used as the initial buffer size). MAXSIZE is
380 the maximum amount of memory this function is allowed to allocate,
381 or 0 if no upper limit is to be enforced.
383 This function should be used as a building block for other
384 functions -- see fd_read_line as a simple example. */
387 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
389 long bufsize = sizehint;
390 char *hunk = xmalloc (bufsize);
391 int tail = 0; /* tail position in HUNK */
393 assert (maxsize >= bufsize);
398 int pklen, rdlen, remain;
400 /* First, peek at the available data. */
402 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
408 end = terminator (hunk, tail, pklen);
411 /* The data contains the terminator: we'll drain the data up
412 to the end of the terminator. */
413 remain = end - (hunk + tail);
416 /* No more data needs to be read. */
420 if (bufsize - 1 < tail + remain)
422 bufsize = tail + remain + 1;
423 hunk = xrealloc (hunk, bufsize);
427 /* No terminator: simply read the data we know is (or should
431 /* Now, read the data. Note that we make no assumptions about
432 how much data we'll get. (Some TCP stacks are notorious for
433 read returning less data than the previous MSG_PEEK.) */
435 rdlen = fd_read (fd, hunk + tail, remain, 0);
448 /* EOF without anything having been read */
454 /* EOF seen: return the data we've read. */
457 if (end && rdlen == remain)
458 /* The terminator was seen and the remaining data drained --
459 we got what we came for. */
462 /* Keep looping until all the data arrives. */
464 if (tail == bufsize - 1)
466 /* Double the buffer size, but refuse to allocate more than
468 if (maxsize && bufsize >= maxsize)
475 if (maxsize && bufsize > maxsize)
477 hunk = xrealloc (hunk, bufsize);
483 line_terminator (const char *hunk, int oldlen, int peeklen)
485 const char *p = memchr (hunk + oldlen, '\n', peeklen);
487 /* p+1 because we want the line to include '\n' */
492 /* The maximum size of the single line we agree to accept. This is
493 not meant to impose an arbitrary limit, but to protect the user
494 from Wget slurping up available memory upon encountering malicious
495 or buggy server output. Define it to 0 to remove the limit. */
496 #define FD_READ_LINE_MAX 4096
498 /* Read one line from FD and return it. The line is allocated using
499 malloc, but is never larger than FD_READ_LINE_MAX.
501 If an error occurs, or if no data can be read, NULL is returned.
502 In the former case errno indicates the error condition, and in the
503 latter case, errno is NULL. */
506 fd_read_line (int fd)
508 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
511 /* Return a printed representation of the download rate, as
512 appropriate for the speed. If PAD is non-zero, strings will be
513 padded to the width of 7 characters (xxxx.xx). */
515 retr_rate (wgint bytes, double msecs, int pad)
518 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
521 double dlrate = calc_rate (bytes, msecs, &units);
522 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
527 /* Calculate the download rate and trim it as appropriate for the
528 speed. Appropriate means that if rate is greater than 1K/s,
529 kilobytes are used, and if rate is greater than 1MB/s, megabytes
532 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
535 calc_rate (wgint bytes, double msecs, int *units)
543 /* If elapsed time is exactly zero, it means we're under the
544 granularity of the timer. This often happens on systems that
545 use time() for the timer. */
546 msecs = wtimer_granularity ();
548 dlrate = (double)1000 * bytes / msecs;
551 else if (dlrate < 1024.0 * 1024.0)
552 *units = 1, dlrate /= 1024.0;
553 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
554 *units = 2, dlrate /= (1024.0 * 1024.0);
556 /* Maybe someone will need this, one day. */
557 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
562 /* Maximum number of allowed redirections. 20 was chosen as a
563 "reasonable" value, which is low enough to not cause havoc, yet
564 high enough to guarantee that normal retrievals will not be hurt by
567 #define MAX_REDIRECTIONS 20
569 #define SUSPEND_POST_DATA do { \
570 post_data_suspended = 1; \
571 saved_post_data = opt.post_data; \
572 saved_post_file_name = opt.post_file_name; \
573 opt.post_data = NULL; \
574 opt.post_file_name = NULL; \
577 #define RESTORE_POST_DATA do { \
578 if (post_data_suspended) \
580 opt.post_data = saved_post_data; \
581 opt.post_file_name = saved_post_file_name; \
582 post_data_suspended = 0; \
586 static char *getproxy PARAMS ((struct url *));
588 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
591 /* #### This function should be rewritten so it doesn't return from
595 retrieve_url (const char *origurl, char **file, char **newloc,
596 const char *refurl, int *dt)
600 int location_changed, dummy;
601 char *mynewloc, *proxy;
602 struct url *u, *proxy_url;
603 int up_error_code; /* url parse error code */
605 int redirection_count = 0;
607 int post_data_suspended = 0;
608 char *saved_post_data = NULL;
609 char *saved_post_file_name = NULL;
611 /* If dt is NULL, use local storage. */
617 url = xstrdup (origurl);
623 u = url_parse (url, &up_error_code);
626 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
632 refurl = opt.referer;
641 proxy = getproxy (u);
644 /* Parse the proxy URL. */
645 proxy_url = url_parse (proxy, &up_error_code);
648 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
649 proxy, url_error (up_error_code));
654 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
656 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
657 url_free (proxy_url);
664 if (u->scheme == SCHEME_HTTP
666 || u->scheme == SCHEME_HTTPS
668 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
670 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
672 else if (u->scheme == SCHEME_FTP)
674 /* If this is a redirection, we must not allow recursive FTP
675 retrieval, so we save recursion to oldrec, and restore it
677 int oldrec = opt.recursive;
678 if (redirection_count)
680 result = ftp_loop (u, dt, proxy_url);
681 opt.recursive = oldrec;
683 /* There is a possibility of having HTTP being redirected to
684 FTP. In these cases we must decide whether the text is HTML
685 according to the suffix. The HTML suffixes are `.html',
686 `.htm' and a few others, case-insensitive. */
687 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
689 if (has_html_suffix_p (local_file))
696 url_free (proxy_url);
700 location_changed = (result == NEWLOCATION);
701 if (location_changed)
703 char *construced_newloc;
704 struct url *newloc_parsed;
706 assert (mynewloc != NULL);
711 /* The HTTP specs only allow absolute URLs to appear in
712 redirects, but a ton of boneheaded webservers and CGIs out
713 there break the rules and use relative URLs, and popular
714 browsers are lenient about this, so wget should be too. */
715 construced_newloc = uri_merge (url, mynewloc);
717 mynewloc = construced_newloc;
719 /* Now, see if this new location makes sense. */
720 newloc_parsed = url_parse (mynewloc, &up_error_code);
723 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
724 url_error (up_error_code));
732 /* Now mynewloc will become newloc_parsed->url, because if the
733 Location contained relative paths like .././something, we
734 don't want that propagating as url. */
736 mynewloc = xstrdup (newloc_parsed->url);
738 /* Check for max. number of redirections. */
739 if (++redirection_count > MAX_REDIRECTIONS)
741 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
743 url_free (newloc_parsed);
756 /* If we're being redirected from POST, we don't want to POST
757 again. Many requests answer POST with a redirection to an
758 index page; that redirection is clearly a GET. We "suspend"
759 POST data for the duration of the redirections, and restore
760 it when we're done. */
761 if (!post_data_suspended)
771 register_download (u->url, local_file);
772 if (redirection_count && 0 != strcmp (origurl, u->url))
773 register_redirection (origurl, u->url);
775 register_html (u->url, local_file);
780 *file = local_file ? local_file : NULL;
782 xfree_null (local_file);
786 if (redirection_count)
805 /* Find the URLs in the file and call retrieve_url() for each of
806 them. If HTML is non-zero, treat the file as HTML, and construct
807 the URLs accordingly.
809 If opt.recursive is set, call retrieve_tree() for each file. */
812 retrieve_from_file (const char *file, int html, int *count)
815 struct urlpos *url_list, *cur_url;
817 url_list = (html ? get_urls_html (file, NULL, NULL)
818 : get_urls_file (file));
819 status = RETROK; /* Suppose everything is OK. */
820 *count = 0; /* Reset the URL count. */
822 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
824 char *filename = NULL, *new_file = NULL;
827 if (cur_url->ignore_when_downloading)
830 if (opt.quota && total_downloaded_bytes > opt.quota)
835 if ((opt.recursive || opt.page_requisites)
836 && cur_url->url->scheme != SCHEME_FTP)
837 status = retrieve_tree (cur_url->url->url);
839 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
841 if (filename && opt.delete_after && file_exists_p (filename))
843 DEBUGP (("Removing file due to --delete-after in"
844 " retrieve_from_file():\n"));
845 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
846 if (unlink (filename))
847 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
851 xfree_null (new_file);
852 xfree_null (filename);
855 /* Free the linked list of URL-s. */
856 free_urlpos (url_list);
861 /* Print `giving up', or `retrying', depending on the impending
862 action. N1 and N2 are the attempt number and the attempt limit. */
864 printwhat (int n1, int n2)
866 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
869 /* If opt.wait or opt.waitretry are specified, and if certain
870 conditions are met, sleep the appropriate number of seconds. See
871 the documentation of --wait and --waitretry for more information.
873 COUNT is the count of current retrieval, beginning with 1. */
876 sleep_between_retrievals (int count)
878 static int first_retrieval = 1;
882 /* Don't sleep before the very first retrieval. */
887 if (opt.waitretry && count > 1)
889 /* If opt.waitretry is specified and this is a retry, wait for
890 COUNT-1 number of seconds, or for opt.waitretry seconds. */
891 if (count <= opt.waitretry)
894 xsleep (opt.waitretry);
898 if (!opt.random_wait || count > 1)
899 /* If random-wait is not specified, or if we are sleeping
900 between retries of the same download, sleep the fixed
905 /* Sleep a random amount of time averaging in opt.wait
906 seconds. The sleeping amount ranges from 0 to
907 opt.wait*2, inclusive. */
908 double waitsecs = 2 * opt.wait * random_float ();
909 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
910 opt.wait, waitsecs));
916 /* Free the linked list of urlpos. */
918 free_urlpos (struct urlpos *l)
922 struct urlpos *next = l->next;
925 xfree_null (l->local_name);
931 /* Rotate FNAME opt.backups times */
933 rotate_backups(const char *fname)
935 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
936 char *from = (char *)alloca (maxlen);
937 char *to = (char *)alloca (maxlen);
941 if (stat (fname, &sb) == 0)
942 if (S_ISREG (sb.st_mode) == 0)
945 for (i = opt.backups; i > 1; i--)
947 sprintf (from, "%s.%d", fname, i - 1);
948 sprintf (to, "%s.%d", fname, i);
952 sprintf (to, "%s.%d", fname, 1);
956 static int no_proxy_match PARAMS ((const char *, const char **));
958 /* Return the URL of the proxy appropriate for url U. */
961 getproxy (struct url *u)
965 static char rewritten_storage[1024];
969 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
975 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
979 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
983 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
988 if (!proxy || !*proxy)
991 /* Handle shorthands. `rewritten_storage' is a kludge to allow
992 getproxy() to return static storage. */
993 rewritten_url = rewrite_shorthand_url (proxy);
996 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
997 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
998 proxy = rewritten_storage;
1004 /* Should a host be accessed through proxy, concerning no_proxy? */
1006 no_proxy_match (const char *host, const char **no_proxy)
1011 return !sufmatch (no_proxy, host);