1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
//! Support for stream based connections.
use core::ops::{ControlFlow, Deref};
use core::time::Duration;
use std::io;
use std::net::SocketAddr;
use std::sync::Arc;
use octseq::Octets;
use tokio::io::{
AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt, ReadHalf, WriteHalf,
};
use tokio::sync::mpsc::error::TrySendError;
use tokio::sync::mpsc::Sender;
use tokio::sync::{mpsc, watch};
use tokio::time::Instant;
use tokio::time::{sleep_until, timeout};
use tracing::Level;
use tracing::{debug, enabled, error, trace, warn};
use crate::base::wire::Composer;
use crate::base::{Message, StreamTarget};
use crate::net::server::buf::BufSource;
use crate::net::server::message::CommonMessageFlow;
use crate::net::server::message::Request;
use crate::net::server::metrics::ServerMetrics;
use crate::net::server::middleware::chain::MiddlewareChain;
use crate::net::server::service::{
CallResult, Service, ServiceError, ServiceFeedback,
};
use crate::net::server::util::to_pcap_text;
use crate::utils::config::DefMinMax;
use super::message::{NonUdpTransportContext, TransportSpecificContext};
use super::middleware::builder::MiddlewareBuilder;
use super::stream::Config as ServerConfig;
use super::ServerCommand;
use std::fmt::Display;
/// Limit on the amount of time to allow between client requests.
///
/// According to [RFC 7766]:
/// - "A timeout of at least a few seconds is advisable for normal
/// operations".
/// - "Servers MAY use zero timeouts when they are experiencing heavy load or
/// are under attack".
/// - "Servers MAY allow idle connections to remain open for longer periods as
/// resources permit".
///
/// The value has to be between zero and 30 days with a default of 30 seconds.
/// The default and minimum values are the same as those of the Unbound 1.19.2
/// `tcp-idle-timeout` configuration setting. The upper bound is a guess at
/// something reasonable.
///
/// [RFC 7766]: https://datatracker.ietf.org/doc/html/rfc7766#section-6.2.3
//
// Note: Unbound 1.19.2 has another setting, edns-tcp-keepalive-timeout,
// which if set and edns-tcp-keepalive is set to yes, then Unbound has a
// default timeout value of 2 minutes instead of 30 seconds. Internally both
// Unbound options configure the same timeout limit, the tcp-idle-timeout
// setting may exist for backward compatibility. TO DO: Should we increase
// the default timeout value to 2 minutes instead of 30 seconds?
const IDLE_TIMEOUT: DefMinMax<Duration> = DefMinMax::new(
Duration::from_secs(30),
Duration::from_millis(200),
Duration::from_secs(30 * 24 * 60 * 60),
);
/// Limit on the amount of time to wait for writing a response to complete.
///
/// The value has to be between 1 millisecond and 1 hour with a default of 30
/// seconds. These values are guesses at something reasonable. The default is
/// based on the Unbound 1.19.2 default value for its `tcp-idle-timeout`
/// setting.
const RESPONSE_WRITE_TIMEOUT: DefMinMax<Duration> = DefMinMax::new(
Duration::from_secs(30),
Duration::from_millis(1),
Duration::from_secs(60 * 60),
);
/// Limit on the number of DNS responses queued for writing to the client.
///
/// The value has to be between zero and 1,024. The default value is 10. These
/// numbers are just a guess at something reasonable.
///
/// If the limit is hit handling of client requests will block until space
/// becomes available.
const MAX_QUEUED_RESPONSES: DefMinMax<usize> = DefMinMax::new(10, 0, 1024);
//----------- Config ---------------------------------------------------------
/// Configuration for a stream server connection.
pub struct Config<RequestOctets, Target> {
/// Limit on the amount of time to allow between client requests.
///
/// This setting can be overridden on a per connection basis by a
/// [`Service`] by returning a [`ServerCommand::Reconfigure`] command
/// with a response message, for example to adjust it per [RFC 7828]
/// section 3.3.1 "Receivomg queries" which says:
///
/// A DNS server that receives a query using TCP transport that includes
/// the edns-tcp-keepalive option MAY modify the local idle timeout
/// associated with that TCP session if resources permit. idle_timeout:
/// Duration,
///
/// [RFC 7828]: https://datatracker.ietf.org/doc/html/rfc7828#section-3.3.1
idle_timeout: Duration,
/// Limit on the amount of time to wait for writing a response to
/// complete.
///
/// The value has to be between 1 millisecond and 1 hour with a default of
/// 30 seconds. These values are guesses at something reasonable. The
/// default is based on the Unbound 1.19.2 default value for its
/// `tcp-idle-timeout` setting.
response_write_timeout: Duration,
/// Limit on the number of DNS responses queued for wriing to the client.
max_queued_responses: usize,
/// The middleware chain used to pre-process requests and post-process
/// responses.
middleware_chain: MiddlewareChain<RequestOctets, Target>,
}
impl<RequestOctets, Target> Config<RequestOctets, Target>
where
RequestOctets: Octets,
Target: Composer + Default,
{
/// Creates a new, default config.
#[allow(dead_code)]
pub fn new() -> Self {
Default::default()
}
/// Set the limit on the amount of time to allow between client requests.
///
/// According to [RFC 7766]:
/// - "A timeout of at least a few seconds is advisable for normal
/// operations".
/// - "Servers MAY use zero timeouts when they are experiencing heavy load
/// or are under attack".
/// - "Servers MAY allow idle connections to remain open for longer
/// periods as resources permit".
///
/// The value has to be between zero and 30 days with a default of 30
/// seconds. The default and minimum values are the same as those of the
/// Unbound 1.19.2 `tcp-idle-timeout` configuration setting. The upper
/// bound is a guess at something reasonable.
///
/// # Reconfigure
///
/// On [`StreamServer::reconfigure`] the current idle period will NOT be
/// affected. Subsequent idle periods (after the next message is received
/// or response is sent, assuming that happens within the current idle
/// period) will use the new timeout value.
///
/// [RFC 7766]:
/// https://datatracker.ietf.org/doc/html/rfc7766#section-6.2.3
///
/// [`StreamServer::reconfigure`]:
/// super::stream::StreamServer::reconfigure()
#[allow(dead_code)]
pub fn set_idle_timeout(&mut self, value: Duration) {
self.idle_timeout = value;
}
/// Set the limit on the amount of time to wait for writing a response to
/// complete.
///
/// The value has to be between 1 millisecond and 1 hour with a default of
/// 30 seconds. These values are guesses at something reasonable. The
/// default is based on the Unbound 1.19.2 default value for its
/// `tcp-idle-timeout` setting.
///
/// # Reconfigure
///
/// On [`StreamServer::reconfigure`] any responses currently being
/// written will NOT use the new timeout, it will only apply to responses
/// that start being sent after the timeout is changed.
///
/// [`StreamServer::reconfigure`]:
/// super::stream::StreamServer::reconfigure()
#[allow(dead_code)]
pub fn set_response_write_timeout(&mut self, value: Duration) {
self.response_write_timeout = value;
}
/// Set the limit on the number of DNS responses queued for writing to the
/// client.
///
/// The value has to be between zero and 1,024. The default value is 10.
/// These numbers are just a guess at something reasonable.
///
/// DNS response messages will be discarded if they cannot be queued for
/// sending because the queue is full.
///
/// # Reconfigure
///
/// On [`StreamServer::reconfigure`] only new connections created after
/// this setting is changed will use the new value, existing connections
/// will continue to use their exisitng queue at its existing size.
///
/// [`StreamServer::reconfigure`]:
/// super::stream::StreamServer::reconfigure()
#[allow(dead_code)]
pub fn set_max_queued_responses(&mut self, value: usize) {
self.max_queued_responses = value;
}
/// Set the middleware chain used to pre-process requests and post-process
/// responses.
///
/// # Reconfigure
///
/// On [`StreamServer::reconfigure`] only new connections created after
/// this setting is changed will use the new value, existing connections
/// and in-flight requests (and their responses) will continue to use
/// their current middleware chain.
///
/// [`StreamServer::reconfigure`]:
/// super::stream::StreamServer::reconfigure()
pub fn set_middleware_chain(
&mut self,
value: MiddlewareChain<RequestOctets, Target>,
) {
self.middleware_chain = value;
}
}
//--- Default
impl<RequestOctets, Target> Default for Config<RequestOctets, Target>
where
RequestOctets: Octets,
Target: Composer + Default,
{
fn default() -> Self {
Self {
idle_timeout: IDLE_TIMEOUT.default(),
response_write_timeout: RESPONSE_WRITE_TIMEOUT.default(),
max_queued_responses: MAX_QUEUED_RESPONSES.default(),
middleware_chain: MiddlewareBuilder::default().build(),
}
}
}
//--- Clone
impl<RequestOctets, Target> Clone for Config<RequestOctets, Target> {
fn clone(&self) -> Self {
Self {
idle_timeout: self.idle_timeout,
response_write_timeout: self.response_write_timeout,
max_queued_responses: self.max_queued_responses,
middleware_chain: self.middleware_chain.clone(),
}
}
}
//------------ Connection -----------------------------------------------
/// A handler for a single stream connection between client and server.
pub struct Connection<Stream, Buf, Svc>
where
Buf: BufSource,
Svc: Service<Buf::Output>,
{
/// Flag used by the Drop impl to track if the metric count has to be
/// decreased or not.
active: bool,
/// A [`BufSource`] for creating buffers on demand. e.g. to hold response
/// messages.
buf: Buf,
/// User supplied settings that influence our behaviour.
///
/// Note: Some reconfiguration is possible at runtime via
/// [`ServerCommand::Reconfigure`] and [`ServiceFeedback::Reconfigure`].
config: Config<Buf::Output, Svc::Target>,
/// The address of the connected client.
addr: SocketAddr,
/// The incoming connection stream from the client.
///
/// Note: Though this is an Option it should never be None.
stream_rx: Option<ReadHalf<Stream>>,
/// The outgoing connection stream to the client.
stream_tx: WriteHalf<Stream>,
/// The reader for consuming from the queue of responses waiting to be
/// written back to the client.
result_q_rx: mpsc::Receiver<CallResult<Svc::Target>>,
/// The writer for pushing ready responses onto the queue waiting
/// to be written back the client.
result_q_tx: mpsc::Sender<CallResult<Svc::Target>>,
/// A [`Service`] for handling received requests and generating responses.
service: Svc,
/// DNS protocol idle time out tracking.
idle_timer: IdleTimer,
/// [`ServerMetrics`] describing the status of the server.
metrics: Arc<ServerMetrics>,
}
/// Creation
///
impl<Stream, Buf, Svc> Connection<Stream, Buf, Svc>
where
Stream: AsyncRead + AsyncWrite,
Buf: BufSource,
Buf::Output: Octets,
Svc: Service<Buf::Output>,
Svc::Target: Composer + Default,
{
/// Creates a new handler for an accepted stream connection.
#[must_use]
#[allow(dead_code)]
pub fn new(
service: Svc,
buf: Buf,
metrics: Arc<ServerMetrics>,
stream: Stream,
addr: SocketAddr,
) -> Self {
Self::with_config(
service,
buf,
metrics,
stream,
addr,
Config::default(),
)
}
/// Create a new connection handler with a given configuration.
#[must_use]
pub fn with_config(
service: Svc,
buf: Buf,
metrics: Arc<ServerMetrics>,
stream: Stream,
addr: SocketAddr,
config: Config<Buf::Output, Svc::Target>,
) -> Self {
let (stream_rx, stream_tx) = tokio::io::split(stream);
let (result_q_tx, result_q_rx) =
mpsc::channel(config.max_queued_responses);
let idle_timer = IdleTimer::new();
// Place the ReadHalf of the stream into an Option so that we can take
// it out (as we can't clone it and we can't place it into an Arc
// (even though it is Send and Sync) because AsyncRead::poll_read()
// takes Pin<&mut Self> which can't be obtained from an Arc without
// having the only Arc). We want to take it out so that we can use it
// without taking a reference to self as that conflicts with other
// uses of self we have to do while running.
let stream_rx = Some(stream_rx);
Self {
active: false,
buf,
config,
addr,
stream_rx,
stream_tx,
result_q_rx,
result_q_tx,
service,
idle_timer,
metrics,
}
}
}
/// Control
///
impl<Stream, Buf, Svc> Connection<Stream, Buf, Svc>
where
Stream: AsyncRead + AsyncWrite + Send + Sync + 'static,
Buf: BufSource + Send + Sync + Clone + 'static,
Buf::Output: Octets + Send + Sync,
Svc: Service<Buf::Output> + Send + Sync + 'static,
Svc::Target: Send + Composer + Default,
{
/// Start reading requests and writing responses to the stream.
///
/// # Shutdown behaviour
///
/// When the parent server is shutdown (explicitly or via Drop) the child
/// connections will also see the [`ServerCommand::Shutdown`] signal and
/// shutdown and flush any pending writes to the output stream.
///
/// Any requests received after the shutdown signal or requests still
/// in-flight will continue processing and then fail to queue the response
/// for writing.
pub async fn run(
mut self,
command_rx: watch::Receiver<
ServerCommand<ServerConfig<Buf::Output, Svc::Target>>,
>,
) where
Svc::Future: Send,
{
self.metrics.inc_num_connections();
// Flag that we have to decrease the metric count on Drop.
self.active = true;
self.run_until_error(command_rx).await;
}
}
//--- Internal details
impl<Stream, Buf, Svc> Connection<Stream, Buf, Svc>
where
Stream: AsyncRead + AsyncWrite + Send + Sync + 'static,
Buf: BufSource + Send + Sync + Clone + 'static,
Buf::Output: Octets + Send + Sync,
Svc: Service<Buf::Output> + Send + Sync + 'static,
Svc::Future: Send,
Svc::Target: Send + Composer + Default,
{
/// Connection handler main loop.
async fn run_until_error(
mut self,
mut command_rx: watch::Receiver<
ServerCommand<ServerConfig<Buf::Output, Svc::Target>>,
>,
) {
// SAFETY: This unwrap is safe because we always put a Some value into
// self.stream_rx in [`Self::with_config`] above (and thus also in
// [`Self::new`] which calls [`Self::with_config`]).
let stream_rx = self.stream_rx.take().unwrap();
let mut dns_msg_receiver =
DnsMessageReceiver::new(self.buf.clone(), stream_rx);
'outer: loop {
// Create a read future that will survive when other
// tokio::select! branches resolve before the branch awaiting this
// future resolves. This ensures that in-progress non-cancel-safe
// reads do not get cancelled. This works because it
// avoids creating a new future each time as would happen if we
// called transceive() in a tokio::select! branch.
let msg_recv = dns_msg_receiver.recv();
tokio::pin!(msg_recv);
'inner: loop {
let res = tokio::select! {
biased;
res = command_rx.changed() => {
self.process_server_command(res, &mut command_rx)
}
res = self.result_q_rx.recv() => {
self.process_queued_result(res).await
}
_ = sleep_until(self.idle_timer.idle_timeout_deadline(self.config.idle_timeout)) => {
self.process_dns_idle_timeout()
}
res = &mut msg_recv => {
let res = self.process_read_request(res).await;
if res.is_ok() {
// Set up to receive another message
break 'inner;
} else {
res
}
}
};
if let Err(err) = res {
match err {
ConnectionEvent::DisconnectWithoutFlush => {
break 'outer;
}
ConnectionEvent::DisconnectWithFlush => {
self.flush_write_queue().await;
break 'outer;
}
ConnectionEvent::ServiceError(err) => {
error!("Service error: {}", err);
}
}
}
}
}
trace!("Shutting down the write stream.");
if let Err(err) = self.stream_tx.shutdown().await {
warn!("Error while shutting down the write stream: {err}");
}
trace!("Connection terminated.");
#[cfg(test)]
if dns_msg_receiver.cancelled() {
panic!("Async not-cancel-safe code was cancelled");
}
}
/// Decide what to do with a received [`ServerCommand`].
fn process_server_command(
&mut self,
res: Result<(), watch::error::RecvError>,
command_rx: &mut watch::Receiver<
ServerCommand<ServerConfig<Buf::Output, Svc::Target>>,
>,
) -> Result<(), ConnectionEvent> {
// If the parent server no longer exists but was not cleanly shutdown
// then the command channel will be closed and attempting to check for
// a new command will fail. Advise the caller to break the connection
// and cleanup if such a problem occurs.
res.map_err(|_err| ConnectionEvent::DisconnectWithFlush)?;
// Get the changed command.
let lock = command_rx.borrow_and_update();
let command = lock.deref();
// And process it.
match command {
ServerCommand::Init => {
// The initial "Init" value in the watch channel is never
// actually seen because changed() is required to return true
// before we call borrow_and_update() but the initial value in
// the channel, Init, is not considered a "change". So the
// only way to end up here would be if we somehow wrongly
// placed another ServerCommand::Init value into the watch
// channel after the initial one.
unreachable!()
}
ServerCommand::CloseConnection => {
// TODO: Should we flush in this case or not?
return Err(ConnectionEvent::DisconnectWithFlush);
}
ServerCommand::Reconfigure(ServerConfig {
connection_config:
Config {
idle_timeout,
response_write_timeout,
max_queued_responses: _,
middleware_chain: _,
},
.. // Ignore the Server specific configuration settings
}) => {
// Support RFC 7828 "The edns-tcp-keepalive EDNS0 Option".
// This cannot be done by the caller as it requires knowing
// (a) when the last message was received and (b) when all
// pending messages have been sent, neither of which is known
// to the caller. However we also don't want to parse and
// understand DNS messages in this layer, it is left to the
// caller to process received messages and construct
// appropriate responses. If the caller detects an EDNS0
// edns-tcp-keepalive option it can use this reconfigure
// mechanism to signal to us that we should adjust the point
// at which we will consider the connectin to be idle and thus
// potentially worthy of timing out.
debug!("Server connection timeout reconfigured to {idle_timeout:?}");
self.config.idle_timeout = *idle_timeout;
self.config.response_write_timeout = *response_write_timeout;
}
ServerCommand::Shutdown => {
// The parent server has been shutdown. Close this connection
// but ensure that we write any pending responses to the
// stream first.
//
// TODO: Should we also wait for any in-flight requests to
// complete before shutting down? And if so how should we
// respond to any requests received in the meantime? Should we
// even stop reading from the stream?
return Err(ConnectionEvent::DisconnectWithFlush);
}
}
Ok(())
}
/// Stop queueing new responses and process those already in the queue.
async fn flush_write_queue(&mut self) {
debug!("Flushing connection write queue.");
// Stop accepting new response messages (should we check for in-flight
// messages that haven't generated a response yet but should be
// allowed to do so?) so that we can flush the write queue and exit
// this connection handler.
trace!("Stop queueing up new results.");
self.result_q_rx.close();
trace!("Process already queued results.");
while let Some(call_result) = self.result_q_rx.recv().await {
trace!("Processing queued result.");
if let Err(err) =
self.process_queued_result(Some(call_result)).await
{
warn!("Error while processing queued result: {err}");
} else {
trace!("Result processed");
}
}
debug!("Connection write queue flush complete.");
}
/// Process a single queued response.
async fn process_queued_result(
&mut self,
call_result: Option<CallResult<Svc::Target>>,
) -> Result<(), ConnectionEvent> {
// If we failed to read the results of requests processed by the
// service because the queue holding those results is empty and can no
// longer be read from, then there is no point continuing to read from
// the input stream because we will not be able to access the result
// of processing the request. I'm not sure when this could happen,
// perhaps if we were dropped?
let Some(call_result) = call_result else {
return Err(ConnectionEvent::DisconnectWithFlush);
};
let (response, feedback) = call_result.into_inner();
if let Some(feedback) = feedback {
self.process_service_feedback(feedback).await;
}
if let Some(response) = response {
self.write_response_to_stream(response.finish()).await;
}
Ok(())
}
/// Write a response back to the caller over the network stream.
async fn write_response_to_stream(
&mut self,
msg: StreamTarget<Svc::Target>,
) {
if enabled!(Level::TRACE) {
let bytes = msg.as_dgram_slice();
let pcap_text = to_pcap_text(bytes, bytes.len());
trace!(addr = %self.addr, pcap_text, "Sending response");
}
match timeout(
self.config.response_write_timeout,
self.stream_tx.write_all(msg.as_stream_slice()),
)
.await
{
Err(_) => {
error!(
"Write timed out (>{:?})",
self.config.response_write_timeout
);
// TODO: Push it to the back of the queue to retry it?
}
Ok(Err(err)) => {
error!("Write error: {err}");
}
Ok(Ok(_)) => {
self.metrics.inc_num_sent_responses();
}
}
self.metrics.dec_num_pending_writes();
if self.result_q_tx.capacity() == self.result_q_tx.max_capacity() {
self.idle_timer.response_queue_emptied();
}
}
/// Decide what to do with received [`ServiceFeedback`].
async fn process_service_feedback(&mut self, cmd: ServiceFeedback) {
match cmd {
ServiceFeedback::Reconfigure { idle_timeout } => {
if let Some(idle_timeout) = idle_timeout {
debug!(
"Reconfigured connection timeout to {idle_timeout:?}"
);
self.config.idle_timeout = idle_timeout;
}
}
}
}
/// Implemnt DNS rules regarding timing out of idle connections.
///
/// Disconnects the current connection of the timer is expired, flushing
/// pending responses first.
fn process_dns_idle_timeout(&self) -> Result<(), ConnectionEvent> {
// DNS idle timeout elapsed, or was it reset?
if self
.idle_timer
.idle_timeout_expired(self.config.idle_timeout)
{
Err(ConnectionEvent::DisconnectWithoutFlush)
} else {
Ok(())
}
}
/// Process a received request message.
async fn process_read_request(
&mut self,
res: Result<Buf::Output, ConnectionEvent>,
) -> Result<(), ConnectionEvent>
where
Svc::Future: Send,
{
res.and_then(|msg| {
let received_at = Instant::now();
if enabled!(Level::TRACE) {
let pcap_text = to_pcap_text(&msg, msg.as_ref().len());
trace!(addr = %self.addr, pcap_text, "Received message");
}
self.metrics.inc_num_received_requests();
// Message received, reset the DNS idle timer
self.idle_timer.full_msg_received();
// Process the received message
self.process_request(
msg,
received_at,
self.addr,
self.config.middleware_chain.clone(),
&self.service,
self.metrics.clone(),
self.result_q_tx.clone(),
)
.map_err(ConnectionEvent::ServiceError)
})
}
}
//--- Drop
impl<Stream, Buf, Svc> Drop for Connection<Stream, Buf, Svc>
where
Buf: BufSource,
Svc: Service<Buf::Output>,
{
fn drop(&mut self) {
if self.active {
self.active = false;
self.metrics.dec_num_connections();
}
}
}
//--- CommonMessageFlow
impl<Stream, Buf, Svc> CommonMessageFlow<Buf, Svc>
for Connection<Stream, Buf, Svc>
where
Buf: BufSource,
Buf::Output: Octets + Send + Sync + 'static,
Svc: Service<Buf::Output> + Send + Sync + 'static,
Svc::Target: Send,
{
type Meta = Sender<CallResult<Svc::Target>>;
/// Add information to the request that relates to the type of server we
/// are and our state where relevant.
fn add_context_to_request(
&self,
request: Message<Buf::Output>,
received_at: Instant,
addr: SocketAddr,
) -> Request<Buf::Output> {
let ctx = NonUdpTransportContext::new(Some(self.config.idle_timeout));
let ctx = TransportSpecificContext::NonUdp(ctx);
Request::new(addr, received_at, request, ctx)
}
/// Process the result from the middleware -> service -> middleware call
/// tree.
fn process_call_result(
_request: &Request<Buf::Output>,
call_result: CallResult<Svc::Target>,
tx: Self::Meta,
metrics: Arc<ServerMetrics>,
) {
// We can't send in a spawned async task as then we would just
// accumlate tasks even if the target queue is full. We can't call
// `tx.blocking_send()` as that would block the Tokio runtime. So
// instead we try and send and if that fails because the queue is full
// then we abort.
match tx.try_send(call_result) {
Ok(()) => {
metrics.set_num_pending_writes(
tx.max_capacity() - tx.capacity(),
);
}
Err(TrySendError::Closed(_msg)) => {
// TODO: How should we properly communicate this to the operator?
error!("Unable to queue message for sending: server is shutting down.");
}
Err(TrySendError::Full(_msg)) => {
// TODO: How should we properly communicate this to the operator?
error!("Unable to queue message for sending: queue is full.");
}
}
}
}
//----------- DnsMessageReceiver ---------------------------------------------
/// The [`DnsMessageReceiver`] state machine.
#[derive(Copy, Clone, Debug, PartialEq, Eq)]
enum Status {
/// Initial state.
New,
/// Waiting to receive a DNS message header.
WaitingForMessageHeader,
/// Waiting to receive a DNS message body.
WaitingForMessageBody,
/// A full DNS message header and body has been received.
MessageReceived,
}
/// A cancel safe DNS message receiver.
///
/// If the message is received in bits, e.g. header then body, this receiver
/// ensures that any part of the request already received is not lost if the
/// read operation is cancelled by Tokio and then a new read operation is
/// started.
struct DnsMessageReceiver<Stream, Buf> {
/// A buffer to record the total expected size of the message currently
/// being received. DNS TCP streams preceed the DNS message by bytes
/// indicating the length of the message that follows.
msg_size_buf: [u8; 2],
/// A [`BufSource`] for creating buffers on demand. e.g. to hold response
/// messages.
buf: Buf,
/// The incoming connection stream from the client.
stream_rx: ReadHalf<Stream>,
/// Our state machine state.
status: Status,
#[cfg(test)]
/// A flag used only during testing that will be set if a read operation
/// is started but detects that a previous read operation didn't complete,
/// i.e. the async operation was cancelled.
cancelled: bool,
}
impl<Stream, Buf> DnsMessageReceiver<Stream, Buf>
where
Stream: AsyncRead + AsyncWrite + Send + Sync + 'static,
Buf: BufSource + Send + Sync + 'static + Clone,
Buf::Output: Send + Sync + 'static,
{
/// Creates a new message receiver.
fn new(buf: Buf, stream_rx: ReadHalf<Stream>) -> Self {
Self {
msg_size_buf: [0; 2],
buf,
stream_rx,
status: Status::New,
#[cfg(test)]
cancelled: false,
}
}
#[cfg(test)]
/// Was a read operation using this receiver cancelled at some point?
pub fn cancelled(&self) -> bool {
self.cancelled
}
/// Receive a single DNS message.
///
/// # Cancel safety
///
/// This function is NOT cancel safe.
pub async fn recv(&mut self) -> Result<Buf::Output, ConnectionEvent> {
#[cfg(test)]
if self.status == Status::WaitingForMessageBody {
self.cancelled = true;
}
self.status = Status::WaitingForMessageHeader;
Self::recv_n_bytes(&mut self.stream_rx, &mut self.msg_size_buf)
.await?;
let msg_len = u16::from_be_bytes(self.msg_size_buf) as usize;
let mut msg_buf = self.buf.create_sized(msg_len);
self.status = Status::WaitingForMessageBody;
Self::recv_n_bytes(&mut self.stream_rx, &mut msg_buf).await?;
self.status = Status::MessageReceived;
Ok(msg_buf)
}
/// Receive exactly as many bytes as the given buffer can hold.
///
/// # Cancel safety
///
/// This function is NOT cancel safe.
async fn recv_n_bytes<T: AsMut<[u8]>>(
stream_rx: &mut ReadHalf<Stream>,
buf: &mut T,
) -> Result<(), ConnectionEvent> {
loop {
match stream_rx.read_exact(buf.as_mut()).await {
// The stream read succeeded. Return to the caller
// so that it can process the bytes written to the
// buffer.
Ok(_size) => return Ok(()),
Err(err) => match Self::process_io_error(err) {
ControlFlow::Continue(_) => continue,
ControlFlow::Break(err) => return Err(err),
},
}
}
}
/// Handle I/O errors by deciding whether to log them, and whethr to
/// continue or abort.
#[must_use]
fn process_io_error(err: io::Error) -> ControlFlow<ConnectionEvent> {
match err.kind() {
io::ErrorKind::UnexpectedEof => {
// The client disconnected. Per RFC 7766 6.2.4 pending
// responses MUST NOT be sent to the client.
ControlFlow::Break(ConnectionEvent::DisconnectWithoutFlush)
}
io::ErrorKind::TimedOut | io::ErrorKind::Interrupted => {
// These errors might be recoverable, try again.
ControlFlow::Continue(())
}
_ => {
// Everything else is either unrecoverable or unknown to us at
// the time of writing and so we can't guess how to handle it,
// so abort.
error!("I/O error: {}", err);
ControlFlow::Break(ConnectionEvent::DisconnectWithoutFlush)
}
}
}
}
//------------ ConnectionEvent -----------------------------------------------
/// An event that occurred while the connection handler was handling the
/// connection.
enum ConnectionEvent {
/// RFC 7766 6.2.4 "Under normal operation DNS clients typically initiate
/// connection closing on idle connections; however, DNS servers can close
/// the connection if the idle timeout set by local policy is exceeded.
/// Also, connections can be closed by either end under unusual conditions
/// such as defending against an attack or system failure reboot."
///
/// And: RFC 7766 3 "A DNS server considers an established DNS-over-TCP
/// session to be idle when it has sent responses to all the queries it
/// has received on that connection."
DisconnectWithoutFlush,
/// RFC 7766 6.2.3 "If a DNS server finds that a DNS client has closed a
/// TCP session (or if the session has been otherwise interrupted) before
/// all pending responses have been sent, then the server MUST NOT attempt
/// to send those responses. Of course, the DNS server MAY cache those
/// responses."
DisconnectWithFlush,
/// A [`Service`] specific error occurred while the service was processing
/// a request message.
ServiceError(ServiceError),
}
//--- Display
impl Display for ConnectionEvent {
fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
match self {
ConnectionEvent::DisconnectWithoutFlush => {
write!(f, "Disconnect without flush")
}
ConnectionEvent::DisconnectWithFlush => {
write!(f, "Disconnect with flush")
}
ConnectionEvent::ServiceError(err) => {
write!(f, "Service error: {err}")
}
}
}
}
//------------ IdleTimer -----------------------------------------------------
/// RFC 7766 section 6.2.3 / RFC 7828 section 3 idle time out tracking.
pub struct IdleTimer {
/// The instant when the timer was last reset.
idle_timer_reset_at: Instant,
}
impl IdleTimer {
/// Creates a new idle timer.
///
/// Sets the last reset instant to now.
#[must_use]
fn new() -> Self {
Self {
idle_timer_reset_at: Instant::now(),
}
}
/// How long from now should this connection be timed out?
///
/// When we (will) have been sat idle for longer than the configured idle
/// timeout for this connection.
#[must_use]
pub fn idle_timeout_deadline(&self, timeout: Duration) -> Instant {
self.idle_timer_reset_at
.checked_add(timeout)
.unwrap_or_else(|| {
warn!("Unable to reset idle timer: value out of bounds");
Instant::now()
})
}
/// Did the idle timeout expire?
#[must_use]
pub fn idle_timeout_expired(&self, timeout: Duration) -> bool {
self.idle_timeout_deadline(timeout) <= Instant::now()
}
/// Reset the idle timer to now.
fn reset_idle_timer(&mut self) {
self.idle_timer_reset_at = Instant::now();
}
/// Act on the fact that a complete DNS message was received.
///
/// Per RFC 7766 this resets the idle timer.
fn full_msg_received(&mut self) {
// RFC 7766 6.2.3: "DNS messages delivered over TCP might arrive in
// multiple segments. A DNS server that resets its idle timeout after
// receiving a single segment might be vulnerable to a "slow-read
// attack". For this reason, servers SHOULD reset the idle timeout on
// the receipt of a full DNS message, rather than on receipt of any
// part of a DNS message."
self.reset_idle_timer()
}
/// Act on the fact that the connection handler caught up with processing
/// all queued responses.
///
/// Per RFC 7766 this resets the idle timer.
fn response_queue_emptied(&mut self) {
// RFC 7766 3: "A DNS server considers an established DNS-over-TCP
// session to be idle when it has sent responses to all the queries it
// has received on that connection."
self.reset_idle_timer()
}
}