Websocket: Validate UTF-8 in text payloads (#418)

Validate text payloads, as required by [RFC-6455 Section 5.6](https://www.rfc-editor.org/rfc/rfc6455#section-5.6)

include/aws/http/private/websocket_decoder.h

@@ -26,6 +26,7 @@ enum aws_websocket_decoder_state {
     AWS_WEBSOCKET_DECODER_STATE_MASKING_KEY,
     AWS_WEBSOCKET_DECODER_STATE_PAYLOAD_CHECK,
     AWS_WEBSOCKET_DECODER_STATE_PAYLOAD,
+    AWS_WEBSOCKET_DECODER_STATE_FRAME_END,
     AWS_WEBSOCKET_DECODER_STATE_DONE,
 };
 
@@ -38,6 +39,11 @@ struct aws_websocket_decoder {
 
     bool expecting_continuation_data_frame; /* True when the next data frame must be CONTINUATION frame */
 
+    /* True while processing a TEXT "message" (from the start of a TEXT frame,
+     * until the end of the TEXT or CONTINUATION frame with the FIN bit set). */
+    bool processing_text_message;
+    struct aws_utf8_validator *text_message_validator;
+
     void *user_data;
     aws_websocket_decoder_frame_fn *on_frame;
     aws_websocket_decoder_payload_fn *on_payload;
@@ -48,19 +54,22 @@ AWS_EXTERN_C_BEGIN
 AWS_HTTP_API
 void aws_websocket_decoder_init(
     struct aws_websocket_decoder *decoder,
+    struct aws_allocator *alloc,
     aws_websocket_decoder_frame_fn *on_frame,
     aws_websocket_decoder_payload_fn *on_payload,
     void *user_data);
 
+AWS_HTTP_API
+void aws_websocket_decoder_clean_up(struct aws_websocket_decoder *decoder);
+
 /**
  * Returns when all data is processed, or a frame and its payload have completed.
  * `data` will be advanced to reflect the amount of data processed by this call.
  * `frame_complete` will be set true if this call returned due to completion of a frame.
  * The `on_frame` and `on_payload` callbacks may each be invoked once as a result of this call.
  * If an error occurs, the decoder is invalid forevermore.
  */
-AWS_HTTP_API
-int aws_websocket_decoder_process(
+AWS_HTTP_API int aws_websocket_decoder_process(
     struct aws_websocket_decoder *decoder,
     struct aws_byte_cursor *data,
     bool *frame_complete);

source/websocket.c

@@ -304,7 +304,8 @@ struct aws_websocket *aws_websocket_handler_new(const struct aws_websocket_handl
 
     aws_websocket_encoder_init(&websocket->thread_data.encoder, s_encoder_stream_outgoing_payload, websocket);
 
-    aws_websocket_decoder_init(&websocket->thread_data.decoder, s_decoder_on_frame, s_decoder_on_payload, websocket);
+    aws_websocket_decoder_init(
+        &websocket->thread_data.decoder, options->allocator, s_decoder_on_frame, s_decoder_on_payload, websocket);
 
     aws_linked_list_init(&websocket->synced_data.outgoing_frame_list);
 
@@ -346,6 +347,7 @@ static void s_handler_destroy(struct aws_channel_handler *handler) {
 
     AWS_LOGF_TRACE(AWS_LS_HTTP_WEBSOCKET, "id=%p: Destroying websocket.", (void *)websocket);
 
+    aws_websocket_decoder_clean_up(&websocket->thread_data.decoder);
     aws_byte_buf_clean_up(&websocket->thread_data.incoming_ping_payload);
     aws_mutex_clean_up(&websocket->synced_data.lock);
     aws_mem_release(websocket->alloc, websocket);

source/websocket_decoder.c

@@ -5,6 +5,8 @@
 
 #include <aws/http/private/websocket_decoder.h>
 
+#include <aws/common/encoding.h>
+
 #include <inttypes.h>
 
 typedef int(state_fn)(struct aws_websocket_decoder *decoder, struct aws_byte_cursor *data);
@@ -86,6 +88,10 @@ static int s_state_opcode_byte(struct aws_websocket_decoder *decoder, struct aws
         }
     }
 
+    if (decoder->current_frame.opcode == AWS_WEBSOCKET_OPCODE_TEXT) {
+        decoder->processing_text_message = true;
+    }
+
     decoder->state = AWS_WEBSOCKET_DECODER_STATE_LENGTH_BYTE;
     return AWS_OP_SUCCESS;
 }
@@ -234,7 +240,7 @@ static int s_state_payload_check(struct aws_websocket_decoder *decoder, struct a
         decoder->state_bytes_processed = 0;
         decoder->state = AWS_WEBSOCKET_DECODER_STATE_PAYLOAD;
     } else {
-        decoder->state = AWS_WEBSOCKET_DECODER_STATE_DONE;
+        decoder->state = AWS_WEBSOCKET_DECODER_STATE_FRAME_END;
     }
 
     return AWS_OP_SUCCESS;
@@ -266,9 +272,16 @@ static int s_state_payload(struct aws_websocket_decoder *decoder, struct aws_byt
         }
     }
 
-    /* TODO: validate utf-8 */
     /* TODO: validate payload of CLOSE frame */
 
+    /* Validate the UTF-8 for TEXT messages (a TEXT frame and any subsequent CONTINUATION frames) */
+    if (decoder->processing_text_message && aws_websocket_is_data_frame(decoder->current_frame.opcode)) {
+        if (aws_utf8_validator_update(decoder->text_message_validator, payload)) {
+            AWS_LOGF_ERROR(AWS_LS_HTTP_WEBSOCKET, "id=%p: Received invalid UTF-8", (void *)decoder->user_data);
+            return aws_raise_error(AWS_ERROR_HTTP_WEBSOCKET_PROTOCOL_ERROR);
+        }
+    }
+
     /* Invoke on_payload() callback to inform user of payload data */
     int err = decoder->on_payload(payload, decoder->user_data);
     if (err) {
@@ -280,9 +293,34 @@ static int s_state_payload(struct aws_websocket_decoder *decoder, struct aws_byt
 
     /* If all data consumed, proceed to next state. */
     if (decoder->state_bytes_processed == decoder->current_frame.payload_length) {
-        decoder->state++;
+        decoder->state = AWS_WEBSOCKET_DECODER_STATE_FRAME_END;
+    }
+
+    return AWS_OP_SUCCESS;
+}
+
+/* FRAME_END: Perform checks once we reach the end of the frame. */
+static int s_state_frame_end(struct aws_websocket_decoder *decoder, struct aws_byte_cursor *data) {
+    (void)data;
+
+    /* If we're done processing a text message (a TEXT frame and any subsequent CONTINUATION frames),
+     * complete the UTF-8 validation */
+    if (decoder->processing_text_message && aws_websocket_is_data_frame(decoder->current_frame.opcode) &&
+        decoder->current_frame.fin) {
+
+        if (aws_utf8_validator_finalize(decoder->text_message_validator)) {
+            AWS_LOGF_ERROR(
+                AWS_LS_HTTP_WEBSOCKET,
+                "id=%p: Received invalid UTF-8 (incomplete encoding)",
+                (void *)decoder->user_data);
+            return aws_raise_error(AWS_ERROR_HTTP_WEBSOCKET_PROTOCOL_ERROR);
+        }
+
+        decoder->processing_text_message = false;
     }
 
+    /* Done! */
+    decoder->state = AWS_WEBSOCKET_DECODER_STATE_DONE;
     return AWS_OP_SUCCESS;
 }
 
@@ -295,6 +333,7 @@ static state_fn *s_state_functions[AWS_WEBSOCKET_DECODER_STATE_DONE] = {
     s_state_masking_key,
     s_state_payload_check,
     s_state_payload,
+    s_state_frame_end,
 };
 
 int aws_websocket_decoder_process(
@@ -330,6 +369,7 @@ int aws_websocket_decoder_process(
 
 void aws_websocket_decoder_init(
     struct aws_websocket_decoder *decoder,
+    struct aws_allocator *alloc,
     aws_websocket_decoder_frame_fn *on_frame,
     aws_websocket_decoder_payload_fn *on_payload,
     void *user_data) {
@@ -338,4 +378,10 @@ void aws_websocket_decoder_init(
     decoder->user_data = user_data;
     decoder->on_frame = on_frame;
     decoder->on_payload = on_payload;
+    decoder->text_message_validator = aws_utf8_validator_new(alloc);
+}
+
+void aws_websocket_decoder_clean_up(struct aws_websocket_decoder *decoder) {
+    aws_utf8_validator_destroy(decoder->text_message_validator);
+    AWS_ZERO_STRUCT(*decoder);
 }

tests/CMakeLists.txt

@@ -167,6 +167,10 @@ add_test_case(websocket_decoder_fail_on_unknown_opcode)
 add_test_case(websocket_decoder_fragmented_message)
 add_test_case(websocket_decoder_fail_on_bad_fragmentation)
 add_test_case(websocket_decoder_control_frame_cannot_be_fragmented)
+add_test_case(websocket_decoder_utf8_text)
+add_test_case(websocket_decoder_fail_on_bad_utf8_text)
+add_test_case(websocket_decoder_fragmented_utf8_text)
+add_test_case(websocket_decoder_fail_on_fragmented_bad_utf8_text)
 add_test_case(websocket_decoder_on_frame_callback_can_fail_decoder)
 add_test_case(websocket_decoder_on_payload_callback_can_fail_decoder)
 add_test_case(websocket_encoder_sanity_check)