mod_ext_filter.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。

     * and QUERY_STRING_UNESCAPED
     */
    ap_add_cgi_vars(f->r);
    ap_add_common_vars(f->r);
    apr_table_setn(f->r->subprocess_env, "DOCUMENT_URI", f->r->uri);
    apr_table_setn(f->r->subprocess_env, "DOCUMENT_PATH_INFO", f->r->path_info);
    if (f->r->args) {
            /* QUERY_STRING is added by ap_add_cgi_vars */
        char *arg_copy = apr_pstrdup(f->r->pool, f->r->args);
        ap_unescape_url(arg_copy);
        apr_table_setn(f->r->subprocess_env, "QUERY_STRING_UNESCAPED",
                       ap_escape_shell_cmd(f->r->pool, arg_copy));
    }

    env = (const char * const *) ap_create_environment(ctx->p,
                                                       f->r->subprocess_env);

    rc = apr_proc_create(ctx->proc, 
                            ctx->filter->command, 
                            (const char * const *)ctx->filter->args, 
                            env, /* environment */
                            ctx->procattr, 
                            ctx->p);
    if (rc != APR_SUCCESS) {
        ap_log_rerror(APLOG_MARK, APLOG_ERR, rc, f->r,
                      "couldn't create child process to run `%s'",
                      ctx->filter->command);
        return rc;
    }

    apr_pool_note_subprocess(ctx->p, ctx->proc, APR_KILL_AFTER_TIMEOUT);

    /* We don't want the handle to the child's stdin inherited by any
     * other processes created by httpd.  Otherwise, when we close our
     * handle, the child won't see EOF because another handle will still
     * be open.
     */

    apr_pool_cleanup_register(ctx->p, ctx->proc->in, 
                         apr_pool_cleanup_null, /* other mechanism */
                         ef_close_file);

#if APR_FILES_AS_SOCKETS
    {
        apr_socket_t *newsock;

        rc = apr_poll_setup(&ctx->pollset, 2, ctx->p);
        ap_assert(rc == APR_SUCCESS);
        rc = apr_socket_from_file(&newsock, ctx->proc->in);
        ap_assert(rc == APR_SUCCESS);
        rc = apr_poll_socket_add(ctx->pollset, newsock, APR_POLLOUT);
        ap_assert(rc == APR_SUCCESS);
        rc = apr_socket_from_file(&newsock, ctx->proc->out);
        ap_assert(rc == APR_SUCCESS);
        rc = apr_poll_socket_add(ctx->pollset, newsock, APR_POLLIN);
        ap_assert(rc == APR_SUCCESS);
    }
#endif

    return APR_SUCCESS;
}

static const char *get_cfg_string(ef_dir_t *dc, ef_filter_t *filter, apr_pool_t *p)
{
    const char *debug_str = dc->debug == -1 ? 
        "DebugLevel=0" : apr_psprintf(p, "DebugLevel=%d", dc->debug);
    const char *log_stderr_str = dc->log_stderr < 1 ?
        "NoLogStderr" : "LogStderr";
    const char *preserve_content_length_str = filter->preserves_content_length ?
        "PreservesContentLength" : "!PreserveContentLength";
    const char *intype_str = !filter->intype ?
        "*/*" : filter->intype;
    const char *outtype_str = !filter->outtype ?
        "(unchanged)" : filter->outtype;
    
    return apr_psprintf(p,
                        "ExtFilterOptions %s %s %s ExtFilterInType %s "
                        "ExtFilterOuttype %s",
                        debug_str, log_stderr_str, preserve_content_length_str,
                        intype_str, outtype_str);
}

static ef_filter_t *find_filter_def(const server_rec *s, const char *fname)
{
    ef_server_t *sc;
    ef_filter_t *f;

    sc = ap_get_module_config(s->module_config, &ext_filter_module);
    f = apr_hash_get(sc->h, fname, APR_HASH_KEY_STRING);
    if (!f && s != main_server) {
        s = main_server;
        sc = ap_get_module_config(s->module_config, &ext_filter_module);
        f = apr_hash_get(sc->h, fname, APR_HASH_KEY_STRING);
    }
    return f;
}

static apr_status_t init_filter_instance(ap_filter_t *f)
{
    ef_ctx_t *ctx;
    ef_dir_t *dc;
    apr_status_t rv;

    f->ctx = ctx = apr_pcalloc(f->r->pool, sizeof(ef_ctx_t));
    dc = ap_get_module_config(f->r->per_dir_config,
                              &ext_filter_module);
    ctx->dc = dc;
    /* look for the user-defined filter */
    ctx->filter = find_filter_def(f->r->server, f->frec->name);
    if (!ctx->filter) {
        ap_log_rerror(APLOG_MARK, APLOG_ERR, 0, f->r,
                      "couldn't find definition of filter '%s'",
                      f->frec->name);
        return APR_EINVAL;
    }
    ctx->p = f->r->pool;
    if (ctx->filter->intype &&
        ctx->filter->intype != INTYPE_ALL) {
        if (!f->r->content_type) {
            ctx->noop = 1;
        }
        else {
            const char *ctypes = f->r->content_type;
            const char *ctype = ap_getword(f->r->pool, &ctypes, ';');

            if (strcasecmp(ctx->filter->intype, ctype)) {
                /* wrong IMT for us; don't mess with the output */
                ctx->noop = 1;
            }
        }
    }
    if (ctx->filter->enable_env &&
        !apr_table_get(f->r->subprocess_env, ctx->filter->enable_env)) {
        /* an environment variable that enables the filter isn't set; bail */
        ctx->noop = 1;
    }
    if (ctx->filter->disable_env &&
        apr_table_get(f->r->subprocess_env, ctx->filter->disable_env)) {
        /* an environment variable that disables the filter is set; bail */
        ctx->noop = 1;
    }
    if (!ctx->noop) {
        rv = init_ext_filter_process(f);
        if (rv != APR_SUCCESS) {
            return rv;
        }
        if (ctx->filter->outtype &&
            ctx->filter->outtype != OUTTYPE_UNCHANGED) {
            ap_set_content_type(f->r, ctx->filter->outtype);
        }
        if (ctx->filter->preserves_content_length != 1) {
            /* nasty, but needed to avoid confusing the browser 
             */
            apr_table_unset(f->r->headers_out, "Content-Length");
        }
    }

    if (dc->debug >= DBGLVL_SHOWOPTIONS) {
        ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, f->r,
                      "%sfiltering `%s' of type `%s' through `%s', cfg %s",
                      ctx->noop ? "NOT " : "",
                      f->r->uri ? f->r->uri : f->r->filename,
                      f->r->content_type ? f->r->content_type : "(unspecified)",
                      ctx->filter->command,
                      get_cfg_string(dc, ctx->filter, f->r->pool));
    }

    return APR_SUCCESS;
}

/* drain_available_output(): 
 *
 * if any data is available from the filter, read it and pass it
 * to the next filter
 */
static apr_status_t drain_available_output(ap_filter_t *f)
{
    request_rec *r = f->r;
    conn_rec *c = r->connection;
    ef_ctx_t *ctx = f->ctx;
    ef_dir_t *dc = ctx->dc;
    apr_size_t len;
    char buf[4096];
    apr_status_t rv;
    apr_bucket_brigade *bb;
    apr_bucket *b;

    while (1) {
        len = sizeof(buf);
        rv = apr_file_read(ctx->proc->out,
                      buf,
                      &len);
        if ((rv && !APR_STATUS_IS_EAGAIN(rv)) ||
            dc->debug >= DBGLVL_GORY) {
            ap_log_rerror(APLOG_MARK, APLOG_DEBUG, rv, r,
                          "apr_file_read(child output), len %" APR_SIZE_T_FMT,
                          !rv ? len : -1);
        }
        if (rv != APR_SUCCESS) {
            return rv;
        }
        bb = apr_brigade_create(r->pool, c->bucket_alloc);
        b = apr_bucket_transient_create(buf, len, c->bucket_alloc);
        APR_BRIGADE_INSERT_TAIL(bb, b);
        if ((rv = ap_pass_brigade(f->next, bb)) != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
                          "ap_pass_brigade()");
            return rv;
        }
    }
    /* we should never get here; if we do, a bogus error message would be
     * the least of our problems 
     */
    return APR_ANONYMOUS;
}

static apr_status_t pass_data_to_filter(ap_filter_t *f, const char *data, 
                                        apr_size_t len)
{
    ef_ctx_t *ctx = f->ctx;
    ef_dir_t *dc = ctx->dc;
    apr_status_t rv;
    apr_size_t bytes_written = 0;
    apr_size_t tmplen;
    
    do {
        tmplen = len - bytes_written;
        rv = apr_file_write(ctx->proc->in,
                       (const char *)data + bytes_written,
                       &tmplen);
        bytes_written += tmplen;
        if (rv != APR_SUCCESS && !APR_STATUS_IS_EAGAIN(rv)) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, f->r,
                          "apr_file_write(child input), len %" APR_SIZE_T_FMT,
                          tmplen);
            return rv;
        }
        if (APR_STATUS_IS_EAGAIN(rv)) {
            /* XXX handle blocking conditions here...  if we block, we need 
             * to read data from the child process and pass it down to the
             * next filter!
             */
            rv = drain_available_output(f);
            if (APR_STATUS_IS_EAGAIN(rv)) {
#if APR_FILES_AS_SOCKETS
                int num_events;
                
                rv = apr_poll(ctx->pollset, 2,
                              &num_events, f->r->server->timeout);
                if (rv || dc->debug >= DBGLVL_GORY) {
                    ap_log_rerror(APLOG_MARK, APLOG_DEBUG,
                                  rv, f->r, "apr_poll()");
                }
                if (rv != APR_SUCCESS && !APR_STATUS_IS_EINTR(rv)) { 
                    /* some error such as APR_TIMEUP */
                    return rv;
                }
#else /* APR_FILES_AS_SOCKETS */
                /* Yuck... I'd really like to wait until I can read
                 * or write, but instead I have to sleep and try again 
                 */
                apr_sleep(100000); /* 100 milliseconds */
                if (dc->debug >= DBGLVL_GORY) {
                    ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 
                                  0, f->r, "apr_sleep()");
                }
#endif /* APR_FILES_AS_SOCKETS */
            }
            else if (rv != APR_SUCCESS) {
                return rv;
            }
        }
    } while (bytes_written < len);
    return rv;
}

static apr_status_t ef_output_filter(ap_filter_t *f, apr_bucket_brigade *bb)
{
    request_rec *r = f->r;
    conn_rec *c = r->connection;
    ef_ctx_t *ctx = f->ctx;
    apr_bucket *b;
    ef_dir_t *dc;
    apr_size_t len;
    const char *data;
    apr_status_t rv;
    char buf[4096];
    apr_bucket *eos = NULL;

    if (!ctx) {
        if ((rv = init_filter_instance(f)) != APR_SUCCESS) {
            return rv;
        }
        ctx = f->ctx;
    }
    if (ctx->noop) {
        ap_remove_output_filter(f);
        return ap_pass_brigade(f->next, bb);
    }
    dc = ctx->dc;

    APR_BRIGADE_FOREACH(b, bb) {

        if (APR_BUCKET_IS_EOS(b)) {
            eos = b;
            break;
        }

        rv = apr_bucket_read(b, &data, &len, APR_BLOCK_READ);
        if (rv != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r, "apr_bucket_read()");
            return rv;
        }

        /* Good cast, we just tested len isn't negative */
        if (len > 0 &&
            (rv = pass_data_to_filter(f, data, (apr_size_t)len)) 
                != APR_SUCCESS) {
            return rv;
        }
    }

    apr_brigade_destroy(bb);

    /* XXX What we *really* need to do once we've hit eos is create a pipe bucket
     * from the child output pipe and pass down the pipe bucket + eos.
     */
    if (eos) {
        /* close the child's stdin to signal that no more data is coming;
         * that will cause the child to finish generating output
         */
        if ((rv = apr_file_close(ctx->proc->in)) != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
                          "apr_file_close(child input)");
            return rv;
        }
        /* since we've seen eos and closed the child's stdin, set the proper pipe 
         * timeout; we don't care if we don't return from apr_file_read() for a while... 
         */
        rv = apr_file_pipe_timeout_set(ctx->proc->out, 
                                       r->server->timeout);
        if (rv) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
                          "apr_file_pipe_timeout_set(child output)");
            return rv;
        }
    }

    do {
        len = sizeof(buf);
        rv = apr_file_read(ctx->proc->out,
                      buf,
                      &len);
        if ((rv && !APR_STATUS_IS_EOF(rv) && !APR_STATUS_IS_EAGAIN(rv)) ||
            dc->debug >= DBGLVL_GORY) {
            ap_log_rerror(APLOG_MARK, APLOG_DEBUG, rv, r,
                          "apr_file_read(child output), len %" APR_SIZE_T_FMT,
                          !rv ? len : -1);
        }
        if (APR_STATUS_IS_EAGAIN(rv)) {
            if (eos) {
                /* should not occur, because we have an APR timeout in place */
                AP_DEBUG_ASSERT(1 != 1);
            }
            return APR_SUCCESS;
        }
        
        if (rv == APR_SUCCESS) {
            bb = apr_brigade_create(r->pool, c->bucket_alloc);
            b = apr_bucket_transient_create(buf, len, c->bucket_alloc);
            APR_BRIGADE_INSERT_TAIL(bb, b);
            if ((rv = ap_pass_brigade(f->next, bb)) != APR_SUCCESS) {
                ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
                              "ap_pass_brigade(filtered buffer) failed");
                return rv;
            }
        }
    } while (rv == APR_SUCCESS);

    if (!APR_STATUS_IS_EOF(rv)) {
        return rv;
    }

    if (eos) {
        /* pass down eos */
        bb = apr_brigade_create(r->pool, c->bucket_alloc);
        b = apr_bucket_eos_create(c->bucket_alloc);
        APR_BRIGADE_INSERT_TAIL(bb, b);
        if ((rv = ap_pass_brigade(f->next, bb)) != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
                          "ap_pass_brigade(eos) failed");
            return rv;
        }
    }

    return APR_SUCCESS;
}

#if 0
static int ef_input_filter(ap_filter_t *f, apr_bucket_brigade *bb, 
                           ap_input_mode_t mode, apr_read_type_e block,
                           apr_off_t readbytes)
{
    apr_status_t rv;
    apr_bucket *b;
    char *buf;
    apr_ssize_t len;
    char *zero;

    rv = ap_get_brigade(f->next, bb, mode, block, readbytes);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    APR_BRIGADE_FOREACH(b, bb) {
        if (!APR_BUCKET_IS_EOS(b)) {
            if ((rv = apr_bucket_read(b, (const char **)&buf, &len, APR_BLOCK_READ)) != APR_SUCCESS) {
                ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, f->r, "apr_bucket_read() failed");
                return rv;
            }
            ap_log_error(APLOG_MARK, APLOG_ERR, 0, NULL, "apr_bucket_read -> %d bytes",
                         len);
            while ((zero = memchr(buf, '0', len))) {
                *zero = 'a';
            }
        }
        else
            ap_log_error(APLOG_MARK, APLOG_ERR, 0, NULL, "got eos bucket");
    }

    return rv;
}
#endif

module AP_MODULE_DECLARE_DATA ext_filter_module =
{
    STANDARD20_MODULE_STUFF,
    create_ef_dir_conf,
    merge_ef_dir_conf,
    create_ef_server_conf,
    NULL,
    cmds,
    register_hooks
};