mod_charset_lite.c

Apache HTTP Server 是一个功能强大的灵活的与HTTP/1.1相兼容的web服务器.这里给出的是Apache HTTP服务器的源码。

        if (reqinfo->output_ctx && !configured_on_output(r, XLATEOUT_FILTER_NAME)) {
            ap_add_output_filter(XLATEOUT_FILTER_NAME, reqinfo->output_ctx, r, 
                                 r->connection);
        }
        else if (dc->debug >= DBGLVL_FLOW) {
            ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
                          "xlate output filter not added implicitly because %s",
                          !reqinfo->output_ctx ? 
                          "no output configuration available" :
                          "another module added the filter");
        }

        if (reqinfo->input_ctx && !configured_on_input(r, XLATEIN_FILTER_NAME)) {
            ap_add_input_filter(XLATEIN_FILTER_NAME, reqinfo->input_ctx, r,
                                r->connection);
        }
        else if (dc->debug >= DBGLVL_FLOW) {
            ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
                          "xlate input filter not added implicitly because %s",
                          !reqinfo->input_ctx ?
                          "no input configuration available" :
                          "another module added the filter");
        }
    }
}

/* stuff that sucks that I know of:
 *
 * bucket handling:
 *  why create an eos bucket when we see it come down the stream?  just send the one
 *  passed as input...  news flash: this will be fixed when xlate_out_filter() starts
 *  using the more generic xlate_brigade()
 *
 * translation mechanics:
 *   we don't handle characters that straddle more than two buckets; an error
 *   will be generated
 */

/* send_downstream() is passed the translated data; it puts it in a single-
 * bucket brigade and passes the brigade to the next filter
 */
static apr_status_t send_downstream(ap_filter_t *f, const char *tmp, apr_size_t len)
{
    request_rec *r = f->r;
    conn_rec *c = r->connection;
    apr_bucket_brigade *bb;
    apr_bucket *b;
    charset_filter_ctx_t *ctx = f->ctx;
    apr_status_t rv;

    bb = apr_brigade_create(r->pool, c->bucket_alloc);
    b = apr_bucket_transient_create(tmp, len, c->bucket_alloc);
    APR_BRIGADE_INSERT_TAIL(bb, b);
    rv = ap_pass_brigade(f->next, bb);
    if (rv != APR_SUCCESS) {
        ctx->ees = EES_DOWNSTREAM;
    }
    return rv;
}

static apr_status_t send_eos(ap_filter_t *f)
{
    request_rec *r = f->r;
    conn_rec *c = r->connection;
    apr_bucket_brigade *bb;
    apr_bucket *b;
    charset_filter_ctx_t *ctx = f->ctx;
    apr_status_t rv;

    bb = apr_brigade_create(r->pool, c->bucket_alloc);
    b = apr_bucket_eos_create(c->bucket_alloc);
    APR_BRIGADE_INSERT_TAIL(bb, b);
    rv = ap_pass_brigade(f->next, bb);
    if (rv != APR_SUCCESS) {
        ctx->ees = EES_DOWNSTREAM;
    }
    return rv;
}

static apr_status_t set_aside_partial_char(charset_filter_ctx_t *ctx, 
                                           const char *partial,
                                           apr_size_t partial_len)
{
    apr_status_t rv;

    if (sizeof(ctx->buf) > partial_len) {
        ctx->saved = partial_len;
        memcpy(ctx->buf, partial, partial_len);
        rv = APR_SUCCESS;
    }
    else {
        rv = APR_INCOMPLETE;
        ctx->ees = EES_LIMIT; /* we don't handle chars this wide which straddle 
                               * buckets 
                               */
    }
    return rv;
}

static apr_status_t finish_partial_char(charset_filter_ctx_t *ctx,
                                        /* input buffer: */
                                        const char **cur_str, 
                                        apr_size_t *cur_len,
                                        /* output buffer: */
                                        char **out_str,
                                        apr_size_t *out_len)
{
    apr_status_t rv;
    apr_size_t tmp_input_len;

    /* Keep adding bytes from the input string to the saved string until we
     *    1) finish the input char
     *    2) get an error
     * or 3) run out of bytes to add
     */

    do {
        ctx->buf[ctx->saved] = **cur_str;
        ++ctx->saved;
        ++*cur_str;
        --*cur_len;
        tmp_input_len = ctx->saved;
        rv = apr_xlate_conv_buffer(ctx->xlate,
                                   ctx->buf,
                                   &tmp_input_len,
                                   *out_str,
                                   out_len);
    } while (rv == APR_INCOMPLETE && *cur_len);

    if (rv == APR_SUCCESS) {
        ctx->saved = 0;
    }
    else {
        ctx->ees = EES_LIMIT; /* code isn't smart enough to handle chars
                               * straddling more than two buckets
                               */
    }

    return rv;
}

static void log_xlate_error(ap_filter_t *f, apr_status_t rv)
{
    charset_filter_ctx_t *ctx = f->ctx;
    const char *msg;
    char msgbuf[100];
    int cur;

    switch(ctx->ees) {
    case EES_LIMIT:
        rv = 0;
        msg = "xlate filter - a built-in restriction was encountered";
        break;
    case EES_BAD_INPUT:
        rv = 0;
        msg = "xlate filter - an input character was invalid";
        break;
    case EES_BUCKET_READ:
        rv = 0;
        msg = "xlate filter - bucket read routine failed";
        break;
    case EES_INCOMPLETE_CHAR:
        rv = 0;
        strcpy(msgbuf, "xlate filter - incomplete char at end of input - ");
        cur = 0;
        while ((apr_size_t)cur < ctx->saved) {
            apr_snprintf(msgbuf + strlen(msgbuf), sizeof(msgbuf) - strlen(msgbuf), 
                         "%02X", (unsigned)ctx->buf[cur]);
            ++cur;
        }
        msg = msgbuf;
        break;
    case EES_DOWNSTREAM:
        msg = "xlate filter - an error occurred in a lower filter";
        break;
    default:
        msg = "xlate filter - returning error";
    }
    ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, f->r,
                  "%s", msg);
}

/* chk_filter_chain() is called once per filter instance; it tries to
 * determine if the current filter instance should be disabled because
 * its translation is incompatible with the translation of an existing
 * instance of the translate filter
 *
 * Example bad scenario:
 *
 *   configured filter chain for the request:
 *     INCLUDES XLATEOUT(8859-1->UTS-16)
 *   configured filter chain for the subrequest:
 *     XLATEOUT(8859-1->UTS-16)
 *
 *   When the subrequest is processed, the filter chain will be
 *     XLATEOUT(8859-1->UTS-16) XLATEOUT(8859-1->UTS-16)
 *   This makes no sense, so the instance of XLATEOUT added for the
 *   subrequest will be noop-ed.
 *
 * Example good scenario:
 *
 *   configured filter chain for the request:
 *     INCLUDES XLATEOUT(8859-1->UTS-16)
 *   configured filter chain for the subrequest:
 *     XLATEOUT(IBM-1047->8859-1)
 *
 *   When the subrequest is processed, the filter chain will be
 *     XLATEOUT(IBM-1047->8859-1) XLATEOUT(8859-1->UTS-16)
 *   This makes sense, so the instance of XLATEOUT added for the
 *   subrequest will be left alone and it will translate from
 *   IBM-1047->8859-1.
 */
static void chk_filter_chain(ap_filter_t *f)
{
    ap_filter_t *curf;
    charset_filter_ctx_t *curctx, *last_xlate_ctx = NULL,
        *ctx = f->ctx;
    int debug = ctx->dc->debug;
    int output = !strcasecmp(f->frec->name, XLATEOUT_FILTER_NAME);

    if (ctx->noop) {
        return;
    }

    /* walk the filter chain; see if it makes sense for our filter to
     * do any translation
     */
    curf = output ? f->r->output_filters : f->r->input_filters;
    while (curf) {
        if (!strcasecmp(curf->frec->name, f->frec->name) &&
            curf->ctx) {
            curctx = (charset_filter_ctx_t *)curf->ctx;
            if (!last_xlate_ctx) {
                last_xlate_ctx = curctx;
            }
            else {
                if (strcmp(last_xlate_ctx->dc->charset_default,
                           curctx->dc->charset_source)) {
                    /* incompatible translation 
                     * if our filter instance is incompatible with an instance
                     * already in place, noop our instance
                     * Notes: 
                     * . We are only willing to noop our own instance.
                     * . It is possible to noop another instance which has not
                     *   yet run, but this is not currently implemented.
                     *   Hopefully it will not be needed.
                     * . It is not possible to noop an instance which has 
                     *   already run.
                     */
                    if (last_xlate_ctx == f->ctx) {
                        last_xlate_ctx->noop = 1;
                        if (debug >= DBGLVL_PMC) {
                            const char *symbol = output ? "->" : "<-";

                            ap_log_rerror(APLOG_MARK, APLOG_DEBUG,
                                          0, f->r,
                                          "%s %s - disabling "
                                          "translation %s%s%s; existing "
                                          "translation %s%s%s",
                                          f->r->uri ? "uri" : "file",
                                          f->r->uri ? f->r->uri : f->r->filename,
                                          last_xlate_ctx->dc->charset_source,
                                          symbol,
                                          last_xlate_ctx->dc->charset_default,
                                          curctx->dc->charset_source,
                                          symbol,
                                          curctx->dc->charset_default);
                        }
                    }
                    else {
                        const char *symbol = output ? "->" : "<-";

                        ap_log_rerror(APLOG_MARK, APLOG_ERR,
                                      0, f->r,
                                      "chk_filter_chain() - can't disable "
                                      "translation %s%s%s; existing "
                                      "translation %s%s%s",
                                      last_xlate_ctx->dc->charset_source,
                                      symbol,
                                      last_xlate_ctx->dc->charset_default,
                                      curctx->dc->charset_source,
                                      symbol,
                                      curctx->dc->charset_default);
                    }
                    break;
                }
            }
        }
        curf = curf->next;
    }
}

/* xlate_brigade() is used to filter request and response bodies
 *
 * we'll stop when one of the following occurs:
 * . we run out of buckets
 * . we run out of space in the output buffer
 * . we hit an error
 *
 * inputs:
 *   bb:               brigade to process
 *   buffer:           storage to hold the translated characters
 *   buffer_size:      size of buffer
 *   (and a few more uninteresting parms)
 *
 * outputs:
 *   return value:     APR_SUCCESS or some error code
 *   bb:               we've removed any buckets representing the
 *                     translated characters; the eos bucket, if
 *                     present, will be left in the brigade
 *   buffer:           filled in with translated characters
 *   buffer_size:      updated with the bytes remaining
 *   hit_eos:          did we hit an EOS bucket?
 */
static apr_status_t xlate_brigade(charset_filter_ctx_t *ctx,
                                  apr_bucket_brigade *bb,
                                  char *buffer, 
                                  apr_size_t *buffer_avail,
                                  int *hit_eos)
{
    apr_bucket *b = NULL; /* set to NULL only to quiet some gcc */
    apr_bucket *consumed_bucket;
    const char *bucket;
    apr_size_t bytes_in_bucket; /* total bytes read from current bucket */
    apr_size_t bucket_avail;    /* bytes left in current bucket */
    apr_status_t rv = APR_SUCCESS;

    *hit_eos = 0;
    bucket_avail = 0;
    consumed_bucket = NULL;
    while (1) {
        if (!bucket_avail) { /* no bytes left to process in the current bucket... */
            if (consumed_bucket) {
                apr_bucket_delete(consumed_bucket);
                consumed_bucket = NULL;
            }
            b = APR_BRIGADE_FIRST(bb);
            if (b == APR_BRIGADE_SENTINEL(bb) ||
                APR_BUCKET_IS_EOS(b)) {
                break;
            }
            rv = apr_bucket_read(b, &bucket, &bytes_in_bucket, APR_BLOCK_READ);
            if (rv != APR_SUCCESS) {
                ctx->ees = EES_BUCKET_READ;
                break;
            }
            bucket_avail = bytes_in_bucket;
            consumed_bucket = b;   /* for axing when we're done reading it */
        }
        if (bucket_avail) {
            /* We've got data, so translate it. */
            if (ctx->saved) {
                /* Rats... we need to finish a partial character from the previous
                 * bucket.
                 *
                 * Strangely, finish_partial_char() increments the input buffer
                 * pointer but does not increment the output buffer pointer.
                 */
                apr_size_t old_buffer_avail = *buffer_avail;
                rv = finish_partial_char(ctx,
                                         &bucket, &bucket_avail,