[wget] / src / retr.c

/* File retrieval.
   Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001 Free Software Foundation, Inc.

This file is part of GNU Wget.

GNU Wget is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or (at
your option) any later version.

GNU Wget is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with Wget; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

In addition, as a special exception, the Free Software Foundation
gives permission to link the code of its release of Wget with the
OpenSSL project's "OpenSSL" library (or with modified versions of it
that use the same license as the "OpenSSL" library), and distribute
the linked executables.  You must obey the GNU General Public License
in all respects for all of the code used other than "OpenSSL".  If you
modify this file, you may extend this exception to your version of the
file, but you are not obligated to do so.  If you do not wish to do
so, delete this exception statement from your version.  */

#include <config.h>

#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif /* HAVE_UNISTD_H */
#include <errno.h>
#ifdef HAVE_STRING_H
# include <string.h>
#else
# include <strings.h>
#endif /* HAVE_STRING_H */
#include <assert.h>

#include "wget.h"
#include "utils.h"
#include "retr.h"
#include "progress.h"
#include "url.h"
#include "recur.h"
#include "ftp.h"
#include "host.h"
#include "connect.h"
#include "hash.h"

#ifdef HAVE_SSL
# include "gen_sslfunc.h"       /* for ssl_iread */
#endif

#ifndef errno
extern int errno;
#endif

/* See the comment in gethttp() why this is needed. */
int global_download_count;

\f
static struct {
  long bytes;
  double dltime;
} limit_data;

static void
limit_bandwidth_reset (void)
{
  limit_data.bytes  = 0;
  limit_data.dltime = 0;
}

/* Limit the bandwidth by pausing the download for an amount of time.
   BYTES is the number of bytes received from the network, and DELTA
   is how long it took to receive them.  */

static void
limit_bandwidth (long bytes, double delta)
{
  double expected;

  limit_data.bytes += bytes;
  limit_data.dltime += delta;

  expected = 1000.0 * limit_data.bytes / opt.limit_rate;

  if (expected > limit_data.dltime)
    {
      double slp = expected - limit_data.dltime;
      if (slp < 200)
        {
          DEBUGP (("deferring a %.2f ms sleep (%ld/%.2f).\n",
                   slp, limit_data.bytes, limit_data.dltime));
          return;
        }
      DEBUGP (("sleeping %.2f ms\n", slp));
      usleep ((unsigned long) (1000 * slp));
    }

  limit_data.bytes = 0;
  limit_data.dltime = 0;
}

#define MIN(i, j) ((i) <= (j) ? (i) : (j))

/* Reads the contents of file descriptor FD, until it is closed, or a
   read error occurs.  The data is read in 8K chunks, and stored to
   stream fp, which should have been open for writing.  If BUF is
   non-NULL and its file descriptor is equal to FD, flush RBUF first.
   This function will *not* use the rbuf_* functions!

   The EXPECTED argument is passed to show_progress() unchanged, but
   otherwise ignored.

   If opt.verbose is set, the progress is also shown.  RESTVAL
   represents a value from which to start downloading (which will be
   shown accordingly).  If RESTVAL is non-zero, the stream should have
   been open for appending.

   The function exits and returns codes of 0, -1 and -2 if the
   connection was closed, there was a read error, or if it could not
   write to the output stream, respectively.

   IMPORTANT: The function flushes the contents of the buffer in
   rbuf_flush() before actually reading from fd.  If you wish to read
   from fd immediately, flush or discard the buffer.  */
int
get_contents (int fd, FILE *fp, long *len, long restval, long expected,
              struct rbuf *rbuf, int use_expected, double *elapsed)
{
  int res = 0;

  static char dlbuf[16384];
  int dlbufsize = sizeof (dlbuf);

  void *progress = NULL;
  struct wget_timer *timer = wtimer_allocate ();
  double dltime = 0, last_dltime = 0;

  *len = restval;

  if (opt.verbose)
    progress = progress_create (restval, expected);

  if (rbuf && RBUF_FD (rbuf) == fd)
    {
      int sz = 0;
      while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
        {
          fwrite (dlbuf, 1, res, fp);
          *len += res;
          sz += res;
        }
      if (sz)
        fflush (fp);
      if (ferror (fp))
        {
          res = -2;
          goto out;
        }
      if (progress)
        progress_update (progress, sz, 0);
    }

  if (opt.limit_rate)
    limit_bandwidth_reset ();
  wtimer_reset (timer);

  /* If we're limiting the download, set our buffer size to the
     limit.  */
  if (opt.limit_rate && opt.limit_rate < dlbufsize)
    dlbufsize = opt.limit_rate;

  /* Read from fd while there is available data.

     Normally, if expected is 0, it means that it is not known how
     much data is expected.  However, if use_expected is specified,
     then expected being zero means exactly that.  */
  while (!use_expected || (*len < expected))
    {
      int amount_to_read = (use_expected
                            ? MIN (expected - *len, dlbufsize) : dlbufsize);
#ifdef HAVE_SSL
      if (rbuf->ssl!=NULL)
        res = ssl_iread (rbuf->ssl, dlbuf, amount_to_read);
      else
#endif /* HAVE_SSL */
        res = iread (fd, dlbuf, amount_to_read);

      if (res <= 0)
        break;

      fwrite (dlbuf, 1, res, fp);
      /* Always flush the contents of the network packet.  This should
         not hinder performance: fast downloads will be received in
         16K chunks (which stdio would write out anyway), and slow
         downloads won't be limited with disk performance.  */
      fflush (fp);
      if (ferror (fp))
        {
          res = -2;
          goto out;
        }

      /* If bandwidth is not limited, one call to wtimer_elapsed is
         sufficient.  */
      dltime = wtimer_elapsed (timer);
      if (opt.limit_rate)
        {
          limit_bandwidth (res, dltime - last_dltime);
          dltime = wtimer_elapsed (timer);
          last_dltime = dltime;
        }

      if (progress)
        progress_update (progress, res, dltime);
      *len += res;
    }
  if (res < -1)
    res = -1;

 out:
  if (progress)
    progress_finish (progress, dltime);
  if (elapsed)
    *elapsed = dltime;
  wtimer_delete (timer);

  return res;
}
\f
/* Return a printed representation of the download rate, as
   appropriate for the speed.  If PAD is non-zero, strings will be
   padded to the width of 7 characters (xxxx.xx).  */
char *
retr_rate (long bytes, double msecs, int pad)
{
  static char res[20];
  static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
  int units = 0;

  double dlrate = calc_rate (bytes, msecs, &units);
  sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);

  return res;
}

/* Calculate the download rate and trim it as appropriate for the
   speed.  Appropriate means that if rate is greater than 1K/s,
   kilobytes are used, and if rate is greater than 1MB/s, megabytes
   are used.

   UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
   GB/s.  */
double
calc_rate (long bytes, double msecs, int *units)
{
  double dlrate;

  assert (msecs >= 0);
  assert (bytes >= 0);

  if (msecs == 0)
    /* If elapsed time is exactly zero, it means we're under the
       granularity of the timer.  This often happens on systems that
       use time() for the timer.  */
    msecs = wtimer_granularity ();

  dlrate = (double)1000 * bytes / msecs;
  if (dlrate < 1024.0)
    *units = 0;
  else if (dlrate < 1024.0 * 1024.0)
    *units = 1, dlrate /= 1024.0;
  else if (dlrate < 1024.0 * 1024.0 * 1024.0)
    *units = 2, dlrate /= (1024.0 * 1024.0);
  else
    /* Maybe someone will need this, one day. */
    *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);

  return dlrate;
}
\f
/* Maximum number of allowed redirections.  20 was chosen as a
   "reasonable" value, which is low enough to not cause havoc, yet
   high enough to guarantee that normal retrievals will not be hurt by
   the check.  */

#define MAX_REDIRECTIONS 20

#define SUSPEND_POST_DATA do {                  \
  post_data_suspended = 1;                      \
  saved_post_data = opt.post_data;              \
  saved_post_file_name = opt.post_file_name;    \
  opt.post_data = NULL;                         \
  opt.post_file_name = NULL;                    \
} while (0)

#define RESTORE_POST_DATA do {                          \
  if (post_data_suspended)                              \
    {                                                   \
      opt.post_data = saved_post_data;                  \
      opt.post_file_name = saved_post_file_name;        \
      post_data_suspended = 0;                          \
    }                                                   \
} while (0)

/* Retrieve the given URL.  Decides which loop to call -- HTTP, FTP,
   FTP, proxy, etc.  */

/* #### This function should be rewritten so it doesn't return from
   multiple points. */

uerr_t
retrieve_url (const char *origurl, char **file, char **newloc,
              const char *refurl, int *dt)
{
  uerr_t result;
  char *url;
  int location_changed, dummy;
  char *mynewloc, *proxy;
  struct url *u, *proxy_url;
  int up_error_code;            /* url parse error code */
  char *local_file;
  int redirection_count = 0;

  int post_data_suspended = 0;
  char *saved_post_data = NULL;
  char *saved_post_file_name = NULL;

  /* If dt is NULL, just ignore it.  */
  if (!dt)
    dt = &dummy;
  url = xstrdup (origurl);
  if (newloc)
    *newloc = NULL;
  if (file)
    *file = NULL;

  u = url_parse (url, &up_error_code);
  if (!u)
    {
      logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
      xfree (url);
      return URLERROR;
    }

  if (!refurl)
    refurl = opt.referer;

 redirected:

  result = NOCONERROR;
  mynewloc = NULL;
  local_file = NULL;
  proxy_url = NULL;

  proxy = getproxy (u);
  if (proxy)
    {
      /* Parse the proxy URL.  */
      proxy_url = url_parse (proxy, &up_error_code);
      if (!proxy_url)
        {
          logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
                     proxy, url_error (up_error_code));
          xfree (url);
          RESTORE_POST_DATA;
          return PROXERR;
        }
      if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
        {
          logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
          url_free (proxy_url);
          xfree (url);
          RESTORE_POST_DATA;
          return PROXERR;
        }
    }

  if (u->scheme == SCHEME_HTTP
#ifdef HAVE_SSL
      || u->scheme == SCHEME_HTTPS
#endif
      || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
    {
      result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
    }
  else if (u->scheme == SCHEME_FTP)
    {
      /* If this is a redirection, we must not allow recursive FTP
         retrieval, so we save recursion to oldrec, and restore it
         later.  */
      int oldrec = opt.recursive;
      if (redirection_count)
        opt.recursive = 0;
      result = ftp_loop (u, dt, proxy_url);
      opt.recursive = oldrec;

      /* There is a possibility of having HTTP being redirected to
         FTP.  In these cases we must decide whether the text is HTML
         according to the suffix.  The HTML suffixes are `.html',
         `.htm' and a few others, case-insensitive.  */
      if (redirection_count && local_file && u->scheme == SCHEME_FTP)
        {
          if (has_html_suffix_p (local_file))
            *dt |= TEXTHTML;
        }
    }

  if (proxy_url)
    {
      url_free (proxy_url);
      proxy_url = NULL;
    }

  location_changed = (result == NEWLOCATION);
  if (location_changed)
    {
      char *construced_newloc;
      struct url *newloc_parsed;

      assert (mynewloc != NULL);

      if (local_file)
        xfree (local_file);

      /* The HTTP specs only allow absolute URLs to appear in
         redirects, but a ton of boneheaded webservers and CGIs out
         there break the rules and use relative URLs, and popular
         browsers are lenient about this, so wget should be too. */
      construced_newloc = uri_merge (url, mynewloc);
      xfree (mynewloc);
      mynewloc = construced_newloc;

      /* Now, see if this new location makes sense. */
      newloc_parsed = url_parse (mynewloc, &up_error_code);
      if (!newloc_parsed)
        {
          logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
                     url_error (up_error_code));
          url_free (u);
          xfree (url);
          xfree (mynewloc);
          RESTORE_POST_DATA;
          return result;
        }

      /* Now mynewloc will become newloc_parsed->url, because if the
         Location contained relative paths like .././something, we
         don't want that propagating as url.  */
      xfree (mynewloc);
      mynewloc = xstrdup (newloc_parsed->url);

      /* Check for max. number of redirections.  */
      if (++redirection_count > MAX_REDIRECTIONS)
        {
          logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
                     MAX_REDIRECTIONS);
          url_free (newloc_parsed);
          url_free (u);
          xfree (url);
          xfree (mynewloc);
          RESTORE_POST_DATA;
          return WRONGCODE;
        }

      xfree (url);
      url = mynewloc;
      url_free (u);
      u = newloc_parsed;

      /* If we're being redirected from POST, we don't want to POST
         again.  Many requests answer POST with a redirection to an
         index page; that redirection is clearly a GET.  We "suspend"
         POST data for the duration of the redirections, and restore
         it when we're done. */
      if (!post_data_suspended)
        SUSPEND_POST_DATA;

      goto redirected;
    }

  if (local_file)
    {
      if (*dt & RETROKF)
        {
          register_download (u->url, local_file);
          if (redirection_count && 0 != strcmp (origurl, u->url))
            register_redirection (origurl, u->url);
          if (*dt & TEXTHTML)
            register_html (u->url, local_file);
        }
    }

  if (file)
    *file = local_file ? local_file : NULL;
  else
    FREE_MAYBE (local_file);

  url_free (u);

  if (redirection_count)
    {
      if (newloc)
        *newloc = url;
      else
        xfree (url);
    }
  else
    {
      if (newloc)
        *newloc = NULL;
      xfree (url);
    }

  ++global_download_count;
  RESTORE_POST_DATA;

  return result;
}

/* Find the URLs in the file and call retrieve_url() for each of
   them.  If HTML is non-zero, treat the file as HTML, and construct
   the URLs accordingly.

   If opt.recursive is set, call recursive_retrieve() for each file.  */
uerr_t
retrieve_from_file (const char *file, int html, int *count)
{
  uerr_t status;
  struct urlpos *url_list, *cur_url;

  url_list = (html ? get_urls_html (file, NULL, NULL)
              : get_urls_file (file));
  status = RETROK;             /* Suppose everything is OK.  */
  *count = 0;                  /* Reset the URL count.  */

  for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
    {
      char *filename = NULL, *new_file = NULL;
      int dt;

      if (cur_url->ignore_when_downloading)
        continue;

      if (downloaded_exceeds_quota ())
        {
          status = QUOTEXC;
          break;
        }
      if (opt.recursive && cur_url->url->scheme != SCHEME_FTP)
        status = retrieve_tree (cur_url->url->url);
      else
        status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);

      if (filename && opt.delete_after && file_exists_p (filename))
        {
          DEBUGP (("Removing file due to --delete-after in"
                   " retrieve_from_file():\n"));
          logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
          if (unlink (filename))
            logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
          dt &= ~RETROKF;
        }

      FREE_MAYBE (new_file);
      FREE_MAYBE (filename);
    }

  /* Free the linked list of URL-s.  */
  free_urlpos (url_list);

  return status;
}

/* Print `giving up', or `retrying', depending on the impending
   action.  N1 and N2 are the attempt number and the attempt limit.  */
void
printwhat (int n1, int n2)
{
  logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
}

/* Increment opt.downloaded by BY_HOW_MUCH.  If an overflow occurs,
   set opt.downloaded_overflow to 1. */
void
downloaded_increase (unsigned long by_how_much)
{
  VERY_LONG_TYPE old;
  if (opt.downloaded_overflow)
    return;
  old = opt.downloaded;
  opt.downloaded += by_how_much;
  if (opt.downloaded < old)     /* carry flag, where are you when I
                                   need you? */
    {
      /* Overflow. */
      opt.downloaded_overflow = 1;
      opt.downloaded = ~((VERY_LONG_TYPE)0);
    }
}

/* Return non-zero if the downloaded amount of bytes exceeds the
   desired quota.  If quota is not set or if the amount overflowed, 0
   is returned. */
int
downloaded_exceeds_quota (void)
{
  if (!opt.quota)
    return 0;
  if (opt.downloaded_overflow)
    /* We don't really know.  (Wildly) assume not. */
    return 0;

  return opt.downloaded > opt.quota;
}

/* If opt.wait or opt.waitretry are specified, and if certain
   conditions are met, sleep the appropriate number of seconds.  See
   the documentation of --wait and --waitretry for more information.

   COUNT is the count of current retrieval, beginning with 1. */

void
sleep_between_retrievals (int count)
{
  static int first_retrieval = 1;

  if (first_retrieval)
    {
      /* Don't sleep before the very first retrieval. */
      first_retrieval = 0;
      return;
    }

  if (opt.waitretry && count > 1)
    {
      /* If opt.waitretry is specified and this is a retry, wait for
         COUNT-1 number of seconds, or for opt.waitretry seconds.  */
      if (count <= opt.waitretry)
        sleep (count - 1);
      else
        sleep (opt.waitretry);
    }
  else if (opt.wait)
    {
      if (!opt.random_wait || count > 1)
        /* If random-wait is not specified, or if we are sleeping
           between retries of the same download, sleep the fixed
           interval.  */
        sleep (opt.wait);
      else
        {
          /* Sleep a random amount of time averaging in opt.wait
             seconds.  The sleeping amount ranges from 0 to
             opt.wait*2, inclusive.  */
          int waitsecs = random_number (opt.wait * 2 + 1);

          DEBUGP (("sleep_between_retrievals: norm=%ld,fuzz=%ld,sleep=%d\n",
                   opt.wait, waitsecs - opt.wait, waitsecs));

          if (waitsecs)
            sleep (waitsecs);
        }
    }
}