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

         * mod_mime will _always_ provide us the base name in the
         * ap-mime-exception-list, if it processed anything.  If
         * this list is empty, give up immediately, there was
         * nothing interesting.  For example, looking at the files
         * readme.txt and readme.foo, we will throw away .foo if
         * it's an insignificant file (e.g. did not identify a
         * language, charset, encoding, content type or handler,)
         */
        exception_list =
            (apr_array_header_t *)apr_table_get(sub_req->notes,
                                                "ap-mime-exceptions-list");

        if (!exception_list) {
            ap_destroy_sub_req(sub_req);
            continue;
        }

        /* Each unregonized bit better match our base name, in sequence.
         * A test of index.html.foo will match index.foo or index.html.foo,
         * but it will never transpose the segments and allow index.foo.html
         * because that would introduce too much CPU consumption.  Better that
         * we don't attempt a many-to-many match here.
         */
        {
            int nexcept = exception_list->nelts;
            char **cur_except = (char**)exception_list->elts;
            char *segstart = filp, *segend, saveend;

            while (*segstart && nexcept) {
                if (!(segend = strchr(segstart, '.')))
                    segend = strchr(segstart, '\0');
                saveend = *segend;
                *segend = '\0';

#ifdef CASE_BLIND_FILESYSTEM
                if (strcasecmp(segstart, *cur_except) == 0) {
#else
                if (strcmp(segstart, *cur_except) == 0) {
#endif
                    --nexcept;
                    ++cur_except;
                }

                if (!saveend)
                    break;

                *segend = saveend;
                segstart = segend + 1;
            }

            if (nexcept) {
                /* Something you don't know is, something you don't know...
                 */
                ap_destroy_sub_req(sub_req);
                continue;
            }
        }

        /*
         * ###: be warned, the _default_ content type is already
         * picked up here!  If we failed the subrequest, or don't
         * know what we are serving, then continue.
         */
        if (sub_req->status != HTTP_OK || (!sub_req->content_type)) {
            ap_destroy_sub_req(sub_req);
            continue;
        }

        /* If it's a map file, we use that instead of the map
         * we're building...
         */
        if (((sub_req->content_type) &&
             !strcmp(sub_req->content_type, MAP_FILE_MAGIC_TYPE)) ||
            ((sub_req->handler) &&
             !strcmp(sub_req->handler, "type-map"))) {

            apr_dir_close(dirp);
            neg->avail_vars->nelts = 0;
            if (sub_req->status != HTTP_OK) {
                return sub_req->status;
            }
            return read_type_map(NULL, neg, sub_req);
        }

        /* Have reasonable variant --- gather notes. */

        mime_info.sub_req = sub_req;
        mime_info.file_name = apr_pstrdup(neg->pool, dirent.name);
        if (sub_req->content_encoding) {
            mime_info.content_encoding = sub_req->content_encoding;
        }
        if (sub_req->content_languages) {
            mime_info.content_languages = sub_req->content_languages;
        }

        get_entry(neg->pool, &accept_info, sub_req->content_type);
        set_mime_fields(&mime_info, &accept_info);

        new_var = apr_array_push(neg->avail_vars);
        memcpy(new_var, (void *) &mime_info, sizeof(var_rec));

        neg->count_multiviews_variants++;

        clean_var_rec(&mime_info);
    }

    apr_dir_close(dirp);

    /* We found some file names that matched.  None could be served.
     * Rather than fall out to autoindex or some other mapper, this
     * request must die.
     */
    if (anymatch && !neg->avail_vars->nelts) {
        ap_log_rerror(APLOG_MARK, APLOG_ERR, 0, r,
                      "Negotiation: discovered file(s) matching request: %s"
                      " (None could be negotiated).",
                      r->filename);
        return HTTP_NOT_FOUND;
    }

    set_vlist_validator(r, r);

    /* Sort the variants into a canonical order.  The negotiation
     * result sometimes depends on the order of the variants.  By
     * sorting the variants into a canonical order, rather than using
     * the order in which readdir() happens to return them, we ensure
     * that the negotiation result will be consistent over filesystem
     * backup/restores and over all mirror sites.
     */

    qsort((void *) neg->avail_vars->elts, neg->avail_vars->nelts,
          sizeof(var_rec), (int (*)(const void *, const void *)) variantsortf);

    return OK;
}


/*****************************************************************
 * And now for the code you've been waiting for... actually
 * finding a match to the client's requirements.
 */

/* Matching MIME types ... the star/star and foo/star commenting conventions
 * are implemented here.  (You know what I mean by star/star, but just
 * try mentioning those three characters in a C comment).  Using strcmp()
 * is legit, because everything has already been smashed to lowercase.
 *
 * Note also that if we get an exact match on the media type, we update
 * level_matched for use in level_cmp below...
 *
 * We also give a value for mime_stars, which is used later. It should
 * be 1 for star/star, 2 for type/star and 3 for type/subtype.
 */

static int mime_match(accept_rec *accept_r, var_rec *avail)
{
    const char *accept_type = accept_r->name;
    const char *avail_type = avail->mime_type;
    int len = strlen(accept_type);

    if (accept_type[0] == '*') {        /* Anything matches star/star */
        if (avail->mime_stars < 1) {
            avail->mime_stars = 1;
        }
        return 1;
    }
    else if ((accept_type[len - 1] == '*') &&
             !strncmp(accept_type, avail_type, len - 2)) {
        if (avail->mime_stars < 2) {
            avail->mime_stars = 2;
        }
        return 1;
    }
    else if (!strcmp(accept_type, avail_type)
             || (!strcmp(accept_type, "text/html")
                 && (!strcmp(avail_type, INCLUDES_MAGIC_TYPE)
                     || !strcmp(avail_type, INCLUDES_MAGIC_TYPE3)))) {
        if (accept_r->level >= avail->level) {
            avail->level_matched = avail->level;
            avail->mime_stars = 3;
            return 1;
        }
    }

    return OK;
}

/* This code implements a piece of the tie-breaking algorithm between
 * variants of equal quality.  This piece is the treatment of variants
 * of the same base media type, but different levels.  What we want to
 * return is the variant at the highest level that the client explicitly
 * claimed to accept.
 *
 * If all the variants available are at a higher level than that, or if
 * the client didn't say anything specific about this media type at all
 * and these variants just got in on a wildcard, we prefer the lowest
 * level, on grounds that that's the one that the client is least likely
 * to choke on.
 *
 * (This is all motivated by treatment of levels in HTML --- we only
 * want to give level 3 to browsers that explicitly ask for it; browsers
 * that don't, including HTTP/0.9 browsers that only get the implicit
 * "Accept: * / *" [space added to avoid confusing cpp --- no, that
 * syntax doesn't really work] should get HTML2 if available).
 *
 * (Note that this code only comes into play when we are choosing among
 * variants of equal quality, where the draft standard gives us a fair
 * bit of leeway about what to do.  It ain't specified by the standard;
 * rather, it is a choice made by this server about what to do in cases
 * where the standard does not specify a unique course of action).
 */

static int level_cmp(var_rec *var1, var_rec *var2)
{
    /* Levels are only comparable between matching media types */

    if (var1->is_pseudo_html && !var2->is_pseudo_html) {
        return 0;
    }

    if (!var1->is_pseudo_html && strcmp(var1->mime_type, var2->mime_type)) {
        return 0;
    }
    /* The result of the above if statements is that, if we get to
     * here, both variants have the same mime_type or both are
     * pseudo-html.
     */

    /* Take highest level that matched, if either did match. */

    if (var1->level_matched > var2->level_matched) {
        return 1;
    }
    if (var1->level_matched < var2->level_matched) {
        return -1;
    }

    /* Neither matched.  Take lowest level, if there's a difference. */

    if (var1->level < var2->level) {
        return 1;
    }
    if (var1->level > var2->level) {
        return -1;
    }

    /* Tied */

    return 0;
}

/* Finding languages.  The main entry point is set_language_quality()
 * which is called for each variant. It sets two elements in the
 * variant record:
 *    language_quality  - the 'q' value of the 'best' matching language
 *                        from Accept-Language: header (HTTP/1.1)
 *    lang_index    -     Non-negotiated language priority, using
 *                        position of language on the Accept-Language:
 *                        header, if present, else LanguagePriority
 *                        directive order.
 *
 * When we do the variant checking for best variant, we use language
 * quality first, and if a tie, language_index next (this only applies
 * when _not_ using the RVSA/1.0 algorithm). If using the RVSA/1.0
 * algorithm, lang_index is never used.
 *
 * set_language_quality() calls find_lang_index() and find_default_index()
 * to set lang_index.
 */

static int find_lang_index(apr_array_header_t *accept_langs, char *lang)
{
    const char **alang;
    int i;

    if (!lang || !accept_langs) {
        return -1;
    }

    alang = (const char **) accept_langs->elts;

    for (i = 0; i < accept_langs->nelts; ++i) {
        if (!strncmp(lang, *alang, strlen(*alang))) {
            return i;
        }
        alang += (accept_langs->elt_size / sizeof(char*));
    }

    return -1;
}

/* set_default_lang_quality() sets the quality we apply to variants
 * which have no language assigned to them. If none of the variants
 * have a language, we are not negotiating on language, so all are
 * acceptable, and we set the default q value to 1.0. However if
 * some of the variants have languages, we set this default to 0.0001.
 * The value of this default will be applied to all variants with
 * no explicit language -- which will have the effect of making them
 * acceptable, but only if no variants with an explicit language
 * are acceptable. The default q value set here is assigned to variants
 * with no language type in set_language_quality().
 *
 * Note that if using the RVSA/1.0 algorithm, we don't use this
 * fiddle.
 */

static void set_default_lang_quality(negotiation_state *neg)
{
    var_rec *avail_recs = (var_rec *) neg->avail_vars->elts;
    int j;

    if (!neg->dont_fiddle_headers) {
        for (j = 0; j < neg->avail_vars->nelts; ++j) {
            var_rec *variant = &avail_recs[j];
            if (variant->content_languages &&
                variant->content_languages->nelts) {
                neg->default_lang_quality = 0.0001f;
                return;
            }
        }
    }

    neg->default_lang_quality = 1.0f;
}

/* Set the language_quality value in the variant record. Also
 * assigns lang_index for ForceLanguagePriority.
 *
 * To find the language_quality value, we look for the 'q' value
 * of the 'best' matching language on the Accept-Language
 * header. The 'best' match is the language on Accept-Language
 * header which matches the language of this variant either fully,
 * or as far as the prefix marker (-). If two or more languages
 * match, use the longest string from the Accept-Language header
 * (see HTTP/1.1 [14.4])
 *
 * When a variant has multiple languages, we find the 'best'
 * match for each variant language tag as above, then select the
 * one with the highest q value. Because both the accept-header
 * and variant can have multiple languages, we now have a hairy
 * loop-within-a-loop here.
 *
 * If the variant has no language and we have no Accept-Language
 * items, leave the quality at 1.0 and return.
 *
 * If the variant has no language, we use the default as set by
 * set_default_lang_quality() (1.0 if we are not negotiating on
 * language, 0.001 if we are).
 *
 * Following the setting of the language quality, we drop through to
 * set the old 'lang_index'. This is set based on either the order
 * of the languages on the Accept-Language header, or the
 * order on the LanguagePriority directive. This is only used
 * in the negotiation if the language qualities tie.
 */

static void set_language_quality(negotiation_state *neg, var_rec *variant)
{
    int forcepriority = neg->conf->forcelangpriority;
    if (forcepriority == FLP_UNDEF) {
        forcepriority = FLP_DEFAULT;
    }

    if (!variant->content_languages || !variant->content_languages->nelts) {
        /* This variant has no content-language, so use the default
         * quality factor for variants with no content-language
         * (previously set by set_default_lang_quality()).
         * Leave the factor alone (it remains at 1.0) when we may not fiddle
         * with the headers.
         */
        if (!neg->dont_fiddle_headers) {
            variant->lang_quality = neg->default_lang_quality;
        }
        if (!neg->accept_langs) {
            return;             /* no accept-language header */
        }
        return;
    }
    else {
        /* Variant has one (or more) languages.  Look for the best
         * match. We do this by going through each language on the
         * variant description looking for a match on the
         * Accept-Language header. The best match is the longest
         * matching language on the header. The final result is the
         * best q value from all the languages on the variant
         * description.
         */

        if (!neg->accept_langs) {
            /* no accept-language header makes the variant indefinite */
            variant->definite = 0;
        }
        else {    /* There is an accept-language with 0 or more items */