cookies.c

wget讓你可以在console介面下

      /* According to netscape's specification, expiry time in the
	 past means that discarding of a matching cookie is
	 requested.  */
      if (cookie->expiry_time < cookies_now)
	cookie->discard_requested = 1;

      return 1;
    }
  else if (NAME_IS ("max-age"))
    {
      double maxage = -1;
      char *value_copy;

      if (!VALUE_NON_EMPTY)
	return 0;
      BOUNDED_TO_ALLOCA (value_b, value_e, value_copy);

      sscanf (value_copy, "%lf", &maxage);
      if (maxage == -1)
	/* something went wrong. */
	return 0;
      cookie->permanent = 1;
      cookie->expiry_time = cookies_now + maxage;

      /* According to rfc2109, a cookie with max-age of 0 means that
	 discarding of a matching cookie is requested.  */
      if (maxage == 0)
	cookie->discard_requested = 1;

      return 1;
    }
  else if (NAME_IS ("secure"))
    {
      /* ignore value completely */
      cookie->secure = 1;
      return 1;
    }
  else
    /* Unrecognized attribute; ignore it. */
    return 1;
}

#undef NAME_IS

/* Returns non-zero for characters that are legal in the name of an
   attribute.  This used to allow only alphanumerics, '-', and '_',
   but we need to be more lenient because a number of sites wants to
   use weirder attribute names.  rfc2965 "informally specifies"
   attribute name (token) as "a sequence of non-special, non-white
   space characters".  So we allow everything except the stuff we know
   could harm us.  */

#define ATTR_NAME_CHAR(c) ((c) > 32 && (c) < 127	\
			   && (c) != '"' && (c) != '='	\
			   && (c) != ';' && (c) != ',')

/* Parse the contents of the `Set-Cookie' header.  The header looks
   like this:

   name1=value1; name2=value2; ...

   Trailing semicolon is optional; spaces are allowed between all
   tokens.  Additionally, values may be quoted.

   A new cookie is returned upon success, NULL otherwise.  The
   specified CALLBACK function (normally `update_cookie_field' is used
   to update the fields of the newly created cookie structure.  */

static struct cookie *
parse_set_cookies (const char *sc,
		   int (*callback) (struct cookie *,
				    const char *, const char *,
				    const char *, const char *),
		   int silent)
{
  struct cookie *cookie = cookie_new ();

  /* #### Hand-written DFAs are no fun to debug.  We'de be better off
     to rewrite this as an inline parser.  */

  enum { S_START, S_NAME, S_NAME_POST,
	 S_VALUE_PRE, S_VALUE, S_QUOTED_VALUE, S_VALUE_TRAILSPACE,
	 S_ATTR_ACTION, S_DONE, S_ERROR
  } state = S_START;

  const char *p = sc;
  char c;

  const char *name_b  = NULL, *name_e  = NULL;
  const char *value_b = NULL, *value_e = NULL;

  c = *p;

  while (state != S_DONE && state != S_ERROR)
    {
      switch (state)
	{
	case S_START:
	  if (!c)
	    state = S_DONE;
	  else if (ISSPACE (c))
	    /* Strip all whitespace preceding the name. */
	    c = *++p;
	  else if (ATTR_NAME_CHAR (c))
	    {
	      name_b = p;
	      state = S_NAME;
	    }
	  else
	    /* empty attr name not allowed */
	    state = S_ERROR;
	  break;
	case S_NAME:
	  if (!c || c == ';' || c == '=' || ISSPACE (c))
	    {
	      name_e = p;
	      state = S_NAME_POST;
	    }
	  else if (ATTR_NAME_CHAR (c))
	    c = *++p;
	  else
	    state = S_ERROR;
	  break;
	case S_NAME_POST:
	  if (!c || c == ';')
	    {
	      value_b = value_e = NULL;
	      if (c == ';')
		c = *++p;
	      state = S_ATTR_ACTION;
	    }
	  else if (c == '=')
	    {
	      c = *++p;
	      state = S_VALUE_PRE;
	    }
	  else if (ISSPACE (c))
	    /* Ignore space and keep the state. */
	    c = *++p;
	  else
	    state = S_ERROR;
	  break;
	case S_VALUE_PRE:
	  if (!c || c == ';')
	    {
	      value_b = value_e = p;
	      if (c == ';')
		c = *++p;
	      state = S_ATTR_ACTION;
	    }
	  else if (c == '"')
	    {
	      c = *++p;
	      value_b = p;
	      state = S_QUOTED_VALUE;
	    }
	  else if (ISSPACE (c))
	    c = *++p;
	  else
	    {
	      value_b = p;
	      value_e = NULL;
	      state = S_VALUE;
	    }
	  break;
	case S_VALUE:
	  if (!c || c == ';' || ISSPACE (c))
	    {
	      value_e = p;
	      state = S_VALUE_TRAILSPACE;
	    }
	  else
	    {
	      value_e = NULL;	/* no trailing space */
	      c = *++p;
	    }
	  break;
	case S_QUOTED_VALUE:
	  if (c == '"')
	    {
	      value_e = p;
	      c = *++p;
	      state = S_VALUE_TRAILSPACE;
	    }
	  else if (!c)
	    state = S_ERROR;
	  else
	    c = *++p;
	  break;
	case S_VALUE_TRAILSPACE:
	  if (c == ';')
	    {
	      c = *++p;
	      state = S_ATTR_ACTION;
	    }
	  else if (!c)
	    state = S_ATTR_ACTION;
	  else if (ISSPACE (c))
	    c = *++p;
	  else
	    state = S_VALUE;
	  break;
	case S_ATTR_ACTION:
	  {
	    int legal = callback (cookie, name_b, name_e, value_b, value_e);
	    if (!legal)
	      {
		if (!silent)
		  {
		    char *name;
		    BOUNDED_TO_ALLOCA (name_b, name_e, name);
		    logprintf (LOG_NOTQUIET,
			       _("Error in Set-Cookie, field `%s'"),
			       escnonprint (name));
		  }
		state = S_ERROR;
		break;
	      }
	    state = S_START;
	  }
	  break;
	case S_DONE:
	case S_ERROR:
	  /* handled by loop condition */
	  break;
	}
    }
  if (state == S_DONE)
    return cookie;

  delete_cookie (cookie);
  if (state != S_ERROR)
    abort ();

  if (!silent)
    logprintf (LOG_NOTQUIET,
	       _("Syntax error in Set-Cookie: %s at position %d.\n"),
	       escnonprint (sc), (int) (p - sc));
  return NULL;
}

/* Sanity checks.  These are important, otherwise it is possible for
   mailcious attackers to destroy important cookie information and/or
   violate your privacy.  */


#define REQUIRE_DIGITS(p) do {			\
  if (!ISDIGIT (*p))				\
    return 0;					\
  for (++p; ISDIGIT (*p); p++)			\
    ;						\
} while (0)

#define REQUIRE_DOT(p) do {			\
  if (*p++ != '.')				\
    return 0;					\
} while (0)

/* Check whether ADDR matches <digits>.<digits>.<digits>.<digits>.

   We don't want to call network functions like inet_addr() because
   all we need is a check, preferrably one that is small, fast, and
   well-defined.  */

static int
numeric_address_p (const char *addr)
{
  const char *p = addr;

  REQUIRE_DIGITS (p);		/* A */
  REQUIRE_DOT (p);		/* . */
  REQUIRE_DIGITS (p);		/* B */
  REQUIRE_DOT (p);		/* . */
  REQUIRE_DIGITS (p);		/* C */
  REQUIRE_DOT (p);		/* . */
  REQUIRE_DIGITS (p);		/* D */

  if (*p != '\0')
    return 0;
  return 1;
}

/* Check whether COOKIE_DOMAIN is an appropriate domain for HOST.
   Originally I tried to make the check compliant with rfc2109, but
   the sites deviated too often, so I had to fall back to "tail
   matching", as defined by the original Netscape's cookie spec.  */

static int
check_domain_match (const char *cookie_domain, const char *host)
{
  DEBUGP (("cdm: 1"));

  /* Numeric address requires exact match.  It also requires HOST to
     be an IP address.  */
  if (numeric_address_p (cookie_domain))
    return 0 == strcmp (cookie_domain, host);

  DEBUGP ((" 2"));

  /* For the sake of efficiency, check for exact match first. */
  if (0 == strcasecmp (cookie_domain, host))
    return 1;

  DEBUGP ((" 3"));

  /* HOST must match the tail of cookie_domain. */
  if (!match_tail (host, cookie_domain, 1))
    return 0;

  /* We know that COOKIE_DOMAIN is a subset of HOST; however, we must
     make sure that somebody is not trying to set the cookie for a
     subdomain shared by many entities.  For example, "company.co.uk"
     must not be allowed to set a cookie for ".co.uk".  On the other
     hand, "sso.redhat.de" should be able to set a cookie for
     ".redhat.de".

     The only marginally sane way to handle this I can think of is to
     reject on the basis of the length of the second-level domain name
     (but when the top-level domain is unknown), with the assumption
     that those of three or less characters could be reserved.  For
     example:

          .co.org -> works because the TLD is known
           .co.uk -> doesn't work because "co" is only two chars long
          .com.au -> doesn't work because "com" is only 3 chars long
          .cnn.uk -> doesn't work because "cnn" is also only 3 chars long (ugh)
          .cnn.de -> doesn't work for the same reason (ugh!!)
         .abcd.de -> works because "abcd" is 4 chars long
      .img.cnn.de -> works because it's not trying to set the 2nd level domain
       .cnn.co.uk -> works for the same reason

    That should prevent misuse, while allowing reasonable usage.  If
    someone knows of a better way to handle this, please let me
    know.  */
  {
    const char *p = cookie_domain;
    int dccount = 1;		/* number of domain components */
    int ldcl  = 0;		/* last domain component length */
    int nldcl = 0;		/* next to last domain component length */
    int out;
    if (*p == '.')
      /* Ignore leading period in this calculation. */
      ++p;
    DEBUGP ((" 4"));
    for (out = 0; !out; p++)
      switch (*p)
	{
	case '\0':
	  out = 1;
	  break;
	case '.':
	  if (ldcl == 0)
	    /* Empty domain component found -- the domain is invalid. */
	    return 0;
	  if (*(p + 1) == '\0')
	    {
	      /* Tolerate trailing '.' by not treating the domain as
		 one ending with an empty domain component.  */
	      out = 1;
	      break;
	    }
	  nldcl = ldcl;
	  ldcl  = 0;
	  ++dccount;
	  break;
	default:
	  ++ldcl;
	}

    DEBUGP ((" 5"));

    if (dccount < 2)
      return 0;

    DEBUGP ((" 6"));

    if (dccount == 2)
      {
	int i;
	int known_toplevel = 0;
	static const char *known_toplevel_domains[] = {
	  ".com", ".edu", ".net", ".org", ".gov", ".mil", ".int"
	};
	for (i = 0; i < countof (known_toplevel_domains); i++)
	  if (match_tail (cookie_domain, known_toplevel_domains[i], 1))
	    {
	      known_toplevel = 1;
	      break;
	    }
	if (!known_toplevel && nldcl <= 3)
	  return 0;
      }
  }

  DEBUGP ((" 7"));

  /* Don't allow the host "foobar.com" to set a cookie for domain
     "bar.com".  */
  if (*cookie_domain != '.')
    {
      int dlen = strlen (cookie_domain);
      int hlen = strlen (host);
      /* cookie host:    hostname.foobar.com */
      /* desired domain:             bar.com */
      /* '.' must be here in host-> ^        */
      if (hlen > dlen && host[hlen - dlen - 1] != '.')