diff --git a/CMakeLists.txt b/CMakeLists.txt index 58767c6..5b1bd3d 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -70,6 +70,16 @@ include_directories (${PROJECT_BINARY_DIR}) add_executable (${PROJECT_NAME} ${PROJECT_NAME}.c http-parser/http_parser.c) target_link_libraries (${PROJECT_NAME} ${project_libraries}) +# Development tools +find_package (LibMagic) +if (LIBMAGIC_FOUND) + include_directories (${LIBMAGIC_INCLUDE_DIRS}) + add_executable (json-rpc-test-server + json-rpc-test-server.c http-parser/http_parser.c) + target_link_libraries (json-rpc-test-server + ${project_libraries} ${LIBMAGIC_LIBRARIES}) +endif () + # The files to be installed include (GNUInstallDirs) install (TARGETS ${PROJECT_NAME} DESTINATION ${CMAKE_INSTALL_BINDIR}) diff --git a/LICENSE b/LICENSE index 438cd83..a069461 100644 --- a/LICENSE +++ b/LICENSE @@ -1,4 +1,4 @@ -Copyright (c) 2014 - 2016, Přemysl Janouch +Copyright (c) 2014 - 2018, Přemysl Janouch Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby granted. diff --git a/README.adoc b/README.adoc index 3112f9f..9c4fec4 100644 --- a/README.adoc +++ b/README.adoc @@ -67,6 +67,12 @@ Note that for versions of CMake before 2.8.9, you need to prefix `cpack` with Run the program with `--help` to obtain usage information. +Test server +----------- +If you install development packages for libmagic, an included test server will +be built but not installed which provides a trivial JSON-RPC 2.0 service with +FastCGI, SCGI, and WebSocket interfaces. It responds to the `ping` method. + Contributing and Support ------------------------ Use https://git.janouch.name/p/json-rpc-shell to report bugs, request features, diff --git a/cmake/FindLibMagic.cmake b/cmake/FindLibMagic.cmake new file mode 100644 index 0000000..9bed3b3 --- /dev/null +++ b/cmake/FindLibMagic.cmake @@ -0,0 +1,10 @@ +# Public Domain + +find_library (LIBMAGIC_LIBRARIES magic) +find_path (LIBMAGIC_INCLUDE_DIRS magic.h) + +include (FindPackageHandleStandardArgs) +find_package_handle_standard_args (LibMagic DEFAULT_MSG + LIBMAGIC_LIBRARIES LIBMAGIC_INCLUDE_DIRS) + +mark_as_advanced (LIBMAGIC_LIBRARIES LIBMAGIC_INCLUDE_DIRS) diff --git a/json-rpc-test-server.c b/json-rpc-test-server.c new file mode 100644 index 0000000..ea38b7b --- /dev/null +++ b/json-rpc-test-server.c @@ -0,0 +1,2928 @@ +/* + * json-rpc-test-server.c: JSON-RPC 2.0 demo server + * + * Copyright (c) 2015 - 2018, Přemysl Janouch + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY + * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION + * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN + * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + * + */ + +#define print_fatal_data ((void *) LOG_ERR) +#define print_error_data ((void *) LOG_ERR) +#define print_warning_data ((void *) LOG_WARNING) +#define print_status_data ((void *) LOG_INFO) +#define print_debug_data ((void *) LOG_DEBUG) + +#define LIBERTY_WANT_SSL +#define LIBERTY_WANT_PROTO_HTTP +#define LIBERTY_WANT_PROTO_WS +#define LIBERTY_WANT_PROTO_SCGI +#define LIBERTY_WANT_PROTO_FASTCGI + +#include "config.h" +#undef PROGRAM_NAME +#define PROGRAM_NAME "json-rpc-test-server" + +#include "liberty/liberty.c" + +#include +#include +#include +#include + +#include +#include +#include + +#include "http-parser/http_parser.h" + +enum { PIPE_READ, PIPE_WRITE }; + +#define FIND_CONTAINER(name, pointer, type, member) \ + type *name = CONTAINER_OF (pointer, type, member) + +// --- Utilities --------------------------------------------------------------- + +static bool +flush_queue (struct write_queue *queue, int fd) +{ + struct iovec vec[queue->len], *vec_iter = vec; + LIST_FOR_EACH (struct write_req, iter, queue->head) + *vec_iter++ = iter->data; + + ssize_t written; +again: + if ((written = writev (fd, vec, N_ELEMENTS (vec))) >= 0) + { + write_queue_processed (queue, written); + return true; + } + if (errno == EINTR) + goto again; + if (errno == EAGAIN) + return true; + + return false; +} + +// --- Logging ----------------------------------------------------------------- + +static void +log_message_syslog (void *user_data, const char *quote, const char *fmt, + va_list ap) +{ + int prio = (int) (intptr_t) user_data; + + va_list va; + va_copy (va, ap); + int size = vsnprintf (NULL, 0, fmt, va); + va_end (va); + if (size < 0) + return; + + char buf[size + 1]; + if (vsnprintf (buf, sizeof buf, fmt, ap) >= 0) + syslog (prio, "%s%s", quote, buf); +} + +// --- FastCGI ----------------------------------------------------------------- +/// @defgroup FastCGI +/// @{ + +enum fcgi_request_state +{ + FCGI_REQUEST_PARAMS, ///< Reading headers + FCGI_REQUEST_STDIN ///< Reading input +}; + +struct fcgi_request +{ + struct fcgi_muxer *muxer; ///< The parent muxer + uint16_t request_id; ///< The ID of this request + uint8_t flags; ///< Request flags + + enum fcgi_request_state state; ///< Parsing state + struct str_map headers; ///< Headers + struct fcgi_nv_parser hdr_parser; ///< Header parser + + struct str output_buffer; ///< Output buffer + + void *handler_data; ///< Handler data +}; + +/// Handles a single FastCGI connection, de/multiplexing requests and responses +struct fcgi_muxer +{ + struct fcgi_parser parser; ///< FastCGI message parser + uint32_t active_requests; ///< The number of active requests + bool in_shutdown; ///< Rejecting new requests + + // Virtual method callbacks: + + /// Write data to the underlying transport + void (*write_cb) (struct fcgi_muxer *, const void *data, size_t len); + + /// Close the underlying transport. You are allowed to destroy the muxer + /// directly from within the callback. + void (*close_cb) (struct fcgi_muxer *); + + /// Start processing a request. Return false if no further action is + /// to be done and the request should be finished. + bool (*request_start_cb) (struct fcgi_request *); + + /// Handle incoming data. "len == 0" means EOF. Returns false if + /// the underlying transport should be closed, this being the last request. + bool (*request_push_cb) + (struct fcgi_request *, const void *data, size_t len); + + /// Destroy the handler's data stored in the request object + void (*request_finalize_cb) (struct fcgi_request *); + + /// Requests assigned to request IDs (may not be FCGI_NULL_REQUEST_ID) + struct fcgi_request *requests[1 << 8]; +}; + +static void +fcgi_muxer_send (struct fcgi_muxer *self, + enum fcgi_type type, uint16_t request_id, const void *data, size_t len) +{ + hard_assert (len <= UINT16_MAX); + + struct str message = str_make (); + static char zeroes[8]; + size_t padding = -len & 7; + + str_pack_u8 (&message, FCGI_VERSION_1); + str_pack_u8 (&message, type); + str_pack_u16 (&message, request_id); + str_pack_u16 (&message, len); // content length + str_pack_u8 (&message, padding); // padding length + str_pack_u8 (&message, 0); // reserved + + str_append_data (&message, data, len); + str_append_data (&message, zeroes, padding); + + // XXX: we should probably have another write_cb that assumes ownership + self->write_cb (self, message.str, message.len); + str_free (&message); +} + +static void +fcgi_muxer_send_end_request (struct fcgi_muxer *self, uint16_t request_id, + uint32_t app_status, enum fcgi_protocol_status protocol_status) +{ + uint8_t content[8] = { app_status >> 24, app_status >> 16, + app_status << 8, app_status, protocol_status }; + fcgi_muxer_send (self, FCGI_END_REQUEST, request_id, + content, sizeof content); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static struct fcgi_request * +fcgi_request_new (void) +{ + struct fcgi_request *self = xcalloc (1, sizeof *self); + + self->headers = str_map_make (free); + + self->hdr_parser = fcgi_nv_parser_make (); + self->hdr_parser.output = &self->headers; + + self->output_buffer = str_make (); + return self; +} + +static void +fcgi_request_destroy (struct fcgi_request *self) +{ + // TODO: consider the case where it hasn't been started yet + self->muxer->request_finalize_cb (self); + + str_map_free (&self->headers); + fcgi_nv_parser_free (&self->hdr_parser); + free (self); +} + +static void +fcgi_request_flush (struct fcgi_request *self) +{ + if (!self->output_buffer.len) + return; + + fcgi_muxer_send (self->muxer, FCGI_STDOUT, self->request_id, + self->output_buffer.str, self->output_buffer.len); + str_reset (&self->output_buffer); +} + +static void +fcgi_request_write (struct fcgi_request *self, const void *data, size_t len) +{ + // We're buffering the output and splitting it into messages + bool need_flush = true; + while (len) + { + size_t to_write = UINT16_MAX - self->output_buffer.len; + if (to_write > len) + { + to_write = len; + need_flush = false; + } + + str_append_data (&self->output_buffer, data, to_write); + data = (uint8_t *) data + to_write; + len -= to_write; + + if (need_flush) + fcgi_request_flush (self); + } +} + +/// Mark the request as done. Returns false if the underlying transport +/// should be closed, this being the last request. +static bool +fcgi_request_finish (struct fcgi_request *self) +{ + fcgi_request_flush (self); + fcgi_muxer_send (self->muxer, FCGI_STDOUT, self->request_id, NULL, 0); + + fcgi_muxer_send_end_request (self->muxer, self->request_id, + 0 /* TODO app_status, although ignored */, + FCGI_REQUEST_COMPLETE /* TODO protocol_status, may be different */); + + bool should_close = !(self->flags & FCGI_KEEP_CONN); + + self->muxer->active_requests--; + self->muxer->requests[self->request_id] = NULL; + fcgi_request_destroy (self); + + return !should_close; +} + +static bool +fcgi_request_push_params + (struct fcgi_request *self, const void *data, size_t len) +{ + if (self->state != FCGI_REQUEST_PARAMS) + { + print_debug ("FastCGI: expected %s, got %s", + STRINGIFY (FCGI_STDIN), STRINGIFY (FCGI_PARAMS)); + return false; + } + + if (len) + fcgi_nv_parser_push (&self->hdr_parser, data, len); + else + { + if (self->hdr_parser.state != FCGI_NV_PARSER_NAME_LEN) + print_debug ("FastCGI: request headers seem to be cut off"); + + self->state = FCGI_REQUEST_STDIN; + if (!self->muxer->request_start_cb (self)) + return fcgi_request_finish (self); + } + return true; +} + +static bool +fcgi_request_push_stdin + (struct fcgi_request *self, const void *data, size_t len) +{ + if (self->state != FCGI_REQUEST_STDIN) + { + print_debug ("FastCGI: expected %s, got %s", + STRINGIFY (FCGI_PARAMS), STRINGIFY (FCGI_STDIN)); + return false; + } + + return self->muxer->request_push_cb (self, data, len); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +typedef bool (*fcgi_muxer_handler_fn) + (struct fcgi_muxer *, const struct fcgi_parser *); + +static bool +fcgi_muxer_on_get_values + (struct fcgi_muxer *self, const struct fcgi_parser *parser) +{ + if (parser->request_id != FCGI_NULL_REQUEST_ID) + { + print_debug ("FastCGI: invalid %s message", + STRINGIFY (FCGI_GET_VALUES)); + return false; + } + + struct str_map values = str_map_make (free); + struct str_map response = str_map_make (free); + + struct fcgi_nv_parser nv_parser = fcgi_nv_parser_make (); + nv_parser.output = &values; + + fcgi_nv_parser_push (&nv_parser, parser->content.str, parser->content.len); + const char *key = NULL; + + // No real-world servers seem to actually use multiplexing + // or even issue this request, but we will implement it anyway + if (str_map_find (&values, (key = FCGI_MPXS_CONNS))) + str_map_set (&response, key, xstrdup ("1")); + + // It's not clear whether FCGI_MAX_REQS means concurrently over all + // connections or over just a single connection (multiplexed), though + // supposedly it's actually per /web server/. Supply the strictest limit. + if (str_map_find (&values, (key = FCGI_MAX_REQS))) + str_map_set (&response, key, + xstrdup_printf ("%zu", N_ELEMENTS (self->requests) - 1)); + + // FCGI_MAX_CONNS would be basically infinity. We don't limit connections. + + struct str content = str_make (); + fcgi_nv_convert (&response, &content); + fcgi_muxer_send (self, FCGI_GET_VALUES_RESULT, parser->request_id, + content.str, content.len); + str_free (&content); + + str_map_free (&values); + str_map_free (&response); + return true; +} + +static bool +fcgi_muxer_on_begin_request + (struct fcgi_muxer *self, const struct fcgi_parser *parser) +{ + struct msg_unpacker unpacker = + msg_unpacker_make (parser->content.str, parser->content.len); + + uint16_t role; + uint8_t flags; + bool success = true; + success &= msg_unpacker_u16 (&unpacker, &role); + success &= msg_unpacker_u8 (&unpacker, &flags); + // Ignoring 5 reserved bytes + + if (!success) + { + print_debug ("FastCGI: invalid %s message", + STRINGIFY (FCGI_BEGIN_REQUEST)); + return false; + } + + struct fcgi_request *request = self->requests[parser->request_id]; + if (parser->request_id == FCGI_NULL_REQUEST_ID || request) + { + print_debug ("FastCGI: unusable request ID in %s message", + STRINGIFY (FCGI_BEGIN_REQUEST)); + return false; + } + + // We can only act as a responder, reject everything else up front + if (role != FCGI_RESPONDER) + { + fcgi_muxer_send_end_request (self, + parser->request_id, 0, FCGI_UNKNOWN_ROLE); + return true; + } + + if (parser->request_id >= N_ELEMENTS (self->requests) + || self->in_shutdown) + { + fcgi_muxer_send_end_request (self, + parser->request_id, 0, FCGI_OVERLOADED); + return true; + } + + request = fcgi_request_new (); + request->muxer = self; + request->request_id = parser->request_id; + request->flags = flags; + + self->requests[parser->request_id] = request; + self->active_requests++; + return true; +} + +static bool +fcgi_muxer_on_abort_request + (struct fcgi_muxer *self, const struct fcgi_parser *parser) +{ + struct fcgi_request *request = self->requests[parser->request_id]; + if (parser->request_id == FCGI_NULL_REQUEST_ID || !request) + { + print_debug ("FastCGI: received %s for an unknown request", + STRINGIFY (FCGI_ABORT_REQUEST)); + return true; // We might have just rejected it + } + + return fcgi_request_finish (request); +} + +static bool +fcgi_muxer_on_params (struct fcgi_muxer *self, const struct fcgi_parser *parser) +{ + struct fcgi_request *request = self->requests[parser->request_id]; + if (parser->request_id == FCGI_NULL_REQUEST_ID || !request) + { + print_debug ("FastCGI: received %s for an unknown request", + STRINGIFY (FCGI_PARAMS)); + return true; // We might have just rejected it + } + + // This may immediately finish and delete the request, but that's fine + return fcgi_request_push_params (request, + parser->content.str, parser->content.len); +} + +static bool +fcgi_muxer_on_stdin (struct fcgi_muxer *self, const struct fcgi_parser *parser) +{ + struct fcgi_request *request = self->requests[parser->request_id]; + if (parser->request_id == FCGI_NULL_REQUEST_ID || !request) + { + print_debug ("FastCGI: received %s for an unknown request", + STRINGIFY (FCGI_STDIN)); + return true; // We might have just rejected it + } + + // At the end of the stream, a zero-length record is received + return fcgi_request_push_stdin (request, + parser->content.str, parser->content.len); +} + +static bool +fcgi_muxer_on_message (const struct fcgi_parser *parser, void *user_data) +{ + struct fcgi_muxer *self = user_data; + + if (parser->version != FCGI_VERSION_1) + { + print_debug ("FastCGI: unsupported version %d", parser->version); + return false; + } + + static const fcgi_muxer_handler_fn handlers[] = + { + [FCGI_GET_VALUES] = fcgi_muxer_on_get_values, + [FCGI_BEGIN_REQUEST] = fcgi_muxer_on_begin_request, + [FCGI_ABORT_REQUEST] = fcgi_muxer_on_abort_request, + [FCGI_PARAMS] = fcgi_muxer_on_params, + [FCGI_STDIN] = fcgi_muxer_on_stdin, + }; + + fcgi_muxer_handler_fn handler; + if (parser->type >= N_ELEMENTS (handlers) + || !(handler = handlers[parser->type])) + { + // Responding in this way even to application records, unspecified + uint8_t content[8] = { parser->type }; + fcgi_muxer_send (self, FCGI_UNKNOWN_TYPE, parser->request_id, + content, sizeof content); + return true; + } + + return handler (self, parser); +} + +static void +fcgi_muxer_init (struct fcgi_muxer *self) +{ + self->parser = fcgi_parser_make (); + self->parser.on_message = fcgi_muxer_on_message; + self->parser.user_data = self; +} + +static void +fcgi_muxer_free (struct fcgi_muxer *self) +{ + for (size_t i = 0; i < N_ELEMENTS (self->requests); i++) + { + if (!self->active_requests) + break; + if (self->requests[i]) + { + fcgi_request_destroy (self->requests[i]); + self->active_requests--; + } + } + + fcgi_parser_free (&self->parser); +} + +static bool +fcgi_muxer_push (struct fcgi_muxer *self, const void *data, size_t len) +{ + return fcgi_parser_push (&self->parser, data, len); +} + +/// @} +// --- WebSockets -------------------------------------------------------------- +/// @defgroup WebSockets +/// @{ + +// WebSockets aren't CGI-compatible, therefore we must handle the initial HTTP +// handshake ourselves. Luckily it's not too much of a bother with http-parser. +// Typically there will be a normal HTTP server in front of us, proxying the +// requests based on the URI. + +enum ws_handler_state +{ + WS_HANDLER_CONNECTING, ///< Parsing HTTP + WS_HANDLER_OPEN, ///< Parsing WebSockets frames + WS_HANDLER_CLOSING, ///< Partial closure by us + WS_HANDLER_FLUSHING, ///< Just waiting for client EOF + WS_HANDLER_CLOSED ///< Dead, both sides closed +}; + +struct ws_handler +{ + enum ws_handler_state state; ///< State + + // HTTP handshake: + + http_parser hp; ///< HTTP parser + bool have_header_value; ///< Parsing header value or field? + struct str field; ///< Field part buffer + struct str value; ///< Value part buffer + struct str_map headers; ///< HTTP Headers + struct str url; ///< Request URL + ev_timer handshake_timeout_watcher; ///< Handshake timeout watcher + + // WebSocket frame protocol: + + struct ws_parser parser; ///< Protocol frame parser + bool expecting_continuation; ///< For non-control traffic + + enum ws_opcode message_opcode; ///< Opcode for the current message + struct str message_data; ///< Concatenated message data + + ev_timer ping_timer; ///< Ping timer + bool received_pong; ///< Received PONG since the last PING + + ev_timer close_timeout_watcher; ///< Close timeout watcher + + // Configuration: + + unsigned handshake_timeout; ///< How long to wait for the handshake + unsigned close_timeout; ///< How long to wait for TCP close + unsigned ping_interval; ///< Ping interval in seconds + uint64_t max_payload_len; ///< Maximum length of any message + + // Event callbacks: + + // TODO: void (*on_handshake) (protocols) that will allow the user + // to choose any sub-protocol, if the client has provided any. + // This may render "on_connected" unnecessary. + // Should also enable failing the handshake. + + /// Called after successfuly connecting (handshake complete) + bool (*on_connected) (struct ws_handler *); + + /// Called upon reception of a single full message + bool (*on_message) (struct ws_handler *, + enum ws_opcode type, const void *data, size_t len); + + /// The connection is about to close. @a close_code may, or may not, be one + /// of enum ws_status. The @a reason is never NULL. + void (*on_close) (struct ws_handler *, int close_code, const char *reason); + + // Virtual method callbacks: + + /// Write a chunk of data to the stream + void (*write_cb) (struct ws_handler *, const void *data, size_t len); + + /// Close the connection. If @a half_close is false, you are allowed to + /// destroy the handler directly from within the callback. + void (*close_cb) (struct ws_handler *, bool half_close); +}; + +static void +ws_handler_send_control (struct ws_handler *self, + enum ws_opcode opcode, const void *data, size_t len) +{ + if (len > WS_MAX_CONTROL_PAYLOAD_LEN) + { + print_debug ("truncating output control frame payload" + " from %zu to %zu bytes", len, (size_t) WS_MAX_CONTROL_PAYLOAD_LEN); + len = WS_MAX_CONTROL_PAYLOAD_LEN; + } + + uint8_t header[2] = { 0x80 | (opcode & 0x0F), len }; + self->write_cb (self, header, sizeof header); + self->write_cb (self, data, len); +} + +static void +ws_handler_close (struct ws_handler *self, + enum ws_status close_code, const char *reason, size_t len) +{ + hard_assert (self->state == WS_HANDLER_OPEN); + + struct str payload = str_make (); + str_pack_u16 (&payload, close_code); + // XXX: maybe accept a null-terminated string on input? Has to be UTF-8 a/w + str_append_data (&payload, reason, len); + ws_handler_send_control (self, WS_OPCODE_CLOSE, payload.str, payload.len); + self->close_cb (self, true /* half_close */); + + self->state = WS_HANDLER_CLOSING; + str_free (&payload); +} + +static bool +ws_handler_fail_connection (struct ws_handler *self, enum ws_status close_code) +{ + hard_assert (self->state == WS_HANDLER_OPEN + || self->state == WS_HANDLER_CLOSING); + + if (self->state == WS_HANDLER_OPEN) + ws_handler_close (self, close_code, NULL, 0); + + self->state = WS_HANDLER_FLUSHING; + if (self->on_close) + self->on_close (self, WS_STATUS_ABNORMAL_CLOSURE, ""); + + ev_timer_stop (EV_DEFAULT_ &self->ping_timer); + ev_timer_set (&self->close_timeout_watcher, self->close_timeout, 0.); + ev_timer_start (EV_DEFAULT_ &self->close_timeout_watcher); + return false; +} + +// TODO: add support for fragmented responses +static void +ws_handler_send_frame (struct ws_handler *self, + enum ws_opcode opcode, const void *data, size_t len) +{ + if (!soft_assert (self->state == WS_HANDLER_OPEN)) + return; + + struct str header = str_make (); + str_pack_u8 (&header, 0x80 | (opcode & 0x0F)); + + if (len > UINT16_MAX) + { + str_pack_u8 (&header, 127); + str_pack_u64 (&header, len); + } + else if (len > 125) + { + str_pack_u8 (&header, 126); + str_pack_u16 (&header, len); + } + else + str_pack_u8 (&header, len); + + self->write_cb (self, header.str, header.len); + self->write_cb (self, data, len); + str_free (&header); +} + +static bool +ws_handler_on_frame_header (void *user_data, const struct ws_parser *parser) +{ + struct ws_handler *self = user_data; + + // Note that we aren't expected to send any close frame before closing the + // connection when the frame is unmasked + + if (parser->reserved_1 || parser->reserved_2 || parser->reserved_3 + || !parser->is_masked // client -> server payload must be masked + || (ws_is_control_frame (parser->opcode) && + (!parser->is_fin || parser->payload_len > WS_MAX_CONTROL_PAYLOAD_LEN)) + || (!ws_is_control_frame (parser->opcode) && + (self->expecting_continuation && parser->opcode != WS_OPCODE_CONT)) + || parser->payload_len >= 0x8000000000000000ULL) + return ws_handler_fail_connection (self, WS_STATUS_PROTOCOL_ERROR); + + if (parser->payload_len > self->max_payload_len + || (self->expecting_continuation && + self->message_data.len + parser->payload_len > self->max_payload_len)) + return ws_handler_fail_connection (self, WS_STATUS_MESSAGE_TOO_BIG); + return true; +} + +static bool +ws_handler_on_control_close + (struct ws_handler *self, const struct ws_parser *parser) +{ + hard_assert (self->state == WS_HANDLER_OPEN + || self->state == WS_HANDLER_CLOSING); + struct msg_unpacker unpacker = + msg_unpacker_make (parser->input.str, parser->payload_len); + + char *reason = NULL; + uint16_t close_code = WS_STATUS_NO_STATUS_RECEIVED; + if (parser->payload_len >= 2) + { + (void) msg_unpacker_u16 (&unpacker, &close_code); + reason = xstrndup (parser->input.str + 2, parser->payload_len - 2); + } + else + reason = xstrdup (""); + + if (close_code < 1000 || close_code > 4999) + // XXX: invalid close code: maybe we should fail the connection instead + close_code = WS_STATUS_PROTOCOL_ERROR; + + if (self->state == WS_HANDLER_OPEN) + { + // Close initiated by the client + // FIXME: not sending the potentially different close_code + ws_handler_send_control (self, WS_OPCODE_CLOSE, + parser->input.str, parser->payload_len); + + self->state = WS_HANDLER_FLUSHING; + if (self->on_close) + self->on_close (self, close_code, reason); + } + else + self->state = WS_HANDLER_FLUSHING; + + free (reason); + + ev_timer_stop (EV_DEFAULT_ &self->ping_timer); + ev_timer_set (&self->close_timeout_watcher, self->close_timeout, 0.); + ev_timer_start (EV_DEFAULT_ &self->close_timeout_watcher); + return true; +} + +static bool +ws_handler_on_control_frame + (struct ws_handler *self, const struct ws_parser *parser) +{ + switch (parser->opcode) + { + case WS_OPCODE_CLOSE: + return ws_handler_on_control_close (self, parser); + case WS_OPCODE_PING: + ws_handler_send_control (self, WS_OPCODE_PONG, + parser->input.str, parser->payload_len); + break; + case WS_OPCODE_PONG: + // TODO: check the payload + self->received_pong = true; + break; + default: + // Unknown control frame + return ws_handler_fail_connection (self, WS_STATUS_PROTOCOL_ERROR); + } + return true; +} + +static bool +ws_handler_on_frame (void *user_data, const struct ws_parser *parser) +{ + struct ws_handler *self = user_data; + if (ws_is_control_frame (parser->opcode)) + return ws_handler_on_control_frame (self, parser); + if (!self->expecting_continuation) + self->message_opcode = parser->opcode; + + str_append_data (&self->message_data, + parser->input.str, parser->payload_len); + if ((self->expecting_continuation = !parser->is_fin)) + return true; + + if (self->message_opcode == WS_OPCODE_TEXT + && !utf8_validate (self->message_data.str, self->message_data.len)) + { + return ws_handler_fail_connection + (self, WS_STATUS_INVALID_PAYLOAD_DATA); + } + + bool result = true; + if (self->on_message) + result = self->on_message (self, self->message_opcode, + self->message_data.str, self->message_data.len); + str_reset (&self->message_data); + // TODO: if (!result), either replace this with a state check, + // or make sure to change the state + return result; +} + +static void +ws_handler_on_ping_timer (EV_P_ ev_timer *watcher, int revents) +{ + (void) loop; + (void) revents; + + struct ws_handler *self = watcher->data; + if (!self->received_pong) + ws_handler_fail_connection (self, 4000); + else + { + // TODO: be an annoying server and send a nonce in the data + ws_handler_send_control (self, WS_OPCODE_PING, NULL, 0); + ev_timer_again (EV_A_ watcher); + } +} + +static void +ws_handler_on_close_timeout (EV_P_ ev_timer *watcher, int revents) +{ + (void) loop; + (void) revents; + struct ws_handler *self = watcher->data; + + hard_assert (self->state == WS_HANDLER_OPEN + || self->state == WS_HANDLER_CLOSING); + + if (self->state == WS_HANDLER_CLOSING + && self->on_close) + self->on_close (self, WS_STATUS_ABNORMAL_CLOSURE, "close timeout"); + + self->state = WS_HANDLER_CLOSED; + self->close_cb (self, false /* half_close */); +} + +static void +ws_handler_on_handshake_timeout (EV_P_ ev_timer *watcher, int revents) +{ + (void) loop; + (void) revents; + struct ws_handler *self = watcher->data; + + // XXX: this is a no-op, since this currently doesn't even call shutdown + // immediately but postpones it until later + self->close_cb (self, true /* half_close */); + self->state = WS_HANDLER_FLUSHING; + + if (self->on_close) + self->on_close (self, WS_STATUS_ABNORMAL_CLOSURE, "handshake timeout"); + + self->state = WS_HANDLER_CLOSED; + self->close_cb (self, false /* half_close */); +} + +static void +ws_handler_init (struct ws_handler *self) +{ + memset (self, 0, sizeof *self); + + self->state = WS_HANDLER_CONNECTING; + + http_parser_init (&self->hp, HTTP_REQUEST); + self->hp.data = self; + self->field = str_make (); + self->value = str_make (); + self->headers = str_map_make (free); + self->headers.key_xfrm = tolower_ascii_strxfrm; + self->url = str_make (); + ev_timer_init (&self->handshake_timeout_watcher, + ws_handler_on_handshake_timeout, 0., 0.); + self->handshake_timeout_watcher.data = self; + + self->parser = ws_parser_make (); + self->parser.on_frame_header = ws_handler_on_frame_header; + self->parser.on_frame = ws_handler_on_frame; + self->parser.user_data = self; + self->message_data = str_make (); + + ev_timer_init (&self->ping_timer, + ws_handler_on_ping_timer, 0., 0.); + self->ping_timer.data = self; + ev_timer_init (&self->close_timeout_watcher, + ws_handler_on_close_timeout, 0., 0.); + self->ping_timer.data = self; + // So that the first ping timer doesn't timeout the connection + self->received_pong = true; + + self->handshake_timeout = self->close_timeout = self->ping_interval = 60; + // This is still ridiculously high. Note that the most significant bit + // must always be zero, i.e. the protocol maximum is 0x7FFF FFFF FFFF FFFF. + self->max_payload_len = UINT32_MAX; +} + +/// Stop all timers, not going to use the handler anymore +static void +ws_handler_stop (struct ws_handler *self) +{ + ev_timer_stop (EV_DEFAULT_ &self->handshake_timeout_watcher); + ev_timer_stop (EV_DEFAULT_ &self->ping_timer); + ev_timer_stop (EV_DEFAULT_ &self->close_timeout_watcher); +} + +static void +ws_handler_free (struct ws_handler *self) +{ + ws_handler_stop (self); + + str_free (&self->field); + str_free (&self->value); + str_map_free (&self->headers); + str_free (&self->url); + + ws_parser_free (&self->parser); + str_free (&self->message_data); +} + +static bool +ws_handler_header_field_is_a_list (const char *name) +{ + // This must contain all header fields we use for anything + static const char *concatenable[] = + { SEC_WS_PROTOCOL, SEC_WS_EXTENSIONS, "Connection", "Upgrade" }; + + for (size_t i = 0; i < N_ELEMENTS (concatenable); i++) + if (!strcasecmp_ascii (name, concatenable[i])) + return true; + return false; +} + +static void +ws_handler_on_header_read (struct ws_handler *self) +{ + // The HTTP parser unfolds values and removes preceding whitespace, but + // otherwise doesn't touch the values or the following whitespace. + + // RFC 7230 states that trailing whitespace is not part of a field value + char *value = self->field.str; + size_t len = self->field.len; + while (len--) + if (value[len] == '\t' || value[len] == ' ') + value[len] = '\0'; + else + break; + self->field.len = len; + + const char *field = self->field.str; + const char *current = str_map_find (&self->headers, field); + if (ws_handler_header_field_is_a_list (field) && current) + str_map_set (&self->headers, field, + xstrdup_printf ("%s, %s", current, self->value.str)); + else + // If the field cannot be concatenated, just overwrite the last value. + // Maybe we should issue a warning or something. + str_map_set (&self->headers, field, xstrdup (self->value.str)); +} + +static int +ws_handler_on_header_field (http_parser *parser, const char *at, size_t len) +{ + struct ws_handler *self = parser->data; + if (self->have_header_value) + { + ws_handler_on_header_read (self); + str_reset (&self->field); + str_reset (&self->value); + } + str_append_data (&self->field, at, len); + self->have_header_value = false; + return 0; +} + +static int +ws_handler_on_header_value (http_parser *parser, const char *at, size_t len) +{ + struct ws_handler *self = parser->data; + str_append_data (&self->value, at, len); + self->have_header_value = true; + return 0; +} + +static int +ws_handler_on_headers_complete (http_parser *parser) +{ + struct ws_handler *self = parser->data; + if (self->have_header_value) + ws_handler_on_header_read (self); + + // We strictly require a protocol upgrade + if (!parser->upgrade) + return 2; + + return 0; +} + +static int +ws_handler_on_url (http_parser *parser, const char *at, size_t len) +{ + struct ws_handler *self = parser->data; + str_append_data (&self->value, at, len); + return 0; +} + +#define HTTP_101_SWITCHING_PROTOCOLS "101 Switching Protocols" +#define HTTP_400_BAD_REQUEST "400 Bad Request" +#define HTTP_405_METHOD_NOT_ALLOWED "405 Method Not Allowed" +#define HTTP_417_EXPECTATION_FAILED "407 Expectation Failed" +#define HTTP_426_UPGRADE_REQUIRED "426 Upgrade Required" +#define HTTP_505_VERSION_NOT_SUPPORTED "505 HTTP Version Not Supported" + +static void +ws_handler_http_responsev (struct ws_handler *self, + const char *status, char *const *fields) +{ + hard_assert (status != NULL); + + struct str response = str_make (); + str_append_printf (&response, "HTTP/1.1 %s\r\n", status); + + while (*fields) + str_append_printf (&response, "%s\r\n", *fields++); + + time_t now = time (NULL); + struct tm ts; + gmtime_r (&now, &ts); + + // See RFC 7231, 7.1.1.2. Date + const char *dow[] = { "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat" }; + const char *moy[] = { "Jan", "Feb", "Mar", "Apr", "May", "Jun", + "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" }; + str_append_printf (&response, + "Date: %s, %02d %s %04d %02d:%02d:%02d GMT\r\n", + dow[ts.tm_wday], ts.tm_mday, moy[ts.tm_mon], ts.tm_year + 1900, + ts.tm_hour, ts.tm_min, ts.tm_sec); + + str_append (&response, "Server: " + PROGRAM_NAME "/" PROGRAM_VERSION "\r\n\r\n"); + self->write_cb (self, response.str, response.len); + str_free (&response); +} + +static bool +ws_handler_fail_handshake (struct ws_handler *self, const char *status, ...) +{ + va_list ap; + va_start (ap, status); + + const char *s; + struct strv v = strv_make (); + while ((s = va_arg (ap, const char *))) + strv_append (&v, s); + + va_end (ap); + ws_handler_http_responsev (self, status, v.vector); + strv_free (&v); + + self->close_cb (self, true /* half_close */); + self->state = WS_HANDLER_FLUSHING; + + if (self->on_close) + self->on_close (self, WS_STATUS_ABNORMAL_CLOSURE, status); + return false; +} + +#define FAIL_HANDSHAKE(...) \ + return ws_handler_fail_handshake (self, __VA_ARGS__, NULL) + +static bool +ws_handler_finish_handshake (struct ws_handler *self) +{ + if (self->hp.method != HTTP_GET) + FAIL_HANDSHAKE (HTTP_405_METHOD_NOT_ALLOWED, "Allow: GET"); + + // Technically, it must be /at least/ 1.1 but no other 1.x version of HTTP + // is going to happen and 2.x is entirely incompatible + // XXX: we probably shouldn't use 505 to reject the minor version but w/e + if (self->hp.http_major != 1 || self->hp.http_minor != 1) + FAIL_HANDSHAKE (HTTP_505_VERSION_NOT_SUPPORTED); + + // Your expectations are way too high + if (str_map_find (&self->headers, "Expect")) + FAIL_HANDSHAKE (HTTP_417_EXPECTATION_FAILED); + + // Reject URLs specifying the schema and host; we're not parsing that + // TODO: actually do parse this and let our user decide if it matches + struct http_parser_url url; + if (http_parser_parse_url (self->url.str, self->url.len, false, &url) + || (url.field_set & (1 << UF_SCHEMA | 1 << UF_HOST | 1 << UF_PORT)) + || !str_map_find (&self->headers, "Host")) + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST); + + const char *connection = str_map_find (&self->headers, "Connection"); + if (!connection || strcasecmp_ascii (connection, "Upgrade")) + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST); + + // Check if we can actually upgrade the protocol to WebSockets + const char *upgrade = str_map_find (&self->headers, "Upgrade"); + struct http_protocol *offered_upgrades = NULL; + bool can_upgrade = false; + if (upgrade && http_parse_upgrade (upgrade, &offered_upgrades)) + // Case-insensitive according to RFC 6455; neither RFC 2616 nor 7230 + // say anything at all about case-sensitivity for this field + LIST_FOR_EACH (struct http_protocol, iter, offered_upgrades) + { + if (!iter->version && !strcasecmp_ascii (iter->name, "websocket")) + can_upgrade = true; + http_protocol_destroy (iter); + } + if (!can_upgrade) + FAIL_HANDSHAKE (HTTP_426_UPGRADE_REQUIRED, + "Upgrade: websocket", SEC_WS_VERSION ": 13"); + + // Okay, we're finally past the basic HTTP/1.1 stuff + const char *key = str_map_find (&self->headers, SEC_WS_KEY); + const char *version = str_map_find (&self->headers, SEC_WS_VERSION); +/* + const char *protocol = str_map_find (&self->headers, SEC_WS_PROTOCOL); + const char *extensions = str_map_find (&self->headers, SEC_WS_EXTENSIONS); +*/ + + if (!version) + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST); + if (strcmp (version, "13")) + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST, SEC_WS_VERSION ": 13"); + + struct str tmp = str_make (); + bool key_is_valid = key + && base64_decode (key, false, &tmp) && tmp.len == 16; + str_free (&tmp); + if (!key_is_valid) + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST); + + struct strv fields = strv_make (); + strv_append_args (&fields, + "Upgrade: websocket", + "Connection: Upgrade", + NULL); + + char *response_key = ws_encode_response_key (key); + strv_append_owned (&fields, + xstrdup_printf (SEC_WS_ACCEPT ": %s", response_key)); + free (response_key); + + // TODO: make it possible to choose Sec-Websocket-{Extensions,Protocol} + + ws_handler_http_responsev (self, + HTTP_101_SWITCHING_PROTOCOLS, fields.vector); + + strv_free (&fields); + + self->state = WS_HANDLER_OPEN; + ev_timer_init (&self->ping_timer, ws_handler_on_ping_timer, + self->ping_interval, 0); + ev_timer_start (EV_DEFAULT_ &self->ping_timer); + return true; +} + +/// Tells the handler that the TCP connection has been established so it can +/// timeout when the client handshake doesn't arrive soon enough +static void +ws_handler_start (struct ws_handler *self) +{ + hard_assert (self->state == WS_HANDLER_CONNECTING); + + ev_timer_set (&self->handshake_timeout_watcher, + self->handshake_timeout, 0.); + ev_timer_start (EV_DEFAULT_ &self->handshake_timeout_watcher); +} + +// The client should normally never close the connection, assume that it's +// either received an EOF from our side, or that it doesn't care about our data +// anymore, having called close() already +static bool +ws_handler_push_eof (struct ws_handler *self) +{ + switch (self->state) + { + case WS_HANDLER_CONNECTING: + ev_timer_stop (EV_DEFAULT_ &self->handshake_timeout_watcher); + + self->state = WS_HANDLER_FLUSHING; + if (self->on_close) + self->on_close (self, WS_STATUS_ABNORMAL_CLOSURE, "unexpected EOF"); + break; + case WS_HANDLER_OPEN: + ev_timer_stop (EV_DEFAULT_ &self->ping_timer); + // Fall-through + case WS_HANDLER_CLOSING: + self->state = WS_HANDLER_CLOSED; + if (self->on_close) + self->on_close (self, WS_STATUS_ABNORMAL_CLOSURE, ""); + // Fall-through + case WS_HANDLER_FLUSHING: + ev_timer_stop (EV_DEFAULT_ &self->close_timeout_watcher); + break; + default: + soft_assert(self->state != WS_HANDLER_CLOSED); + } + self->state = WS_HANDLER_CLOSED; + return false; +} + +/// Push data to the WebSocket handler. "len == 0" means EOF. +/// You are expected to close the connection and dispose of the handler +/// when the function returns false. +static bool +ws_handler_push (struct ws_handler *self, const void *data, size_t len) +{ + if (!len) + return ws_handler_push_eof (self); + + if (self->state == WS_HANDLER_FLUSHING) + // We're waiting for an EOF from the client, must not process data + return true; + + if (self->state != WS_HANDLER_CONNECTING) + return soft_assert (self->state != WS_HANDLER_CLOSED) + && ws_parser_push (&self->parser, data, len); + + // The handshake hasn't been done yet, process HTTP headers + static const http_parser_settings http_settings = + { + .on_header_field = ws_handler_on_header_field, + .on_header_value = ws_handler_on_header_value, + .on_headers_complete = ws_handler_on_headers_complete, + .on_url = ws_handler_on_url, + }; + + size_t n_parsed = + http_parser_execute (&self->hp, &http_settings, data, len); + + if (self->hp.upgrade) + { + ev_timer_stop (EV_DEFAULT_ &self->handshake_timeout_watcher); + + // The handshake hasn't been finished, yet there is more data + // to be processed after the headers already + if (len - n_parsed) + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST); + + if (!ws_handler_finish_handshake (self)) + return false; + if (self->on_connected) + return self->on_connected (self); + return true; + } + + enum http_errno err = HTTP_PARSER_ERRNO (&self->hp); + if (n_parsed != len || err != HPE_OK) + { + ev_timer_stop (EV_DEFAULT_ &self->handshake_timeout_watcher); + + if (err == HPE_CB_headers_complete) + print_debug ("WS handshake failed: %s", "missing `Upgrade' field"); + else + print_debug ("WS handshake failed: %s", + http_errno_description (err)); + + FAIL_HANDSHAKE (HTTP_400_BAD_REQUEST); + } + return true; +} + +/// @} +// --- Server ------------------------------------------------------------------ + +static struct simple_config_item g_config_table[] = +{ + { "bind_host", NULL, "Address of the server" }, + { "port_fastcgi", "9000", "Port to bind for FastCGI" }, + { "port_scgi", NULL, "Port to bind for SCGI" }, + { "port_ws", NULL, "Port to bind for WebSockets" }, + { "pid_file", NULL, "Full path for the PID file" }, + // XXX: here belongs something like a web SPA that interfaces with us + { "static_root", NULL, "The root for static content" }, + { NULL, NULL, NULL } +}; + +struct server_context +{ + ev_signal sigterm_watcher; ///< Got SIGTERM + ev_signal sigint_watcher; ///< Got SIGINT + ev_timer quit_timeout_watcher; ///< Quit timeout watcher + bool quitting; ///< User requested quitting + + struct listener *listeners; ///< Listeners + size_t n_listeners; ///< Number of listening sockets + + struct client *clients; ///< Clients + unsigned n_clients; ///< Current number of connections + + struct request_handler *handlers; ///< Request handlers + struct str_map config; ///< Server configuration +}; + +static void initiate_quit (struct server_context *self); +static void try_finish_quit (struct server_context *self); +static void on_quit_timeout (EV_P_ ev_timer *watcher, int revents); +static void close_listeners (struct server_context *self); + +static void +server_context_init (struct server_context *self) +{ + memset (self, 0, sizeof *self); + + self->config = str_map_make (NULL); + simple_config_load_defaults (&self->config, g_config_table); + ev_timer_init (&self->quit_timeout_watcher, on_quit_timeout, 3., 0.); + self->quit_timeout_watcher.data = self; +} + +static void +server_context_free (struct server_context *self) +{ + // We really shouldn't attempt a quit without closing the clients first + soft_assert (!self->clients); + + close_listeners (self); + free (self->listeners); + + str_map_free (&self->config); +} + +// --- JSON-RPC ---------------------------------------------------------------- +/// @defgroup JSON-RPC +/// @{ + +#define JSON_RPC_ERROR_TABLE(XX) \ + XX (-32700, PARSE_ERROR, "Parse error") \ + XX (-32600, INVALID_REQUEST, "Invalid Request") \ + XX (-32601, METHOD_NOT_FOUND, "Method not found") \ + XX (-32602, INVALID_PARAMS, "Invalid params") \ + XX (-32603, INTERNAL_ERROR, "Internal error") + +enum json_rpc_error +{ +#define XX(code, name, message) JSON_RPC_ERROR_ ## name, + JSON_RPC_ERROR_TABLE (XX) +#undef XX + JSON_RPC_ERROR_COUNT +}; + +static json_t * +json_rpc_error (enum json_rpc_error id, json_t *data) +{ +#define XX(code, name, message) { code, message }, + static const struct json_rpc_error + { + int code; + const char *message; + } + errors[JSON_RPC_ERROR_COUNT] = + { + JSON_RPC_ERROR_TABLE (XX) + }; +#undef XX + + json_t *error = json_object (); + json_object_set_new (error, "code", json_integer (errors[id].code)); + json_object_set_new (error, "message", json_string (errors[id].message)); + + if (data) + json_object_set_new (error, "data", data); + + return error; +} + +static json_t * +json_rpc_response (json_t *id, json_t *result, json_t *error) +{ + json_t *x = json_object (); + json_object_set_new (x, "jsonrpc", json_string ("2.0")); + json_object_set_new (x, "id", id ? id : json_null ()); + if (result) json_object_set_new (x, "result", result); + if (error) json_object_set_new (x, "error", error); + return x; +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static bool +validate_json_rpc_content_type (const char *content_type) +{ + char *type = NULL; + char *subtype = NULL; + + struct str_map parameters = str_map_make (free); + parameters.key_xfrm = tolower_ascii_strxfrm; + + bool result = http_parse_media_type + (content_type, &type, &subtype, ¶meters); + if (!result) + goto end; + + if (strcasecmp_ascii (type, "application") + || (strcasecmp_ascii (subtype, "json") && + strcasecmp_ascii (subtype, "json-rpc" /* obsolete */))) + result = false; + + const char *charset = str_map_find (¶meters, "charset"); + if (charset && strcasecmp_ascii (charset, "UTF-8")) + result = false; + + // Currently ignoring all unknown parametrs + +end: + free (type); + free (subtype); + str_map_free (¶meters); + return result; +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +typedef json_t *(*json_rpc_handler_fn) (struct server_context *, json_t *); + +struct json_rpc_handler_info +{ + const char *method_name; ///< JSON-RPC method name + json_rpc_handler_fn handler; ///< Method handler +}; + +static int +json_rpc_handler_info_cmp (const void *first, const void *second) +{ + return strcmp (((struct json_rpc_handler_info *) first)->method_name, + ((struct json_rpc_handler_info *) second)->method_name); +} + +// TODO: a method that sends a response after a certain number of seconds. +// This has to be owned by the server context as a background job that +// removes itself upon completion. + +static json_t * +json_rpc_ping (struct server_context *ctx, json_t *params) +{ + (void) ctx; + (void) params; + + return json_rpc_response (NULL, json_string ("pong"), NULL); +} + +static json_t * +process_json_rpc_request (struct server_context *ctx, json_t *request) +{ + // A list of all available methods; this list has to be ordered. + // Eventually it might be better to move this into a map in the context. + static struct json_rpc_handler_info handlers[] = + { + { "ping", json_rpc_ping }, + }; + + if (!json_is_object (request)) + return json_rpc_response (NULL, NULL, + json_rpc_error (JSON_RPC_ERROR_INVALID_REQUEST, NULL)); + + json_t *v = json_object_get (request, "jsonrpc"); + json_t *m = json_object_get (request, "method"); + json_t *params = json_object_get (request, "params"); + json_t *id = json_object_get (request, "id"); + + const char *version; + const char *method; + + bool ok = true; + ok &= v && (version = json_string_value (v)) && !strcmp (version, "2.0"); + ok &= m && (method = json_string_value (m)); + ok &= !params || json_is_array (params) || json_is_object (params); + ok &= !id || json_is_null (id) || + json_is_string (id) || json_is_number (id); + if (!ok) + return json_rpc_response (id, NULL, + json_rpc_error (JSON_RPC_ERROR_INVALID_REQUEST, NULL)); + + struct json_rpc_handler_info key = { .method_name = method }; + struct json_rpc_handler_info *handler = bsearch (&key, handlers, + N_ELEMENTS (handlers), sizeof key, json_rpc_handler_info_cmp); + if (!handler) + return json_rpc_response (id, NULL, + json_rpc_error (JSON_RPC_ERROR_METHOD_NOT_FOUND, NULL)); + + json_t *response = handler->handler (ctx, params); + if (id) + return response; + + // Notifications don't get responses + json_decref (response); + return NULL; +} + +static void +flush_json (json_t *json, struct str *output) +{ + char *utf8 = json_dumps (json, JSON_ENCODE_ANY); + str_append (output, utf8); + free (utf8); + json_decref (json); +} + +static void +process_json_rpc (struct server_context *ctx, + const void *data, size_t len, struct str *output) +{ + + json_error_t e; + json_t *request; + if (!(request = json_loadb (data, len, JSON_DECODE_ANY, &e))) + { + flush_json (json_rpc_response (NULL, NULL, + json_rpc_error (JSON_RPC_ERROR_PARSE_ERROR, NULL)), + output); + return; + } + + if (json_is_array (request)) + { + if (!json_array_size (request)) + { + flush_json (json_rpc_response (NULL, NULL, + json_rpc_error (JSON_RPC_ERROR_INVALID_REQUEST, NULL)), + output); + return; + } + + json_t *response = json_array (); + json_t *iter; + size_t i; + + json_array_foreach (request, i, iter) + { + json_t *result = process_json_rpc_request (ctx, iter); + if (result) + json_array_append_new (response, result); + } + + if (json_array_size (response)) + flush_json (response, output); + else + json_decref (response); + } + else + { + json_t *result = process_json_rpc_request (ctx, request); + if (result) + flush_json (result, output); + } +} + +/// @} +// --- Requests ---------------------------------------------------------------- +/// @defgroup Requests +/// @{ + +/// A generic CGI request abstraction, writing data indirectly through callbacks +struct request +{ + struct server_context *ctx; ///< Server context + + struct request_handler *handler; ///< Assigned request handler + void *handler_data; ///< User data for the handler + + /// Callback to write some CGI response data to the output + void (*write_cb) (struct request *, const void *data, size_t len); + + /// Callback to close the CGI response, simulates end of program execution. + /// CALLING THIS MAY CAUSE THE REQUEST TO BE DESTROYED. + void (*close_cb) (struct request *); +}; + +/// An interface to detect and handle specific kinds of CGI requests. +/// The server walks through a list of them until it finds one that can serve +/// a particular request. If unsuccessful, the remote client gets a 404 +/// (the default handling). +struct request_handler +{ + LIST_HEADER (struct request_handler) + + /// Install ourselves as the handler for the request, if applicable. + /// Sets @a continue_ to false if further processing should be stopped, + /// meaning the request has already been handled. + bool (*try_handle) (struct request *request, + struct str_map *headers, bool *continue_); + + /// Handle incoming data. "len == 0" means EOF. + /// Returns false if there is no more processing to be done. + // FIXME: the EOF may or may not be delivered when request is cut short, + // we should fix FastCGI not to deliver it on CONTENT_LENGTH mismatch + bool (*push_cb) (struct request *request, const void *data, size_t len); + + /// Destroy the handler's data stored in the request object + void (*finalize_cb) (struct request *request); +}; + +static void +request_init (struct request *self) +{ + memset (self, 0, sizeof *self); +} + +static void +request_free (struct request *self) +{ + if (self->handler) + self->handler->finalize_cb (self); +} + +/// Write request CGI response data, intended for use by request handlers +static void +request_write (struct request *self, const void *data, size_t len) +{ + self->write_cb (self, data, len); +} + +/// This function is only intended to be run from asynchronous event handlers +/// such as timers, not as a direct result of starting the request or receiving +/// request data. CALLING THIS MAY CAUSE THE REQUEST TO BE DESTROYED. +static void +request_finish (struct request *self) +{ + self->close_cb (self); +} + +/// Starts processing a request. Returns false if no further action is to be +/// done and the request should be finished. +static bool +request_start (struct request *self, struct str_map *headers) +{ + // XXX: it feels like this should rather be two steps: + // bool (*can_handle) (request *, headers) + // ... install the handler ... + // bool (*handle) (request *) + // + // However that might cause some stuff to be done twice. + // + // Another way we could get rid of the continue_ argument is via adding + // some way of marking the request as finished from within the handler. + + if (g_debug_mode) + { + struct str_map_iter iter = str_map_iter_make (headers); + const char *value; + while ((value = str_map_iter_next (&iter))) + print_debug ("%s: %s", iter.link->key, value); + print_debug ("--"); + } + + bool continue_ = true; + LIST_FOR_EACH (struct request_handler, handler, self->ctx->handlers) + if (handler->try_handle (self, headers, &continue_)) + { + self->handler = handler; + return continue_; + } + + // Unable to serve the request + struct str response = str_make (); + str_append (&response, "Status: 404 Not Found\n"); + str_append (&response, "Content-Type: text/plain\n\n"); + request_write (self, response.str, response.len); + str_free (&response); + return false; +} + +static bool +request_push (struct request *self, const void *data, size_t len) +{ + if (!soft_assert (self->handler)) + // No handler, nothing to do with any data + return false; + + return self->handler->push_cb (self, data, len); +} + +/// @} +// --- Requests handlers ------------------------------------------------------- + +static bool +request_handler_json_rpc_try_handle + (struct request *request, struct str_map *headers, bool *continue_) +{ + const char *content_type = str_map_find (headers, "CONTENT_TYPE"); + const char *method = str_map_find (headers, "REQUEST_METHOD"); + + if (!method || strcmp (method, "POST") + || !content_type || !validate_json_rpc_content_type (content_type)) + return false; + + struct str *buf = xcalloc (1, sizeof *buf); + *buf = str_make (); + + request->handler_data = buf; + *continue_ = true; + return true; +} + +static bool +request_handler_json_rpc_push + (struct request *request, const void *data, size_t len) +{ + struct str *buf = request->handler_data; + if (len) + { + str_append_data (buf, data, len); + return true; + } + + // TODO: check buf.len against CONTENT_LENGTH; if it's less, then the + // client hasn't been successful in transferring all of its data. + // See also comment on request_handler::push_cb. + + struct str response = str_make (); + str_append (&response, "Status: 200 OK\n"); + str_append_printf (&response, "Content-Type: %s\n\n", "application/json"); + process_json_rpc (request->ctx, buf->str, buf->len, &response); + request_write (request, response.str, response.len); + str_free (&response); + return false; +} + +static void +request_handler_json_rpc_finalize (struct request *request) +{ + struct str *buf = request->handler_data; + str_free (buf); + free (buf); + + request->handler_data = NULL; +} + +struct request_handler g_request_handler_json_rpc = +{ + .try_handle = request_handler_json_rpc_try_handle, + .push_cb = request_handler_json_rpc_push, + .finalize_cb = request_handler_json_rpc_finalize, +}; + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static char * +canonicalize_url_path (const char *path) +{ + // XXX: this strips any slashes at the end + struct strv v = strv_make (); + cstr_split (path, "/", true, &v); + + struct strv canonical = strv_make (); + + // So that the joined path always begins with a slash + strv_append (&canonical, ""); + + for (size_t i = 0; i < v.len; i++) + { + const char *dir = v.vector[i]; + if (!strcmp (dir, ".")) + continue; + + if (strcmp (dir, "..")) + strv_append (&canonical, dir); + else if (canonical.len > 1) + // ".." never goes above the root + strv_remove (&canonical, canonical.len - 1); + } + strv_free (&v); + + char *joined = strv_join (&canonical, "/"); + strv_free (&canonical); + return joined; +} + +static char * +detect_magic (const void *data, size_t len) +{ + magic_t cookie; + char *mime_type = NULL; + + if (!(cookie = magic_open (MAGIC_MIME))) + return NULL; + + const char *magic = NULL; + if (!magic_load (cookie, NULL) + && (magic = magic_buffer (cookie, data, len))) + mime_type = xstrdup (magic); + else + print_debug ("MIME type detection failed: %s", magic_error (cookie)); + + magic_close (cookie); + return mime_type; +} + +static bool +request_handler_static_try_handle + (struct request *request, struct str_map *headers, bool *continue_) +{ + // Serving static files is actually quite complicated as it turns out; + // but this is only meant to serve a few tiny text files + + struct server_context *ctx = request->ctx; + const char *root = str_map_find (&ctx->config, "static_root"); + if (!root) + { + print_debug ("static document root not configured"); + return false; + } + + // TODO: implement HEAD, we don't get that for free; + // probably implies adding Content-Length + const char *method = str_map_find (headers, "REQUEST_METHOD"); + if (!method || strcmp (method, "GET")) + return false; + + // TODO: look at , REQUEST_URI in the headers + const char *path_info = str_map_find (headers, "PATH_INFO"); + if (!path_info) + path_info = str_map_find (headers, "REQUEST_URI"); + if (!path_info) + { + print_debug ("neither PATH_INFO nor REQUEST_URI was defined"); + return false; + } + + // We need to filter the path to stay in our root + // Being able to read /etc/passwd would be rather embarrasing + char *suffix = canonicalize_url_path (path_info); + char *path = xstrdup_printf ("%s%s", root, suffix); + print_debug ("trying to statically serve %s", path); + + // TODO: check that this is a regular file + FILE *fp = fopen (path, "rb"); + if (!fp) + { + struct str response = str_make (); + str_append (&response, "Status: 404 Not Found\n"); + str_append (&response, "Content-Type: text/plain\n\n"); + str_append_printf (&response, + "File %s was not found on this server\n", suffix); + request_write (request, response.str, response.len); + str_free (&response); + + free (suffix); + free (path); + return false; + } + + free (suffix); + free (path); + + uint8_t buf[8192]; + size_t len; + + // Try to detect the Content-Type from the actual contents + char *mime_type = NULL; + if ((len = fread (buf, 1, sizeof buf, fp))) + mime_type = detect_magic (buf, len); + if (!mime_type) + mime_type = xstrdup ("application/octet_stream"); + + struct str response = str_make (); + str_append (&response, "Status: 200 OK\n"); + str_append_printf (&response, "Content-Type: %s\n\n", mime_type); + request_write (request, response.str, response.len); + str_free (&response); + free (mime_type); + + // Write the chunk we've used to help us with magic detection; + // obviously we have to do it after we've written the headers + if (len) + request_write (request, buf, len); + + while ((len = fread (buf, 1, sizeof buf, fp))) + request_write (request, buf, len); + fclose (fp); + + // TODO: this should rather not be returned all at once but in chunks; + // file read requests never return EAGAIN + // TODO: actual file data should really be returned by a callback when + // the socket is writable with nothing to be sent (pumping the entire + // file all at once won't really work if it's huge). + *continue_ = false; + return true; +} + +static bool +request_handler_static_push + (struct request *request, const void *data, size_t len) +{ + (void) request; + (void) data; + + // Aborting on content; we shouldn't receive any (GET) + // FIXME: there should at least be some indication of this happening + return len == 0; +} + +static void +request_handler_static_finalize (struct request *request) +{ + (void) request; + // Nothing to dispose of this far +} + +struct request_handler g_request_handler_static = +{ + .try_handle = request_handler_static_try_handle, + .push_cb = request_handler_static_push, + .finalize_cb = request_handler_static_finalize, +}; + +// --- Client communication handlers ------------------------------------------- + +/// A virtual class for client connections coming either from the web server +/// or directly from the end-client, depending on the protocol in use +struct client +{ + LIST_HEADER (struct client) + + struct client_vtable *vtable; ///< Client behaviour + + int socket_fd; ///< The network socket + bool received_eof; ///< Whether EOF has been received yet + bool flushing; ///< No more data to write, send FIN + bool closing; ///< No more data to read or write + bool half_closed; ///< Conn. half-closed while flushing + struct write_queue write_queue; ///< Write queue + ev_timer close_timeout_watcher; ///< Write queue flush timer + + ev_io read_watcher; ///< The socket can be read from + ev_io write_watcher; ///< The socket can be written to +}; + +/// The concrete behaviour to serve a particular client's requests +struct client_vtable +{ + /// Process incoming data; "len == 0" means EOF. + /// If the method returns false, client_close() is called by the caller. + bool (*push) (struct client *client, const void *data, size_t len); + + // TODO: optional push_error() to inform about network I/O errors + + /// Attempt a graceful shutdown: make any appropriate steps before + /// the client connection times out and gets torn down by force. + /// The client is allowed to destroy itself immediately. + void (*shutdown) (struct client *client); + + /// Do any additional cleanup for the concrete class before destruction + void (*finalize) (struct client *client); +}; + +static void +client_destroy (struct client *self) +{ + // XXX: this codebase halfway pretends there could be other contexts + struct server_context *ctx = ev_userdata (EV_DEFAULT); + LIST_UNLINK (ctx->clients, self); + ctx->n_clients--; + + // First uninitialize the higher-level implementation + self->vtable->finalize (self); + + ev_io_stop (EV_DEFAULT_ &self->read_watcher); + ev_io_stop (EV_DEFAULT_ &self->write_watcher); + xclose (self->socket_fd); + write_queue_free (&self->write_queue); + ev_timer_stop (EV_DEFAULT_ &self->close_timeout_watcher); + free (self); + + try_finish_quit (ctx); +} + +static void +client_write (struct client *self, const void *data, size_t len) +{ + if (!soft_assert (!self->flushing) || len == 0) + return; + + struct write_req *req = xcalloc (1, sizeof *req); + req->data.iov_base = memcpy (xmalloc (len), data, len); + req->data.iov_len = len; + + write_queue_add (&self->write_queue, req); + ev_io_start (EV_DEFAULT_ &self->write_watcher); +} + +/// Half-close the connection from our side once the write_queue is flushed. +/// It is the caller's responsibility to destroy the connection upon EOF. +// XXX: or we might change on_client_readable to do it anyway, seems safe +static void +client_shutdown (struct client *self) +{ + self->flushing = true; + ev_feed_event (EV_DEFAULT_ &self->write_watcher, EV_WRITE); +} + +/// Try to cleanly close the connection, waiting for the remote client to close +/// its own side of the connection as a sign that it has processed all the data +/// it wanted to. The client implementation will not receive any further data. +/// May directly call client_destroy(). +static void +client_close (struct client *self) +{ + if (self->closing) + return; + + self->closing = true; + ev_timer_start (EV_DEFAULT_ &self->close_timeout_watcher); + client_shutdown (self); + + // We assume the remote client doesn't want our data if it half-closes + if (self->received_eof) + client_destroy (self); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static bool +client_read_loop (EV_P_ struct client *client, ev_io *watcher) +{ + char buf[8192]; + ssize_t n_read; +again: + while ((n_read = recv (watcher->fd, buf, sizeof buf, 0)) >= 0) + { + if (!n_read) + { + // Don't deliver the EOF condition repeatedly + ev_io_stop (EV_A_ watcher); + client->received_eof = true; + } + if (!client->closing + && !client->vtable->push (client, buf, n_read)) + { + client_close (client); + return false; + } + if (!n_read) + return true; + } + if (errno == EINTR) + goto again; + if (errno == EAGAIN) + return true; + + client_destroy (client); + return false; +} + +static void +on_client_readable (EV_P_ ev_io *watcher, int revents) +{ + struct client *client = watcher->data; + (void) revents; + + if (client_read_loop (EV_A_ client, watcher) + && client->closing && client->received_eof) + client_destroy (client); +} + +static void +on_client_writable (EV_P_ ev_io *watcher, int revents) +{ + struct client *client = watcher->data; + (void) loop; + (void) revents; + + // TODO: some sort of "on_buffers_flushed" callback for streaming huge + // chunks of external (or generated) data. That will need to be + // forwarded to "struct request_handler". + if (!flush_queue (&client->write_queue, watcher->fd)) + { + client_destroy (client); + return; + } + if (!write_queue_is_empty (&client->write_queue)) + return; + + ev_io_stop (EV_A_ watcher); + if (client->flushing && !client->half_closed) + { + if (!shutdown (client->socket_fd, SHUT_WR)) + client->half_closed = true; + else + client_destroy (client); + } +} + +static void +on_client_timeout (EV_P_ ev_timer *watcher, int revents) +{ + (void) loop; + (void) revents; + + client_destroy (watcher->data); +} + +/// Create a new instance of a subclass with the given size. +/// The superclass is assumed to be the first member of the structure. +static void * +client_new (EV_P_ size_t size, int sock_fd) +{ + struct server_context *ctx = ev_userdata (loop); + struct client *self = xcalloc (1, size); + + self->write_queue = write_queue_make (); + ev_timer_init (&self->close_timeout_watcher, on_client_timeout, 5., 0.); + self->close_timeout_watcher.data = self; + + set_blocking (sock_fd, false); + self->socket_fd = sock_fd; + + ev_io_init (&self->read_watcher, on_client_readable, sock_fd, EV_READ); + ev_io_init (&self->write_watcher, on_client_writable, sock_fd, EV_WRITE); + self->read_watcher.data = self; + self->write_watcher.data = self; + + // We're only interested in reading as the write queue is empty now + ev_io_start (EV_A_ &self->read_watcher); + + LIST_PREPEND (ctx->clients, self); + ctx->n_clients++; + return self; +} + +// --- FastCGI client handler -------------------------------------------------- + +struct client_fcgi +{ + struct client client; ///< Parent class + struct fcgi_muxer muxer; ///< FastCGI de/multiplexer +}; + +struct client_fcgi_request +{ + struct fcgi_request *fcgi_request; ///< FastCGI request + struct request request; ///< Request +}; + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static void +client_fcgi_request_write_cb (struct request *req, const void *data, size_t len) +{ + FIND_CONTAINER (self, req, struct client_fcgi_request, request); + fcgi_request_write (self->fcgi_request, data, len); +} + +static void +client_fcgi_request_close_cb (struct request *req) +{ + FIND_CONTAINER (self, req, struct client_fcgi_request, request); + struct fcgi_muxer *muxer = self->fcgi_request->muxer; + // No more data to send, terminate the substream/request, + // and also the transport if the client didn't specifically ask to keep it + if (!fcgi_request_finish (self->fcgi_request)) + muxer->close_cb (muxer); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static bool +client_fcgi_request_start (struct fcgi_request *fcgi_request) +{ + struct client_fcgi_request *request = + fcgi_request->handler_data = xcalloc (1, sizeof *request); + request->fcgi_request = fcgi_request; + request_init (&request->request); + request->request.ctx = ev_userdata (EV_DEFAULT); + request->request.write_cb = client_fcgi_request_write_cb; + request->request.close_cb = client_fcgi_request_close_cb; + + return request_start (&request->request, &fcgi_request->headers); +} + +static bool +client_fcgi_request_push + (struct fcgi_request *req, const void *data, size_t len) +{ + struct client_fcgi_request *request = req->handler_data; + return request_push (&request->request, data, len) + || fcgi_request_finish (req); +} + +static void +client_fcgi_request_finalize (struct fcgi_request *req) +{ + struct client_fcgi_request *request = req->handler_data; + request_free (&request->request); + free (request); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static void +client_fcgi_write_cb (struct fcgi_muxer *mux, const void *data, size_t len) +{ + FIND_CONTAINER (self, mux, struct client_fcgi, muxer); + client_write (&self->client, data, len); +} + +static void +client_fcgi_close_cb (struct fcgi_muxer *mux) +{ + FIND_CONTAINER (self, mux, struct client_fcgi, muxer); + client_close (&self->client); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static bool +client_fcgi_push (struct client *client, const void *data, size_t len) +{ + FIND_CONTAINER (self, client, struct client_fcgi, client); + return fcgi_muxer_push (&self->muxer, data, len); +} + +static void +client_fcgi_shutdown (struct client *client) +{ + FIND_CONTAINER (self, client, struct client_fcgi, client); + self->muxer.in_shutdown = true; + + // TODO: respond with FCGI_END_REQUEST: FCGI_REQUEST_COMPLETE to everything? + // The FastCGI specification isn't very clear about what we should do. +} + +static void +client_fcgi_finalize (struct client *client) +{ + FIND_CONTAINER (self, client, struct client_fcgi, client); + fcgi_muxer_free (&self->muxer); +} + +static struct client_vtable client_fcgi_vtable = +{ + .push = client_fcgi_push, + .shutdown = client_fcgi_shutdown, + .finalize = client_fcgi_finalize, +}; + +static struct client * +client_fcgi_create (EV_P_ int sock_fd) +{ + struct client_fcgi *self = client_new (EV_A_ sizeof *self, sock_fd); + self->client.vtable = &client_fcgi_vtable; + + fcgi_muxer_init (&self->muxer); + self->muxer.write_cb = client_fcgi_write_cb; + self->muxer.close_cb = client_fcgi_close_cb; + self->muxer.request_start_cb = client_fcgi_request_start; + self->muxer.request_push_cb = client_fcgi_request_push; + self->muxer.request_finalize_cb = client_fcgi_request_finalize; + return &self->client; +} + +// --- SCGI client handler ----------------------------------------------------- + +struct client_scgi +{ + struct client client; ///< Parent class + struct scgi_parser parser; ///< SCGI stream parser + struct request request; ///< Request (only one per connection) + unsigned long remaining_content; ///< Length of input data to be seen +}; + +static void +client_scgi_write_cb (struct request *req, const void *data, size_t len) +{ + FIND_CONTAINER (self, req, struct client_scgi, request); + client_write (&self->client, data, len); +} + +static void +client_scgi_close_cb (struct request *req) +{ + FIND_CONTAINER (self, req, struct client_scgi, request); + // NOTE: this rather really means "close me [the request]" + client_close (&self->client); +} + +static bool +client_scgi_on_headers_read (void *user_data) +{ + struct client_scgi *self = user_data; + const char *cl = str_map_find (&self->parser.headers, "CONTENT_LENGTH"); + if (!cl || !xstrtoul (&self->remaining_content, cl, 10)) + { + print_debug ("SCGI request with invalid or missing CONTENT_LENGTH"); + return false; + } + return request_start (&self->request, &self->parser.headers); +} + +static bool +client_scgi_on_content (void *user_data, const void *data, size_t len) +{ + struct client_scgi *self = user_data; + if (len > self->remaining_content) + { + print_debug ("SCGI request got more data than CONTENT_LENGTH"); + return false; + } + // We're in a slight disagreement with the specification since + // this tries to write output before it has read all the input + if (!request_push (&self->request, data, len)) + return false; + + // Signalise end of input to the request handler + return (self->remaining_content -= len) != 0 + || request_push (&self->request, NULL, 0); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static bool +client_scgi_push (struct client *client, const void *data, size_t len) +{ + struct client_scgi *self = (struct client_scgi *) client; + struct error *e = NULL; + if (scgi_parser_push (&self->parser, data, len, &e)) + return true; + + if (e != NULL) + { + print_debug ("SCGI parser failed: %s", e->message); + error_free (e); + } + return false; +} + +static void +client_scgi_finalize (struct client *client) +{ + struct client_scgi *self = (struct client_scgi *) client; + request_free (&self->request); + scgi_parser_free (&self->parser); +} + +static struct client_vtable client_scgi_vtable = +{ + .push = client_scgi_push, + .finalize = client_scgi_finalize, +}; + +static struct client * +client_scgi_create (EV_P_ int sock_fd) +{ + struct client_scgi *self = client_new (EV_A_ sizeof *self, sock_fd); + self->client.vtable = &client_scgi_vtable; + + request_init (&self->request); + self->request.ctx = ev_userdata (EV_DEFAULT); + self->request.write_cb = client_scgi_write_cb; + self->request.close_cb = client_scgi_close_cb; + + self->parser = scgi_parser_make (); + self->parser.on_headers_read = client_scgi_on_headers_read; + self->parser.on_content = client_scgi_on_content; + self->parser.user_data = self; + return &self->client; +} + +// --- WebSockets client handler ----------------------------------------------- + +struct client_ws +{ + struct client client; ///< Parent class + struct ws_handler handler; ///< WebSockets connection handler +}; + +static bool +client_ws_on_message (struct ws_handler *handler, + enum ws_opcode type, const void *data, size_t len) +{ + FIND_CONTAINER (self, handler, struct client_ws, handler); + if (type != WS_OPCODE_TEXT) + { + return ws_handler_fail_connection + (&self->handler, WS_STATUS_UNSUPPORTED_DATA); + } + + struct server_context *ctx = ev_userdata (EV_DEFAULT); + struct str response = str_make (); + process_json_rpc (ctx, data, len, &response); + if (response.len) + ws_handler_send_frame (&self->handler, + WS_OPCODE_TEXT, response.str, response.len); + str_free (&response); + return true; +} + +static void +client_ws_write_cb (struct ws_handler *handler, const void *data, size_t len) +{ + FIND_CONTAINER (self, handler, struct client_ws, handler); + client_write (&self->client, data, len); +} + +static void +client_ws_close_cb (struct ws_handler *handler, bool half_close) +{ + FIND_CONTAINER (self, handler, struct client_ws, handler); + (half_close ? client_shutdown : client_destroy) (&self->client); +} + +// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + +static bool +client_ws_push (struct client *client, const void *data, size_t len) +{ + FIND_CONTAINER (self, client, struct client_ws, client); + // client_close() will correctly destroy the client on EOF + return ws_handler_push (&self->handler, data, len); +} + +static void +client_ws_shutdown (struct client *client) +{ + FIND_CONTAINER (self, client, struct client_ws, client); + if (self->handler.state == WS_HANDLER_CONNECTING) + // No on_close, no problem + client_destroy (&self->client); + else if (self->handler.state == WS_HANDLER_OPEN) + ws_handler_close (&self->handler, WS_STATUS_GOING_AWAY, NULL, 0); +} + +static void +client_ws_finalize (struct client *client) +{ + FIND_CONTAINER (self, client, struct client_ws, client); + ws_handler_free (&self->handler); +} + +static struct client_vtable client_ws_vtable = +{ + .push = client_ws_push, + .shutdown = client_ws_shutdown, + .finalize = client_ws_finalize, +}; + +static struct client * +client_ws_create (EV_P_ int sock_fd) +{ + struct client_ws *self = client_new (EV_A_ sizeof *self, sock_fd); + self->client.vtable = &client_ws_vtable; + + ws_handler_init (&self->handler); + self->handler.on_message = client_ws_on_message; + self->handler.write_cb = client_ws_write_cb; + self->handler.close_cb = client_ws_close_cb; + + // One mebibyte seems to be a reasonable value + self->handler.max_payload_len = 1 << 10; + + ws_handler_start (&self->handler); + return &self->client; +} + +// --- Basic server stuff ------------------------------------------------------ + +typedef struct client *(*client_create_fn) (EV_P_ int sock_fd); + +struct listener +{ + int fd; ///< Listening socket FD + ev_io watcher; ///< New connection available + client_create_fn create; ///< Client constructor +}; + +static void +close_listeners (struct server_context *self) +{ + for (size_t i = 0; i < self->n_listeners; i++) + { + struct listener *listener = &self->listeners[i]; + if (listener->fd == -1) + continue; + + ev_io_stop (EV_DEFAULT_ &listener->watcher); + xclose (listener->fd); + listener->fd = -1; + } +} + +static void +try_finish_quit (struct server_context *self) +{ + if (!self->quitting || self->clients) + return; + + ev_timer_stop (EV_DEFAULT_ &self->quit_timeout_watcher); + ev_break (EV_DEFAULT_ EVBREAK_ALL); +} + +static void +on_quit_timeout (EV_P_ ev_timer *watcher, int revents) +{ + struct server_context *self = watcher->data; + (void) loop; + (void) revents; + + LIST_FOR_EACH (struct client, iter, self->clients) + client_destroy (iter); +} + +static void +initiate_quit (struct server_context *self) +{ + self->quitting = true; + close_listeners (self); + + // Wait a little while for all clients to clean up, if necessary + LIST_FOR_EACH (struct client, iter, self->clients) + if (iter->vtable->shutdown) + iter->vtable->shutdown (iter); + ev_timer_start (EV_DEFAULT_ &self->quit_timeout_watcher); + try_finish_quit (self); +} + +static void +on_client_available (EV_P_ ev_io *watcher, int revents) +{ + struct server_context *ctx = ev_userdata (loop); + struct listener *listener = watcher->data; + (void) revents; + + while (true) + { + int sock_fd = accept (watcher->fd, NULL, NULL); + if (sock_fd != -1) + listener->create (EV_A_ sock_fd); + else if (errno == EAGAIN) + return; + else if (errno != EINTR && errno != EMFILE + && errno != ECONNRESET && errno != ECONNABORTED) + break; + } + + // Stop accepting connections to prevent busy looping + ev_io_stop (EV_A_ watcher); + + print_fatal ("%s: %s", "accept", strerror (errno)); + initiate_quit (ctx); +} + +// --- Application setup ------------------------------------------------------- + +/// This function handles values that require validation before their first use, +/// or some kind of a transformation (such as conversion to an integer) needs +/// to be done before they can be used directly. +static bool +parse_config (struct server_context *ctx, struct error **e) +{ + (void) ctx; + (void) e; + + return true; +} + +static int +listener_bind (struct addrinfo *gai_iter) +{ + int fd = socket (gai_iter->ai_family, + gai_iter->ai_socktype, gai_iter->ai_protocol); + if (fd == -1) + return -1; + set_cloexec (fd); + + int yes = 1; + soft_assert (setsockopt (fd, SOL_SOCKET, SO_KEEPALIVE, + &yes, sizeof yes) != -1); + soft_assert (setsockopt (fd, SOL_SOCKET, SO_REUSEADDR, + &yes, sizeof yes) != -1); + + char host[NI_MAXHOST], port[NI_MAXSERV]; + host[0] = port[0] = '\0'; + int err = getnameinfo (gai_iter->ai_addr, gai_iter->ai_addrlen, + host, sizeof host, port, sizeof port, + NI_NUMERICHOST | NI_NUMERICSERV); + if (err) + print_debug ("%s: %s", "getnameinfo", gai_strerror (err)); + + char *address = format_host_port_pair (host, port); + if (bind (fd, gai_iter->ai_addr, gai_iter->ai_addrlen)) + print_error ("bind to %s failed: %s", address, strerror (errno)); + else if (listen (fd, 16 /* arbitrary number */)) + print_error ("listen on %s failed: %s", address, strerror (errno)); + else + { + print_status ("listening on %s", address); + free (address); + return fd; + } + + free (address); + xclose (fd); + return -1; +} + +static void +listener_add (struct server_context *ctx, const char *host, const char *port, + const struct addrinfo *gai_hints, client_create_fn create) +{ + struct addrinfo *gai_result, *gai_iter; + int err = getaddrinfo (host, port, gai_hints, &gai_result); + if (err) + { + char *address = format_host_port_pair (host, port); + print_error ("bind to %s failed: %s: %s", + address, "getaddrinfo", gai_strerror (err)); + free (address); + return; + } + + int fd; + for (gai_iter = gai_result; gai_iter; gai_iter = gai_iter->ai_next) + { + if ((fd = listener_bind (gai_iter)) == -1) + continue; + set_blocking (fd, false); + + struct listener *listener = &ctx->listeners[ctx->n_listeners++]; + ev_io_init (&listener->watcher, on_client_available, fd, EV_READ); + ev_io_start (EV_DEFAULT_ &listener->watcher); + listener->watcher.data = listener; + listener->create = create; + listener->fd = fd; + break; + } + freeaddrinfo (gai_result); +} + +static void +get_ports_from_config (struct server_context *ctx, + const char *key, struct strv *out) +{ + const char *ports; + if ((ports = str_map_find (&ctx->config, key))) + cstr_split (ports, ",", true, out); +} + +static bool +setup_listen_fds (struct server_context *ctx, struct error **e) +{ + static const struct addrinfo gai_hints = + { + .ai_socktype = SOCK_STREAM, + .ai_flags = AI_PASSIVE, + }; + + struct strv ports_fcgi = strv_make (); + struct strv ports_scgi = strv_make (); + struct strv ports_ws = strv_make (); + + get_ports_from_config (ctx, "port_fastcgi", &ports_fcgi); + get_ports_from_config (ctx, "port_scgi", &ports_scgi); + get_ports_from_config (ctx, "port_ws", &ports_ws); + + const char *bind_host = str_map_find (&ctx->config, "bind_host"); + size_t n_ports = ports_fcgi.len + ports_scgi.len + ports_ws.len; + ctx->listeners = xcalloc (n_ports, sizeof *ctx->listeners); + + for (size_t i = 0; i < ports_fcgi.len; i++) + listener_add (ctx, bind_host, ports_fcgi.vector[i], + &gai_hints, client_fcgi_create); + for (size_t i = 0; i < ports_scgi.len; i++) + listener_add (ctx, bind_host, ports_scgi.vector[i], + &gai_hints, client_scgi_create); + for (size_t i = 0; i < ports_ws.len; i++) + listener_add (ctx, bind_host, ports_ws.vector[i], + &gai_hints, client_ws_create); + + strv_free (&ports_fcgi); + strv_free (&ports_scgi); + strv_free (&ports_ws); + + if (!ctx->n_listeners) + { + error_set (e, "%s: %s", + "network setup failed", "no ports to listen on"); + return false; + } + return true; +} + +static bool +app_lock_pid_file (struct server_context *ctx, struct error **e) +{ + const char *path = str_map_find (&ctx->config, "pid_file"); + if (!path) + return true; + + char *resolved = resolve_filename (path, resolve_relative_runtime_filename); + bool result = lock_pid_file (resolved, e) != -1; + free (resolved); + return result; +} + +// --- Tests ------------------------------------------------------------------- + +static void +test_misc (void) +{ + soft_assert ( validate_json_rpc_content_type + ("application/JSON; charset=\"utf-8\"")); + soft_assert (!validate_json_rpc_content_type + ("text/html; charset=\"utf-8\"")); + + char *canon = canonicalize_url_path ("///../../../etc/./passwd"); + soft_assert (!strcmp (canon, "/etc/passwd")); + free (canon); +} + +int +test_main (int argc, char *argv[]) +{ + struct test test; + test_init (&test, argc, argv); + + test_add_simple (&test, "/misc", NULL, test_misc); + + // TODO: write more tests + // TODO: test the server handler (happy path) + + return test_run (&test); +} + +// --- Main program ------------------------------------------------------------ + +static void +on_termination_signal (EV_P_ ev_signal *handle, int revents) +{ + struct server_context *ctx = ev_userdata (loop); + (void) handle; + (void) revents; + + if (ctx->quitting) + { + // Double C-c from the terminal accelerates the process + LIST_FOR_EACH (struct client, iter, ctx->clients) + client_destroy (iter); + } + else + initiate_quit (ctx); +} + +static void +setup_signal_handlers (struct server_context *ctx) +{ + ev_signal_init (&ctx->sigterm_watcher, on_termination_signal, SIGTERM); + ev_signal_start (EV_DEFAULT_ &ctx->sigterm_watcher); + + ev_signal_init (&ctx->sigint_watcher, on_termination_signal, SIGINT); + ev_signal_start (EV_DEFAULT_ &ctx->sigint_watcher); + + (void) signal (SIGPIPE, SIG_IGN); +} + +static void +daemonize (struct server_context *ctx) +{ + print_status ("daemonizing..."); + + if (chdir ("/")) + exit_fatal ("%s: %s", "chdir", strerror (errno)); + + // Because of systemd, we need to exit the parent process _after_ writing + // a PID file, otherwise our grandchild would receive a SIGTERM + int sync_pipe[2]; + if (pipe (sync_pipe)) + exit_fatal ("%s: %s", "pipe", strerror (errno)); + + pid_t pid; + if ((pid = fork ()) < 0) + exit_fatal ("%s: %s", "fork", strerror (errno)); + else if (pid) + { + // Wait until all write ends of the pipe are closed, which can mean + // either success or failure, we don't need to care + xclose (sync_pipe[PIPE_WRITE]); + + char dummy; + if (read (sync_pipe[PIPE_READ], &dummy, 1) < 0) + exit_fatal ("%s: %s", "read", strerror (errno)); + + exit (EXIT_SUCCESS); + } + + setsid (); + signal (SIGHUP, SIG_IGN); + + if ((pid = fork ()) < 0) + exit_fatal ("%s: %s", "fork", strerror (errno)); + else if (pid) + exit (EXIT_SUCCESS); + + openlog (PROGRAM_NAME, LOG_NDELAY | LOG_NOWAIT | LOG_PID, 0); + g_log_message_real = log_message_syslog; + + // Write the PID file (if so configured) and get rid of the pipe, so that + // the read() in our grandparent finally returns zero (no write ends) + struct error *e = NULL; + if (!app_lock_pid_file (ctx, &e)) + exit_fatal ("%s", e->message); + + xclose (sync_pipe[PIPE_READ]); + xclose (sync_pipe[PIPE_WRITE]); + + // XXX: we may close our own descriptors this way, crippling ourselves; + // there is no real guarantee that we will start with all three + // descriptors open. In theory we could try to enumerate the descriptors + // at the start of main(). + for (int i = 0; i < 3; i++) + xclose (i); + + int tty = open ("/dev/null", O_RDWR); + if (tty != 0 || dup (0) != 1 || dup (0) != 2) + exit_fatal ("failed to reopen FD's: %s", strerror (errno)); +} + +static void +parse_program_arguments (int argc, char **argv) +{ + static const struct opt opts[] = + { + { 't', "test", NULL, 0, "self-test" }, + { 'd', "debug", NULL, 0, "run in debug mode" }, + { 'h', "help", NULL, 0, "display this help and exit" }, + { 'V', "version", NULL, 0, "output version information and exit" }, + { 'w', "write-default-cfg", "FILENAME", + OPT_OPTIONAL_ARG | OPT_LONG_ONLY, + "write a default configuration file and exit" }, + { 0, NULL, NULL, 0, NULL } + }; + + struct opt_handler oh = + opt_handler_make (argc, argv, opts, NULL, "JSON-RPC 2.0 demo server."); + + int c; + while ((c = opt_handler_get (&oh)) != -1) + switch (c) + { + case 't': + test_main (argc, argv); + exit (EXIT_SUCCESS); + case 'd': + g_debug_mode = true; + break; + case 'h': + opt_handler_usage (&oh, stdout); + exit (EXIT_SUCCESS); + case 'V': + printf (PROGRAM_NAME " " PROGRAM_VERSION "\n"); + exit (EXIT_SUCCESS); + case 'w': + call_simple_config_write_default (optarg, g_config_table); + exit (EXIT_SUCCESS); + default: + print_error ("wrong options"); + opt_handler_usage (&oh, stderr); + exit (EXIT_FAILURE); + } + + argc -= optind; + argv += optind; + + if (argc) + { + opt_handler_usage (&oh, stderr); + exit (EXIT_FAILURE); + } + opt_handler_free (&oh); +} + +int +main (int argc, char *argv[]) +{ + parse_program_arguments (argc, argv); + + print_status (PROGRAM_NAME " " PROGRAM_VERSION " starting"); + + struct server_context ctx; + server_context_init (&ctx); + + struct error *e = NULL; + if (!simple_config_update_from_file (&ctx.config, &e)) + { + print_error ("error loading configuration: %s", e->message); + error_free (e); + exit (EXIT_FAILURE); + } + + struct ev_loop *loop; + if (!(loop = EV_DEFAULT)) + exit_fatal ("libev initialization failed"); + + ev_set_userdata (loop, &ctx); + setup_signal_handlers (&ctx); + + LIST_PREPEND (ctx.handlers, &g_request_handler_static); + LIST_PREPEND (ctx.handlers, &g_request_handler_json_rpc); + + if (!parse_config (&ctx, &e) + || !setup_listen_fds (&ctx, &e)) + { + print_error ("%s", e->message); + error_free (e); + exit (EXIT_FAILURE); + } + + if (!g_debug_mode) + daemonize (&ctx); + else if (!app_lock_pid_file (&ctx, &e)) + exit_fatal ("%s", e->message); + + ev_run (loop, 0); + ev_loop_destroy (loop); + + server_context_free (&ctx); + return EXIT_SUCCESS; +} diff --git a/liberty b/liberty index bb30c7d..bca7167 160000 --- a/liberty +++ b/liberty @@ -1 +1 @@ -Subproject commit bb30c7d86ef7b165c5e00cc8e0ad2e3e85b9e617 +Subproject commit bca7167d037d857448cb18243425d7c61de3bdd5