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, pass 0 as the INITIAL
228 argument to progress_create because the indicator doesn't
229 (yet) know about "skipping" data. */
230 progress = progress_create (skip ? 0 : startpos, startpos + toread);
231 progress_interactive = progress_interactive_p (progress);
235 limit_bandwidth_reset ();
237 /* A timer is needed for tracking progress, for throttling, and for
238 tracking elapsed time. If either of these are requested, start
240 if (progress || opt.limit_rate || elapsed)
242 timer = wtimer_new ();
243 last_successful_read_tm = 0;
246 /* Use a smaller buffer for low requested bandwidths. For example,
247 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
248 data and then sleep for 8s. With buffer size equal to the limit,
249 we never have to sleep for more than one second. */
250 if (opt.limit_rate && opt.limit_rate < dlbufsize)
251 dlbufsize = opt.limit_rate;
253 /* Read from FD while there is data to read. Normally toread==0
254 means that it is unknown how much data is to arrive. However, if
255 EXACT is set, then toread==0 means what it says: that no data
257 while (!exact || (sum_read < toread))
259 int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
260 double tmout = opt.read_timeout;
261 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 if (opt.read_timeout)
270 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
271 if (waittm + tmout > opt.read_timeout)
273 /* Don't let total idle time exceed read timeout. */
274 tmout = opt.read_timeout - waittm;
277 /* We've already exceeded the timeout. */
278 ret = -1, errno = ETIMEDOUT;
284 ret = fd_read (fd, dlbuf, rdsize, tmout);
286 /* when retrieving from http-proxy wget sometimes does not trust the
287 * file length reported by server.
288 * this check is to tell wget not to stubbornly try to read again and
289 * again until another errno code was received. */
290 if ( ret == -1 && errno == ETIMEDOUT && sum_read == toread && toread > 0 )
293 if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
294 break; /* read error */
296 ret = 0; /* read timeout */
298 if (progress || opt.limit_rate)
300 wtimer_update (timer);
302 last_successful_read_tm = wtimer_read (timer);
308 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
316 limit_bandwidth (ret, timer);
319 progress_update (progress, ret, wtimer_read (timer));
321 if (toread > 0 && !opt.quiet)
322 ws_percenttitle (100.0 *
323 (startpos + sum_read) / (startpos + toread));
331 progress_finish (progress, wtimer_read (timer));
334 *elapsed = wtimer_read (timer);
336 wtimer_delete (timer);
339 *qtyread += sum_read;
341 *qtywritten += sum_written;
346 /* Read a hunk of data from FD, up until a terminator. The terminator
347 is whatever the TERMINATOR function determines it to be; for
348 example, it can be a line of data, or the head of an HTTP response.
349 The function returns the data read allocated with malloc.
351 In case of error, NULL is returned. In case of EOF and no data
352 read, NULL is returned and errno set to 0. In case of EOF with
353 data having been read, the data is returned, but it will
354 (obviously) not contain the terminator.
356 The idea is to be able to read a line of input, or otherwise a hunk
357 of text, such as the head of an HTTP request, without crossing the
358 boundary, so that the next call to fd_read etc. reads the data
359 after the hunk. To achieve that, this function does the following:
361 1. Peek at available data.
363 2. Determine whether the peeked data, along with the previously
364 read data, includes the terminator.
366 2a. If yes, read the data until the end of the terminator, and
369 2b. If no, read the peeked data and goto 1.
371 The function is careful to assume as little as possible about the
372 implementation of peeking. For example, every peek is followed by
373 a read. If the read returns a different amount of data, the
374 process is retried until all data arrives safely.
376 BUFSIZE is the size of the initial buffer expected to read all the
377 data in the typical case.
379 This function should be used as a building block for other
380 functions -- see fd_read_line as a simple example. */
383 fd_read_hunk (int fd, hunk_terminator_t hunk_terminator, int bufsize)
385 char *hunk = xmalloc (bufsize);
386 int tail = 0; /* tail position in HUNK */
391 int pklen, rdlen, remain;
393 /* First, peek at the available data. */
395 pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
401 end = hunk_terminator (hunk, tail, pklen);
404 /* The data contains the terminator: we'll drain the data up
405 to the end of the terminator. */
406 remain = end - (hunk + tail);
409 /* No more data needs to be read. */
413 if (bufsize - 1 < tail + remain)
415 bufsize = tail + remain + 1;
416 hunk = xrealloc (hunk, bufsize);
420 /* No terminator: simply read the data we know is (or should
424 /* Now, read the data. Note that we make no assumptions about
425 how much data we'll get. (Some TCP stacks are notorious for
426 read returning less data than the previous MSG_PEEK.) */
428 rdlen = fd_read (fd, hunk + tail, remain, 0);
441 /* EOF without anything having been read */
447 /* EOF seen: return the data we've read. */
450 if (end && rdlen == remain)
451 /* The terminator was seen and the remaining data drained --
452 we got what we came for. */
455 /* Keep looping until all the data arrives. */
457 if (tail == bufsize - 1)
460 hunk = xrealloc (hunk, bufsize);
466 line_terminator (const char *hunk, int oldlen, int peeklen)
468 const char *p = memchr (hunk + oldlen, '\n', peeklen);
470 /* p+1 because we want the line to include '\n' */
475 /* Read one line from FD and return it. The line is allocated using
478 If an error occurs, or if no data can be read, NULL is returned.
479 In the former case errno indicates the error condition, and in the
480 latter case, errno is NULL. */
483 fd_read_line (int fd)
485 return fd_read_hunk (fd, line_terminator, 128);
488 /* Return a printed representation of the download rate, as
489 appropriate for the speed. If PAD is non-zero, strings will be
490 padded to the width of 7 characters (xxxx.xx). */
492 retr_rate (long bytes, double msecs, int pad)
495 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
498 double dlrate = calc_rate (bytes, msecs, &units);
499 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
504 /* Calculate the download rate and trim it as appropriate for the
505 speed. Appropriate means that if rate is greater than 1K/s,
506 kilobytes are used, and if rate is greater than 1MB/s, megabytes
509 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
512 calc_rate (long bytes, double msecs, int *units)
520 /* If elapsed time is exactly zero, it means we're under the
521 granularity of the timer. This often happens on systems that
522 use time() for the timer. */
523 msecs = wtimer_granularity ();
525 dlrate = (double)1000 * bytes / msecs;
528 else if (dlrate < 1024.0 * 1024.0)
529 *units = 1, dlrate /= 1024.0;
530 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
531 *units = 2, dlrate /= (1024.0 * 1024.0);
533 /* Maybe someone will need this, one day. */
534 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
539 /* Maximum number of allowed redirections. 20 was chosen as a
540 "reasonable" value, which is low enough to not cause havoc, yet
541 high enough to guarantee that normal retrievals will not be hurt by
544 #define MAX_REDIRECTIONS 20
546 #define SUSPEND_POST_DATA do { \
547 post_data_suspended = 1; \
548 saved_post_data = opt.post_data; \
549 saved_post_file_name = opt.post_file_name; \
550 opt.post_data = NULL; \
551 opt.post_file_name = NULL; \
554 #define RESTORE_POST_DATA do { \
555 if (post_data_suspended) \
557 opt.post_data = saved_post_data; \
558 opt.post_file_name = saved_post_file_name; \
559 post_data_suspended = 0; \
563 static char *getproxy PARAMS ((struct url *));
565 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
568 /* #### This function should be rewritten so it doesn't return from
572 retrieve_url (const char *origurl, char **file, char **newloc,
573 const char *refurl, int *dt)
577 int location_changed, dummy;
578 char *mynewloc, *proxy;
579 struct url *u, *proxy_url;
580 int up_error_code; /* url parse error code */
582 int redirection_count = 0;
584 int post_data_suspended = 0;
585 char *saved_post_data = NULL;
586 char *saved_post_file_name = NULL;
588 /* If dt is NULL, use local storage. */
594 url = xstrdup (origurl);
600 u = url_parse (url, &up_error_code);
603 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
609 refurl = opt.referer;
618 proxy = getproxy (u);
621 /* Parse the proxy URL. */
622 proxy_url = url_parse (proxy, &up_error_code);
625 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
626 proxy, url_error (up_error_code));
631 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
633 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
634 url_free (proxy_url);
641 if (u->scheme == SCHEME_HTTP
643 || u->scheme == SCHEME_HTTPS
645 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
647 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
649 else if (u->scheme == SCHEME_FTP)
651 /* If this is a redirection, we must not allow recursive FTP
652 retrieval, so we save recursion to oldrec, and restore it
654 int oldrec = opt.recursive;
655 if (redirection_count)
657 result = ftp_loop (u, dt, proxy_url);
658 opt.recursive = oldrec;
660 /* There is a possibility of having HTTP being redirected to
661 FTP. In these cases we must decide whether the text is HTML
662 according to the suffix. The HTML suffixes are `.html',
663 `.htm' and a few others, case-insensitive. */
664 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
666 if (has_html_suffix_p (local_file))
673 url_free (proxy_url);
677 location_changed = (result == NEWLOCATION);
678 if (location_changed)
680 char *construced_newloc;
681 struct url *newloc_parsed;
683 assert (mynewloc != NULL);
688 /* The HTTP specs only allow absolute URLs to appear in
689 redirects, but a ton of boneheaded webservers and CGIs out
690 there break the rules and use relative URLs, and popular
691 browsers are lenient about this, so wget should be too. */
692 construced_newloc = uri_merge (url, mynewloc);
694 mynewloc = construced_newloc;
696 /* Now, see if this new location makes sense. */
697 newloc_parsed = url_parse (mynewloc, &up_error_code);
700 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
701 url_error (up_error_code));
709 /* Now mynewloc will become newloc_parsed->url, because if the
710 Location contained relative paths like .././something, we
711 don't want that propagating as url. */
713 mynewloc = xstrdup (newloc_parsed->url);
715 /* Check for max. number of redirections. */
716 if (++redirection_count > MAX_REDIRECTIONS)
718 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
720 url_free (newloc_parsed);
733 /* If we're being redirected from POST, we don't want to POST
734 again. Many requests answer POST with a redirection to an
735 index page; that redirection is clearly a GET. We "suspend"
736 POST data for the duration of the redirections, and restore
737 it when we're done. */
738 if (!post_data_suspended)
748 register_download (u->url, local_file);
749 if (redirection_count && 0 != strcmp (origurl, u->url))
750 register_redirection (origurl, u->url);
752 register_html (u->url, local_file);
757 *file = local_file ? local_file : NULL;
759 xfree_null (local_file);
763 if (redirection_count)
782 /* Find the URLs in the file and call retrieve_url() for each of
783 them. If HTML is non-zero, treat the file as HTML, and construct
784 the URLs accordingly.
786 If opt.recursive is set, call retrieve_tree() for each file. */
789 retrieve_from_file (const char *file, int html, int *count)
792 struct urlpos *url_list, *cur_url;
794 url_list = (html ? get_urls_html (file, NULL, NULL)
795 : get_urls_file (file));
796 status = RETROK; /* Suppose everything is OK. */
797 *count = 0; /* Reset the URL count. */
799 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
801 char *filename = NULL, *new_file = NULL;
804 if (cur_url->ignore_when_downloading)
807 if (opt.quota && total_downloaded_bytes > opt.quota)
812 if ((opt.recursive || opt.page_requisites)
813 && cur_url->url->scheme != SCHEME_FTP)
814 status = retrieve_tree (cur_url->url->url);
816 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
818 if (filename && opt.delete_after && file_exists_p (filename))
820 DEBUGP (("Removing file due to --delete-after in"
821 " retrieve_from_file():\n"));
822 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
823 if (unlink (filename))
824 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
828 xfree_null (new_file);
829 xfree_null (filename);
832 /* Free the linked list of URL-s. */
833 free_urlpos (url_list);
838 /* Print `giving up', or `retrying', depending on the impending
839 action. N1 and N2 are the attempt number and the attempt limit. */
841 printwhat (int n1, int n2)
843 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
846 /* If opt.wait or opt.waitretry are specified, and if certain
847 conditions are met, sleep the appropriate number of seconds. See
848 the documentation of --wait and --waitretry for more information.
850 COUNT is the count of current retrieval, beginning with 1. */
853 sleep_between_retrievals (int count)
855 static int first_retrieval = 1;
859 /* Don't sleep before the very first retrieval. */
864 if (opt.waitretry && count > 1)
866 /* If opt.waitretry is specified and this is a retry, wait for
867 COUNT-1 number of seconds, or for opt.waitretry seconds. */
868 if (count <= opt.waitretry)
871 xsleep (opt.waitretry);
875 if (!opt.random_wait || count > 1)
876 /* If random-wait is not specified, or if we are sleeping
877 between retries of the same download, sleep the fixed
882 /* Sleep a random amount of time averaging in opt.wait
883 seconds. The sleeping amount ranges from 0 to
884 opt.wait*2, inclusive. */
885 double waitsecs = 2 * opt.wait * random_float ();
886 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
887 opt.wait, waitsecs));
893 /* Free the linked list of urlpos. */
895 free_urlpos (struct urlpos *l)
899 struct urlpos *next = l->next;
902 xfree_null (l->local_name);
908 /* Rotate FNAME opt.backups times */
910 rotate_backups(const char *fname)
912 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
913 char *from = (char *)alloca (maxlen);
914 char *to = (char *)alloca (maxlen);
918 if (stat (fname, &sb) == 0)
919 if (S_ISREG (sb.st_mode) == 0)
922 for (i = opt.backups; i > 1; i--)
924 sprintf (from, "%s.%d", fname, i - 1);
925 sprintf (to, "%s.%d", fname, i);
929 sprintf (to, "%s.%d", fname, 1);
933 static int no_proxy_match PARAMS ((const char *, const char **));
935 /* Return the URL of the proxy appropriate for url U. */
938 getproxy (struct url *u)
942 static char rewritten_storage[1024];
946 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
952 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
956 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
960 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
965 if (!proxy || !*proxy)
968 /* Handle shorthands. `rewritten_storage' is a kludge to allow
969 getproxy() to return static storage. */
970 rewritten_url = rewrite_shorthand_url (proxy);
973 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
974 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
975 proxy = rewritten_storage;
981 /* Should a host be accessed through proxy, concerning no_proxy? */
983 no_proxy_match (const char *host, const char **no_proxy)
988 return !sufmatch (no_proxy, host);