mod_negotiation.c

Apache官方在今天放出产品系列2.2的最新版本2.2.11的源码包 最流行的HTTP服务器软件之一

                }
            }
        }

        return (float)i / 1000.0f;
    }

    return 0.0f;
}

/*
 * Get a single mime type entry --- one media type and parameters;
 * enter the values we recognize into the argument accept_rec
 */

static const char *get_entry(apr_pool_t *p, accept_rec *result,
                             const char *accept_line)
{
    result->quality = 1.0f;
    result->level = 0.0f;
    result->charset = "";

    /*
     * Note that this handles what I gather is the "old format",
     *
     *    Accept: text/html text/plain moo/zot
     *
     * without any compatibility kludges --- if the token after the
     * MIME type begins with a semicolon, we know we're looking at parms,
     * otherwise, we know we aren't.  (So why all the pissing and moaning
     * in the CERN server code?  I must be missing something).
     */

    result->name = ap_get_token(p, &accept_line, 0);
    ap_str_tolower(result->name);     /* You want case insensitive,
                                       * you'll *get* case insensitive.
                                       */

    /* KLUDGE!!! Default HTML to level 2.0 unless the browser
     * *explicitly* says something else.
     */

    if (!strcmp(result->name, "text/html") && (result->level == 0.0)) {
        result->level = 2.0f;
    }
    else if (!strcmp(result->name, INCLUDES_MAGIC_TYPE)) {
        result->level = 2.0f;
    }
    else if (!strcmp(result->name, INCLUDES_MAGIC_TYPE3)) {
        result->level = 3.0f;
    }

    while (*accept_line == ';') {
        /* Parameters ... */

        char *parm;
        char *cp;
        char *end;

        ++accept_line;
        parm = ap_get_token(p, &accept_line, 1);

        /* Look for 'var = value' --- and make sure the var is in lcase. */

        for (cp = parm; (*cp && !apr_isspace(*cp) && *cp != '='); ++cp) {
            *cp = apr_tolower(*cp);
        }

        if (!*cp) {
            continue;           /* No '='; just ignore it. */
        }

        *cp++ = '\0';           /* Delimit var */
        while (*cp && (apr_isspace(*cp) || *cp == '=')) {
            ++cp;
        }

        if (*cp == '"') {
            ++cp;
            for (end = cp;
                 (*end && *end != '\n' && *end != '\r' && *end != '\"');
                 end++);
        }
        else {
            for (end = cp; (*end && !apr_isspace(*end)); end++);
        }
        if (*end) {
            *end = '\0';        /* strip ending quote or return */
        }
        ap_str_tolower(cp);

        if (parm[0] == 'q'
            && (parm[1] == '\0' || (parm[1] == 's' && parm[2] == '\0'))) {
            result->quality = atoq(cp);
        }
        else if (parm[0] == 'l' && !strcmp(&parm[1], "evel")) {
            result->level = (float)atoi(cp);
        }
        else if (!strcmp(parm, "charset")) {
            result->charset = cp;
        }
    }

    if (*accept_line == ',') {
        ++accept_line;
    }

    return accept_line;
}

/*****************************************************************
 *
 * Dealing with header lines ...
 *
 * Accept, Accept-Charset, Accept-Language and Accept-Encoding
 * are handled by do_header_line() - they all have the same
 * basic structure of a list of items of the format
 *    name; q=N; charset=TEXT
 *
 * where charset is only valid in Accept.
 */

static apr_array_header_t *do_header_line(apr_pool_t *p,
                                          const char *accept_line)
{
    apr_array_header_t *accept_recs;

    if (!accept_line) {
        return NULL;
    }

    accept_recs = apr_array_make(p, 40, sizeof(accept_rec));

    while (*accept_line) {
        accept_rec *new = (accept_rec *) apr_array_push(accept_recs);
        accept_line = get_entry(p, new, accept_line);
    }

    return accept_recs;
}

/* Given the text of the Content-Languages: line from the var map file,
 * return an array containing the languages of this variant
 */

static apr_array_header_t *do_languages_line(apr_pool_t *p,
                                             const char **lang_line)
{
    apr_array_header_t *lang_recs = apr_array_make(p, 2, sizeof(char *));

    if (!lang_line) {
        return lang_recs;
    }

    while (**lang_line) {
        char **new = (char **) apr_array_push(lang_recs);
        *new = ap_get_token(p, lang_line, 0);
        ap_str_tolower(*new);
        if (**lang_line == ',' || **lang_line == ';') {
            ++(*lang_line);
        }
    }

    return lang_recs;
}

/*****************************************************************
 *
 * Handling header lines from clients...
 */

static negotiation_state *parse_accept_headers(request_rec *r)
{
    negotiation_state *new =
        (negotiation_state *) apr_pcalloc(r->pool, sizeof(negotiation_state));
    accept_rec *elts;
    apr_table_t *hdrs = r->headers_in;
    int i;

    new->pool = r->pool;
    new->r = r;
    new->conf = (neg_dir_config *)ap_get_module_config(r->per_dir_config,
                                                       &negotiation_module);

    new->dir_name = ap_make_dirstr_parent(r->pool, r->filename);

    new->accepts = do_header_line(r->pool, apr_table_get(hdrs, "Accept"));

    /* calculate new->accept_q value */
    if (new->accepts) {
        elts = (accept_rec *) new->accepts->elts;

        for (i = 0; i < new->accepts->nelts; ++i) {
            if (elts[i].quality < 1.0) {
                new->accept_q = 1;
            }
        }
    }

    new->accept_encodings =
        do_header_line(r->pool, apr_table_get(hdrs, "Accept-Encoding"));
    new->accept_langs =
        do_header_line(r->pool, apr_table_get(hdrs, "Accept-Language"));
    new->accept_charsets =
        do_header_line(r->pool, apr_table_get(hdrs, "Accept-Charset"));

    /* This is possibly overkill for some servers, heck, we have
     * only 33 index.html variants in docs/docroot (today).
     * Make this configurable?
     */
    new->avail_vars = apr_array_make(r->pool, 40, sizeof(var_rec));

    return new;
}


static void parse_negotiate_header(request_rec *r, negotiation_state *neg)
{
    const char *negotiate = apr_table_get(r->headers_in, "Negotiate");
    char *tok;

    /* First, default to no TCN, no Alternates, and the original Apache
     * negotiation algorithm with fiddles for broken browser configs.
     *
     * To save network bandwidth, we do not configure to send an
     * Alternates header to the user agent by default.  User
     * agents that want an Alternates header for agent-driven
     * negotiation will have to request it by sending an
     * appropriate Negotiate header.
     */
    neg->ua_supports_trans   = 0;
    neg->send_alternates     = 0;
    neg->may_choose          = 1;
    neg->use_rvsa            = 0;
    neg->dont_fiddle_headers = 0;

    if (!negotiate)
        return;

    if (strcmp(negotiate, "trans") == 0) {
        /* Lynx 2.7 and 2.8 send 'negotiate: trans' even though they
         * do not support transparent content negotiation, so for Lynx we
         * ignore the negotiate header when its contents are exactly "trans".
         * If future versions of Lynx ever need to say 'negotiate: trans',
         * they can send the equivalent 'negotiate: trans, trans' instead
         * to avoid triggering the workaround below.
         */
        const char *ua = apr_table_get(r->headers_in, "User-Agent");

        if (ua && (strncmp(ua, "Lynx", 4) == 0))
            return;
    }

    neg->may_choose = 0;  /* An empty Negotiate would require 300 response */

    while ((tok = ap_get_list_item(neg->pool, &negotiate)) != NULL) {

        if (strcmp(tok, "trans") == 0 ||
            strcmp(tok, "vlist") == 0 ||
            strcmp(tok, "guess-small") == 0 ||
            apr_isdigit(tok[0]) ||
            strcmp(tok, "*") == 0) {

            /* The user agent supports transparent negotiation */
            neg->ua_supports_trans = 1;

            /* Send-alternates could be configurable, but note
             * that it must be 1 if we have 'vlist' in the
             * negotiate header.
             */
            neg->send_alternates = 1;

            if (strcmp(tok, "1.0") == 0) {
                /* we may use the RVSA/1.0 algorithm, configure for it */
                neg->may_choose = 1;
                neg->use_rvsa = 1;
                neg->dont_fiddle_headers = 1;
            }
            else if (tok[0] == '*') {
                /* we may use any variant selection algorithm, configure
                 * to use the Apache algorithm
                 */
                neg->may_choose = 1;

                /* We disable header fiddles on the assumption that a
                 * client sending Negotiate knows how to send correct
                 * headers which don't need fiddling.
                 */
                neg->dont_fiddle_headers = 1;
            }
        }
    }

#ifdef NEG_DEBUG
    ap_log_error(APLOG_MARK, APLOG_STARTUP, 0, NULL,
            "dont_fiddle_headers=%d use_rvsa=%d ua_supports_trans=%d "
            "send_alternates=%d, may_choose=%d",
            neg->dont_fiddle_headers, neg->use_rvsa,
            neg->ua_supports_trans, neg->send_alternates, neg->may_choose);
#endif

}

/* Sometimes clients will give us no Accept info at all; this routine sets
 * up the standard default for that case, and also arranges for us to be
 * willing to run a CGI script if we find one.  (In fact, we set up to
 * dramatically prefer CGI scripts in cases where that's appropriate,
 * e.g., POST or when URI includes query args or extra path info).
 */
static void maybe_add_default_accepts(negotiation_state *neg,
                                      int prefer_scripts)
{
    accept_rec *new_accept;

    if (!neg->accepts) {
        neg->accepts = apr_array_make(neg->pool, 4, sizeof(accept_rec));

        new_accept = (accept_rec *) apr_array_push(neg->accepts);

        new_accept->name = "*/*";
        new_accept->quality = 1.0f;
        new_accept->level = 0.0f;
    }

    new_accept = (accept_rec *) apr_array_push(neg->accepts);

    new_accept->name = CGI_MAGIC_TYPE;
    if (neg->use_rvsa) {
        new_accept->quality = 0;
    }
    else {
        new_accept->quality = prefer_scripts ? 2.0f : 0.001f;
    }
    new_accept->level = 0.0f;
}

/*****************************************************************
 *
 * Parsing type-map files, in Roy's meta/http format augmented with
 * #-comments.
 */

/* Reading RFC822-style header lines, ignoring #-comments and
 * handling continuations.
 */

enum header_state {
    header_eof, header_seen, header_sep
};

static enum header_state get_header_line(char *buffer, int len, apr_file_t *map)
{
    char *buf_end = buffer + len;
    char *cp;
    char c;

    /* Get a noncommented line */

    do {
        if (apr_file_gets(buffer, MAX_STRING_LEN, map) != APR_SUCCESS) {
            return header_eof;
        }
    } while (buffer[0] == '#');

    /* If blank, just return it --- this ends information on this variant */

    for (cp = buffer; (*cp && apr_isspace(*cp)); ++cp) {
        continue;
    }

    if (*cp == '\0') {
        return header_sep;
    }

    /* If non-blank, go looking for header lines, but note that we still
     * have to treat comments specially...
     */

    cp += strlen(cp);

    /* We need to shortcut the rest of this block following the Body:
     * tag - we will not look for continutation after this line.
     */
    if (!strncasecmp(buffer, "Body:", 5))
        return header_seen;

    while (apr_file_getc(&c, map) != APR_EOF) {
        if (c == '#') {
            /* Comment line */
            while (apr_file_getc(&c, map) != APR_EOF && c != '\n') {
                continue;
            }
        }