We have services that use UndertowOptions.MAX_ENTITY_SIZE to limit the size of incoming bodies.

Our error management handles undertow's RequestTooBigException by returning 413: Payload too large http status code.

This does not work since upgrading to undertow version 2.3.19. Undertow automatically return 400: Bad Request }}and our error management is not notified of{{ {{RequestTooBigException}} anymore.

After troubleshooting I found this change in 2.3.19:

io.undertow.server.protocol.http.HttpTransferEncoding.handleRequestEncoding() method now contains in lines 126 to 129:

if (exchange.getMaxEntitySize() > 0 && exchange.getMaxEntitySize() < contentLength && exchange.isResponseChannelAvailable()) {
    persistentConnection = false;
    Connectors.terminateRequest(exchange);
    throw UndertowMessages.MESSAGES.requestEntityWasTooLarge(exchange.getMaxEntitySize());
} 

The exception is caught in io.undertow.server.protocol.http.HttpReadListener line 267 and 400 is returned and connection closed.

In 2.3.18 RequestTooBigException was thrown in io.undertow.conduits.FixedLengthStreamSourceConduit.requestEntityWasTooLarge(final long size) lines 166 to 173

if (exchange.getMaxEntitySize() > 0 && exchange.getMaxEntitySize() < (state & MASK_COUNT)) {
    //max entity size is exceeded
    //we need to forcibly close the read side
    Connectors.terminateRequest(exchange);
    exchange.setPersistent(false);
    finishListener.handleEvent(this);
    this.state |= FLAG_FINISHED | FLAG_CLOSED;
    throw UndertowMessages.MESSAGES.requestEntityWasTooLarge(exchange.getMaxEntitySize());
} 

This allowed our error management to catch it and return 413: Payload too large.