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 */
60 # include "gen_sslfunc.h" /* for ssl_iread */
67 /* Total size of downloaded files. Used to enforce quota. */
68 LARGE_INT total_downloaded_bytes;
70 /* If non-NULL, the stream to which output should be written. This
71 stream is initialized when `-O' is used. */
74 /* Whether output_document is a regular file we can manipulate,
75 i.e. not `-' or a device file. */
76 int output_stream_regular;
85 limit_bandwidth_reset (void)
87 limit_data.chunk_bytes = 0;
88 limit_data.chunk_start = 0;
91 /* Limit the bandwidth by pausing the download for an amount of time.
92 BYTES is the number of bytes received from the network, and TIMER
93 is the timer that started at the beginning of download. */
96 limit_bandwidth (wgint bytes, struct ptimer *timer)
98 double delta_t = ptimer_read (timer) - limit_data.chunk_start;
101 limit_data.chunk_bytes += bytes;
103 /* Calculate the amount of time we expect downloading the chunk
104 should take. If in reality it took less time, sleep to
105 compensate for the difference. */
106 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
108 if (expected > delta_t)
110 double slp = expected - delta_t + limit_data.sleep_adjust;
114 DEBUGP (("deferring a %.2f ms sleep (%s/%.2f).\n",
115 slp, number_to_static_string (limit_data.chunk_bytes),
119 DEBUGP (("\nsleeping %.2f ms for %s bytes, adjust %.2f ms\n",
120 slp, number_to_static_string (limit_data.chunk_bytes),
121 limit_data.sleep_adjust));
123 t0 = ptimer_read (timer);
125 t1 = ptimer_measure (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 = ptimer_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 ptimer *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 = ptimer_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 = (ptimer_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 if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
289 break; /* read error */
291 ret = 0; /* read timeout */
293 if (progress || opt.limit_rate)
295 ptimer_measure (timer);
297 last_successful_read_tm = ptimer_read (timer);
303 if (!write_data (out, dlbuf, ret, &skip, &sum_written))
311 limit_bandwidth (ret, timer);
314 progress_update (progress, ret, ptimer_read (timer));
316 if (toread > 0 && !opt.quiet)
317 ws_percenttitle (100.0 *
318 (startpos + sum_read) / (startpos + toread));
326 progress_finish (progress, ptimer_read (timer));
329 *elapsed = ptimer_read (timer);
331 ptimer_destroy (timer);
334 *qtyread += sum_read;
336 *qtywritten += sum_written;
341 /* Read a hunk of data from FD, up until a terminator. The terminator
342 is whatever the TERMINATOR function determines it to be; for
343 example, it can be a line of data, or the head of an HTTP response.
344 The function returns the data read allocated with malloc.
346 In case of error, NULL is returned. In case of EOF and no data
347 read, NULL is returned and errno set to 0. In case of EOF with
348 data having been read, the data is returned, but it will
349 (obviously) not contain the terminator.
351 The idea is to be able to read a line of input, or otherwise a hunk
352 of text, such as the head of an HTTP request, without crossing the
353 boundary, so that the next call to fd_read etc. reads the data
354 after the hunk. To achieve that, this function does the following:
356 1. Peek at available data.
358 2. Determine whether the peeked data, along with the previously
359 read data, includes the terminator.
361 2a. If yes, read the data until the end of the terminator, and
364 2b. If no, read the peeked data and goto 1.
366 The function is careful to assume as little as possible about the
367 implementation of peeking. For example, every peek is followed by
368 a read. If the read returns a different amount of data, the
369 process is retried until all data arrives safely.
371 SIZEHINT is the buffer size sufficient to hold all the data in the
372 typical case (it is used as the initial buffer size). MAXSIZE is
373 the maximum amount of memory this function is allowed to allocate,
374 or 0 if no upper limit is to be enforced.
376 This function should be used as a building block for other
377 functions -- see fd_read_line as a simple example. */
380 fd_read_hunk (int fd, hunk_terminator_t terminator, long sizehint, long maxsize)
382 long bufsize = sizehint;
383 char *hunk = xmalloc (bufsize);
384 int tail = 0; /* tail position in HUNK */
386 assert (maxsize >= bufsize);
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 = 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)
459 /* Double the buffer size, but refuse to allocate more than
461 if (maxsize && bufsize >= maxsize)
468 if (maxsize && bufsize > maxsize)
470 hunk = xrealloc (hunk, bufsize);
476 line_terminator (const char *hunk, int oldlen, int peeklen)
478 const char *p = memchr (hunk + oldlen, '\n', peeklen);
480 /* p+1 because we want the line to include '\n' */
485 /* The maximum size of the single line we agree to accept. This is
486 not meant to impose an arbitrary limit, but to protect the user
487 from Wget slurping up available memory upon encountering malicious
488 or buggy server output. Define it to 0 to remove the limit. */
489 #define FD_READ_LINE_MAX 4096
491 /* Read one line from FD and return it. The line is allocated using
492 malloc, but is never larger than FD_READ_LINE_MAX.
494 If an error occurs, or if no data can be read, NULL is returned.
495 In the former case errno indicates the error condition, and in the
496 latter case, errno is NULL. */
499 fd_read_line (int fd)
501 return fd_read_hunk (fd, line_terminator, 128, FD_READ_LINE_MAX);
504 /* Return a printed representation of the download rate, as
505 appropriate for the speed. If PAD is non-zero, strings will be
506 padded to the width of 7 characters (xxxx.xx). */
508 retr_rate (wgint bytes, double msecs, int pad)
511 static const char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
514 double dlrate = calc_rate (bytes, msecs, &units);
515 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
520 /* Calculate the download rate and trim it as appropriate for the
521 speed. Appropriate means that if rate is greater than 1K/s,
522 kilobytes are used, and if rate is greater than 1MB/s, megabytes
525 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
528 calc_rate (wgint bytes, double msecs, int *units)
536 /* If elapsed time is exactly zero, it means we're under the
537 resolution of the timer. This can easily happen on systems
538 that use time() for the timer. Since the interval lies between
539 0 and the timer's resolution, assume half the resolution. */
540 msecs = ptimer_resolution () / 2.0;
542 dlrate = 1000.0 * bytes / msecs;
545 else if (dlrate < 1024.0 * 1024.0)
546 *units = 1, dlrate /= 1024.0;
547 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
548 *units = 2, dlrate /= (1024.0 * 1024.0);
550 /* Maybe someone will need this, one day. */
551 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
556 /* Maximum number of allowed redirections. 20 was chosen as a
557 "reasonable" value, which is low enough to not cause havoc, yet
558 high enough to guarantee that normal retrievals will not be hurt by
561 #define MAX_REDIRECTIONS 20
563 #define SUSPEND_POST_DATA do { \
564 post_data_suspended = 1; \
565 saved_post_data = opt.post_data; \
566 saved_post_file_name = opt.post_file_name; \
567 opt.post_data = NULL; \
568 opt.post_file_name = NULL; \
571 #define RESTORE_POST_DATA do { \
572 if (post_data_suspended) \
574 opt.post_data = saved_post_data; \
575 opt.post_file_name = saved_post_file_name; \
576 post_data_suspended = 0; \
580 static char *getproxy PARAMS ((struct url *));
582 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
585 /* #### This function should be rewritten so it doesn't return from
589 retrieve_url (const char *origurl, char **file, char **newloc,
590 const char *refurl, int *dt)
594 int location_changed, dummy;
595 char *mynewloc, *proxy;
596 struct url *u, *proxy_url;
597 int up_error_code; /* url parse error code */
599 int redirection_count = 0;
601 int post_data_suspended = 0;
602 char *saved_post_data = NULL;
603 char *saved_post_file_name = NULL;
605 /* If dt is NULL, use local storage. */
611 url = xstrdup (origurl);
617 u = url_parse (url, &up_error_code);
620 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
626 refurl = opt.referer;
635 proxy = getproxy (u);
638 /* Parse the proxy URL. */
639 proxy_url = url_parse (proxy, &up_error_code);
642 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
643 proxy, url_error (up_error_code));
648 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
650 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
651 url_free (proxy_url);
658 if (u->scheme == SCHEME_HTTP
660 || u->scheme == SCHEME_HTTPS
662 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
664 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
666 else if (u->scheme == SCHEME_FTP)
668 /* If this is a redirection, we must not allow recursive FTP
669 retrieval, so we save recursion to oldrec, and restore it
671 int oldrec = opt.recursive;
672 if (redirection_count)
674 result = ftp_loop (u, dt, proxy_url);
675 opt.recursive = oldrec;
677 /* There is a possibility of having HTTP being redirected to
678 FTP. In these cases we must decide whether the text is HTML
679 according to the suffix. The HTML suffixes are `.html',
680 `.htm' and a few others, case-insensitive. */
681 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
683 if (has_html_suffix_p (local_file))
690 url_free (proxy_url);
694 location_changed = (result == NEWLOCATION);
695 if (location_changed)
697 char *construced_newloc;
698 struct url *newloc_parsed;
700 assert (mynewloc != NULL);
705 /* The HTTP specs only allow absolute URLs to appear in
706 redirects, but a ton of boneheaded webservers and CGIs out
707 there break the rules and use relative URLs, and popular
708 browsers are lenient about this, so wget should be too. */
709 construced_newloc = uri_merge (url, mynewloc);
711 mynewloc = construced_newloc;
713 /* Now, see if this new location makes sense. */
714 newloc_parsed = url_parse (mynewloc, &up_error_code);
717 logprintf (LOG_NOTQUIET, "%s: %s.\n", escnonprint_uri (mynewloc),
718 url_error (up_error_code));
726 /* Now mynewloc will become newloc_parsed->url, because if the
727 Location contained relative paths like .././something, we
728 don't want that propagating as url. */
730 mynewloc = xstrdup (newloc_parsed->url);
732 /* Check for max. number of redirections. */
733 if (++redirection_count > MAX_REDIRECTIONS)
735 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
737 url_free (newloc_parsed);
750 /* If we're being redirected from POST, we don't want to POST
751 again. Many requests answer POST with a redirection to an
752 index page; that redirection is clearly a GET. We "suspend"
753 POST data for the duration of the redirections, and restore
754 it when we're done. */
755 if (!post_data_suspended)
765 register_download (u->url, local_file);
766 if (redirection_count && 0 != strcmp (origurl, u->url))
767 register_redirection (origurl, u->url);
769 register_html (u->url, local_file);
774 *file = local_file ? local_file : NULL;
776 xfree_null (local_file);
780 if (redirection_count)
799 /* Find the URLs in the file and call retrieve_url() for each of
800 them. If HTML is non-zero, treat the file as HTML, and construct
801 the URLs accordingly.
803 If opt.recursive is set, call retrieve_tree() for each file. */
806 retrieve_from_file (const char *file, int html, int *count)
809 struct urlpos *url_list, *cur_url;
811 url_list = (html ? get_urls_html (file, NULL, NULL)
812 : get_urls_file (file));
813 status = RETROK; /* Suppose everything is OK. */
814 *count = 0; /* Reset the URL count. */
816 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
818 char *filename = NULL, *new_file = NULL;
821 if (cur_url->ignore_when_downloading)
824 if (opt.quota && total_downloaded_bytes > opt.quota)
829 if ((opt.recursive || opt.page_requisites)
830 && cur_url->url->scheme != SCHEME_FTP)
831 status = retrieve_tree (cur_url->url->url);
833 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
835 if (filename && opt.delete_after && file_exists_p (filename))
837 DEBUGP (("Removing file due to --delete-after in"
838 " retrieve_from_file():\n"));
839 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
840 if (unlink (filename))
841 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
845 xfree_null (new_file);
846 xfree_null (filename);
849 /* Free the linked list of URL-s. */
850 free_urlpos (url_list);
855 /* Print `giving up', or `retrying', depending on the impending
856 action. N1 and N2 are the attempt number and the attempt limit. */
858 printwhat (int n1, int n2)
860 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
863 /* If opt.wait or opt.waitretry are specified, and if certain
864 conditions are met, sleep the appropriate number of seconds. See
865 the documentation of --wait and --waitretry for more information.
867 COUNT is the count of current retrieval, beginning with 1. */
870 sleep_between_retrievals (int count)
872 static int first_retrieval = 1;
876 /* Don't sleep before the very first retrieval. */
881 if (opt.waitretry && count > 1)
883 /* If opt.waitretry is specified and this is a retry, wait for
884 COUNT-1 number of seconds, or for opt.waitretry seconds. */
885 if (count <= opt.waitretry)
888 xsleep (opt.waitretry);
892 if (!opt.random_wait || count > 1)
893 /* If random-wait is not specified, or if we are sleeping
894 between retries of the same download, sleep the fixed
899 /* Sleep a random amount of time averaging in opt.wait
900 seconds. The sleeping amount ranges from 0 to
901 opt.wait*2, inclusive. */
902 double waitsecs = 2 * opt.wait * random_float ();
903 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
904 opt.wait, waitsecs));
910 /* Free the linked list of urlpos. */
912 free_urlpos (struct urlpos *l)
916 struct urlpos *next = l->next;
919 xfree_null (l->local_name);
925 /* Rotate FNAME opt.backups times */
927 rotate_backups(const char *fname)
929 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
930 char *from = (char *)alloca (maxlen);
931 char *to = (char *)alloca (maxlen);
935 if (stat (fname, &sb) == 0)
936 if (S_ISREG (sb.st_mode) == 0)
939 for (i = opt.backups; i > 1; i--)
941 sprintf (from, "%s.%d", fname, i - 1);
942 sprintf (to, "%s.%d", fname, i);
946 sprintf (to, "%s.%d", fname, 1);
950 static int no_proxy_match PARAMS ((const char *, const char **));
952 /* Return the URL of the proxy appropriate for url U. */
955 getproxy (struct url *u)
959 static char rewritten_storage[1024];
963 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
969 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
973 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
977 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
982 if (!proxy || !*proxy)
985 /* Handle shorthands. `rewritten_storage' is a kludge to allow
986 getproxy() to return static storage. */
987 rewritten_url = rewrite_shorthand_url (proxy);
990 strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
991 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
992 proxy = rewritten_storage;
998 /* Should a host be accessed through proxy, concerning no_proxy? */
1000 no_proxy_match (const char *host, const char **no_proxy)
1005 return !sufmatch (no_proxy, host);