tokio/net/unix/stream.rs
1use crate::io::{AsyncRead, AsyncWrite, Interest, PollEvented, ReadBuf, Ready};
2use crate::net::unix::split::{split, ReadHalf, WriteHalf};
3use crate::net::unix::split_owned::{split_owned, OwnedReadHalf, OwnedWriteHalf};
4use crate::net::unix::ucred::{self, UCred};
5use crate::net::unix::SocketAddr;
6
7use std::fmt;
8use std::future::poll_fn;
9use std::io::{self, Read, Write};
10use std::net::Shutdown;
11#[cfg(target_os = "android")]
12use std::os::android::net::SocketAddrExt;
13#[cfg(target_os = "linux")]
14use std::os::linux::net::SocketAddrExt;
15#[cfg(any(target_os = "linux", target_os = "android"))]
16use std::os::unix::ffi::OsStrExt;
17use std::os::unix::io::{AsFd, AsRawFd, BorrowedFd, FromRawFd, IntoRawFd, RawFd};
18use std::os::unix::net::{self, SocketAddr as StdSocketAddr};
19use std::path::Path;
20use std::pin::Pin;
21use std::task::{Context, Poll};
22
23cfg_io_util! {
24 use bytes::BufMut;
25}
26
27cfg_net_unix! {
28 /// A structure representing a connected Unix socket.
29 ///
30 /// This socket can be connected directly with [`UnixStream::connect`] or accepted
31 /// from a listener with [`UnixListener::accept`]. Additionally, a pair of
32 /// anonymous Unix sockets can be created with `UnixStream::pair`.
33 ///
34 /// To shut down the stream in the write direction, you can call the
35 /// [`shutdown()`] method. This will cause the other peer to receive a read of
36 /// length 0, indicating that no more data will be sent. This only closes
37 /// the stream in one direction.
38 ///
39 /// [`shutdown()`]: fn@crate::io::AsyncWriteExt::shutdown
40 /// [`UnixListener::accept`]: crate::net::UnixListener::accept
41 #[cfg_attr(docsrs, doc(alias = "uds"))]
42 pub struct UnixStream {
43 io: PollEvented<mio::net::UnixStream>,
44 }
45}
46
47impl UnixStream {
48 pub(crate) async fn connect_mio(sys: mio::net::UnixStream) -> io::Result<UnixStream> {
49 let stream = UnixStream::new(sys)?;
50
51 // Once we've connected, wait for the stream to be writable as
52 // that's when the actual connection has been initiated. Once we're
53 // writable we check for `take_socket_error` to see if the connect
54 // actually hit an error or not.
55 //
56 // If all that succeeded then we ship everything on up.
57 poll_fn(|cx| stream.io.registration().poll_write_ready(cx)).await?;
58
59 if let Some(e) = stream.io.take_error()? {
60 return Err(e);
61 }
62
63 Ok(stream)
64 }
65
66 /// Connects to the socket named by `path`.
67 ///
68 /// This function will create a new Unix socket and connect to the path
69 /// specified, associating the returned stream with the default event loop's
70 /// handle.
71 pub async fn connect<P>(path: P) -> io::Result<UnixStream>
72 where
73 P: AsRef<Path>,
74 {
75 // On linux, abstract socket paths need to be considered.
76 #[cfg(any(target_os = "linux", target_os = "android"))]
77 let addr = {
78 let os_str_bytes = path.as_ref().as_os_str().as_bytes();
79 if os_str_bytes.starts_with(b"\0") {
80 StdSocketAddr::from_abstract_name(&os_str_bytes[1..])?
81 } else {
82 StdSocketAddr::from_pathname(path)?
83 }
84 };
85 #[cfg(not(any(target_os = "linux", target_os = "android")))]
86 let addr = StdSocketAddr::from_pathname(path)?;
87
88 let stream = mio::net::UnixStream::connect_addr(&addr)?;
89 let stream = UnixStream::new(stream)?;
90
91 poll_fn(|cx| stream.io.registration().poll_write_ready(cx)).await?;
92
93 if let Some(e) = stream.io.take_error()? {
94 return Err(e);
95 }
96
97 Ok(stream)
98 }
99
100 /// Waits for any of the requested ready states.
101 ///
102 /// This function is usually paired with `try_read()` or `try_write()`. It
103 /// can be used to concurrently read / write to the same socket on a single
104 /// task without splitting the socket.
105 ///
106 /// The function may complete without the socket being ready. This is a
107 /// false-positive and attempting an operation will return with
108 /// `io::ErrorKind::WouldBlock`. The function can also return with an empty
109 /// [`Ready`] set, so you should always check the returned value and possibly
110 /// wait again if the requested states are not set.
111 ///
112 /// # Cancel safety
113 ///
114 /// This method is cancel safe. Once a readiness event occurs, the method
115 /// will continue to return immediately until the readiness event is
116 /// consumed by an attempt to read or write that fails with `WouldBlock` or
117 /// `Poll::Pending`.
118 ///
119 /// # Examples
120 ///
121 /// Concurrently read and write to the stream on the same task without
122 /// splitting.
123 ///
124 /// ```no_run
125 /// use tokio::io::Interest;
126 /// use tokio::net::UnixStream;
127 /// use std::error::Error;
128 /// use std::io;
129 ///
130 /// #[tokio::main]
131 /// async fn main() -> Result<(), Box<dyn Error>> {
132 /// let dir = tempfile::tempdir().unwrap();
133 /// let bind_path = dir.path().join("bind_path");
134 /// let stream = UnixStream::connect(bind_path).await?;
135 ///
136 /// loop {
137 /// let ready = stream.ready(Interest::READABLE | Interest::WRITABLE).await?;
138 ///
139 /// if ready.is_readable() {
140 /// let mut data = vec![0; 1024];
141 /// // Try to read data, this may still fail with `WouldBlock`
142 /// // if the readiness event is a false positive.
143 /// match stream.try_read(&mut data) {
144 /// Ok(n) => {
145 /// println!("read {} bytes", n);
146 /// }
147 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
148 /// continue;
149 /// }
150 /// Err(e) => {
151 /// return Err(e.into());
152 /// }
153 /// }
154 ///
155 /// }
156 ///
157 /// if ready.is_writable() {
158 /// // Try to write data, this may still fail with `WouldBlock`
159 /// // if the readiness event is a false positive.
160 /// match stream.try_write(b"hello world") {
161 /// Ok(n) => {
162 /// println!("write {} bytes", n);
163 /// }
164 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
165 /// continue;
166 /// }
167 /// Err(e) => {
168 /// return Err(e.into());
169 /// }
170 /// }
171 /// }
172 /// }
173 /// }
174 /// ```
175 pub async fn ready(&self, interest: Interest) -> io::Result<Ready> {
176 let event = self.io.registration().readiness(interest).await?;
177 Ok(event.ready)
178 }
179
180 /// Waits for the socket to become readable.
181 ///
182 /// This function is equivalent to `ready(Interest::READABLE)` and is usually
183 /// paired with `try_read()`.
184 ///
185 /// # Cancel safety
186 ///
187 /// This method is cancel safe. Once a readiness event occurs, the method
188 /// will continue to return immediately until the readiness event is
189 /// consumed by an attempt to read that fails with `WouldBlock` or
190 /// `Poll::Pending`.
191 ///
192 /// # Examples
193 ///
194 /// ```no_run
195 /// use tokio::net::UnixStream;
196 /// use std::error::Error;
197 /// use std::io;
198 ///
199 /// #[tokio::main]
200 /// async fn main() -> Result<(), Box<dyn Error>> {
201 /// // Connect to a peer
202 /// let dir = tempfile::tempdir().unwrap();
203 /// let bind_path = dir.path().join("bind_path");
204 /// let stream = UnixStream::connect(bind_path).await?;
205 ///
206 /// let mut msg = vec![0; 1024];
207 ///
208 /// loop {
209 /// // Wait for the socket to be readable
210 /// stream.readable().await?;
211 ///
212 /// // Try to read data, this may still fail with `WouldBlock`
213 /// // if the readiness event is a false positive.
214 /// match stream.try_read(&mut msg) {
215 /// Ok(n) => {
216 /// msg.truncate(n);
217 /// break;
218 /// }
219 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
220 /// continue;
221 /// }
222 /// Err(e) => {
223 /// return Err(e.into());
224 /// }
225 /// }
226 /// }
227 ///
228 /// println!("GOT = {:?}", msg);
229 /// Ok(())
230 /// }
231 /// ```
232 pub async fn readable(&self) -> io::Result<()> {
233 self.ready(Interest::READABLE).await?;
234 Ok(())
235 }
236
237 /// Polls for read readiness.
238 ///
239 /// If the unix stream is not currently ready for reading, this method will
240 /// store a clone of the `Waker` from the provided `Context`. When the unix
241 /// stream becomes ready for reading, `Waker::wake` will be called on the
242 /// waker.
243 ///
244 /// Note that on multiple calls to `poll_read_ready` or `poll_read`, only
245 /// the `Waker` from the `Context` passed to the most recent call is
246 /// scheduled to receive a wakeup. (However, `poll_write_ready` retains a
247 /// second, independent waker.)
248 ///
249 /// This function is intended for cases where creating and pinning a future
250 /// via [`readable`] is not feasible. Where possible, using [`readable`] is
251 /// preferred, as this supports polling from multiple tasks at once.
252 ///
253 /// # Return value
254 ///
255 /// The function returns:
256 ///
257 /// * `Poll::Pending` if the unix stream is not ready for reading.
258 /// * `Poll::Ready(Ok(()))` if the unix stream is ready for reading.
259 /// * `Poll::Ready(Err(e))` if an error is encountered.
260 ///
261 /// # Errors
262 ///
263 /// This function may encounter any standard I/O error except `WouldBlock`.
264 ///
265 /// [`readable`]: method@Self::readable
266 pub fn poll_read_ready(&self, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
267 self.io.registration().poll_read_ready(cx).map_ok(|_| ())
268 }
269
270 /// Try to read data from the stream into the provided buffer, returning how
271 /// many bytes were read.
272 ///
273 /// Receives any pending data from the socket but does not wait for new data
274 /// to arrive. On success, returns the number of bytes read. Because
275 /// `try_read()` is non-blocking, the buffer does not have to be stored by
276 /// the async task and can exist entirely on the stack.
277 ///
278 /// Usually, [`readable()`] or [`ready()`] is used with this function.
279 ///
280 /// [`readable()`]: UnixStream::readable()
281 /// [`ready()`]: UnixStream::ready()
282 ///
283 /// # Return
284 ///
285 /// If data is successfully read, `Ok(n)` is returned, where `n` is the
286 /// number of bytes read. If `n` is `0`, then it can indicate one of two scenarios:
287 ///
288 /// 1. The stream's read half is closed and will no longer yield data.
289 /// 2. The specified buffer was 0 bytes in length.
290 ///
291 /// If the stream is not ready to read data,
292 /// `Err(io::ErrorKind::WouldBlock)` is returned.
293 ///
294 /// # Examples
295 ///
296 /// ```no_run
297 /// use tokio::net::UnixStream;
298 /// use std::error::Error;
299 /// use std::io;
300 ///
301 /// #[tokio::main]
302 /// async fn main() -> Result<(), Box<dyn Error>> {
303 /// // Connect to a peer
304 /// let dir = tempfile::tempdir().unwrap();
305 /// let bind_path = dir.path().join("bind_path");
306 /// let stream = UnixStream::connect(bind_path).await?;
307 ///
308 /// loop {
309 /// // Wait for the socket to be readable
310 /// stream.readable().await?;
311 ///
312 /// // Creating the buffer **after** the `await` prevents it from
313 /// // being stored in the async task.
314 /// let mut buf = [0; 4096];
315 ///
316 /// // Try to read data, this may still fail with `WouldBlock`
317 /// // if the readiness event is a false positive.
318 /// match stream.try_read(&mut buf) {
319 /// Ok(0) => break,
320 /// Ok(n) => {
321 /// println!("read {} bytes", n);
322 /// }
323 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
324 /// continue;
325 /// }
326 /// Err(e) => {
327 /// return Err(e.into());
328 /// }
329 /// }
330 /// }
331 ///
332 /// Ok(())
333 /// }
334 /// ```
335 pub fn try_read(&self, buf: &mut [u8]) -> io::Result<usize> {
336 self.io
337 .registration()
338 .try_io(Interest::READABLE, || (&*self.io).read(buf))
339 }
340
341 /// Tries to read data from the stream into the provided buffers, returning
342 /// how many bytes were read.
343 ///
344 /// Data is copied to fill each buffer in order, with the final buffer
345 /// written to possibly being only partially filled. This method behaves
346 /// equivalently to a single call to [`try_read()`] with concatenated
347 /// buffers.
348 ///
349 /// Receives any pending data from the socket but does not wait for new data
350 /// to arrive. On success, returns the number of bytes read. Because
351 /// `try_read_vectored()` is non-blocking, the buffer does not have to be
352 /// stored by the async task and can exist entirely on the stack.
353 ///
354 /// Usually, [`readable()`] or [`ready()`] is used with this function.
355 ///
356 /// [`try_read()`]: UnixStream::try_read()
357 /// [`readable()`]: UnixStream::readable()
358 /// [`ready()`]: UnixStream::ready()
359 ///
360 /// # Return
361 ///
362 /// If data is successfully read, `Ok(n)` is returned, where `n` is the
363 /// number of bytes read. `Ok(0)` indicates the stream's read half is closed
364 /// and will no longer yield data. If the stream is not ready to read data
365 /// `Err(io::ErrorKind::WouldBlock)` is returned.
366 ///
367 /// # Examples
368 ///
369 /// ```no_run
370 /// use tokio::net::UnixStream;
371 /// use std::error::Error;
372 /// use std::io::{self, IoSliceMut};
373 ///
374 /// #[tokio::main]
375 /// async fn main() -> Result<(), Box<dyn Error>> {
376 /// // Connect to a peer
377 /// let dir = tempfile::tempdir().unwrap();
378 /// let bind_path = dir.path().join("bind_path");
379 /// let stream = UnixStream::connect(bind_path).await?;
380 ///
381 /// loop {
382 /// // Wait for the socket to be readable
383 /// stream.readable().await?;
384 ///
385 /// // Creating the buffer **after** the `await` prevents it from
386 /// // being stored in the async task.
387 /// let mut buf_a = [0; 512];
388 /// let mut buf_b = [0; 1024];
389 /// let mut bufs = [
390 /// IoSliceMut::new(&mut buf_a),
391 /// IoSliceMut::new(&mut buf_b),
392 /// ];
393 ///
394 /// // Try to read data, this may still fail with `WouldBlock`
395 /// // if the readiness event is a false positive.
396 /// match stream.try_read_vectored(&mut bufs) {
397 /// Ok(0) => break,
398 /// Ok(n) => {
399 /// println!("read {} bytes", n);
400 /// }
401 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
402 /// continue;
403 /// }
404 /// Err(e) => {
405 /// return Err(e.into());
406 /// }
407 /// }
408 /// }
409 ///
410 /// Ok(())
411 /// }
412 /// ```
413 pub fn try_read_vectored(&self, bufs: &mut [io::IoSliceMut<'_>]) -> io::Result<usize> {
414 self.io
415 .registration()
416 .try_io(Interest::READABLE, || (&*self.io).read_vectored(bufs))
417 }
418
419 cfg_io_util! {
420 /// Tries to read data from the stream into the provided buffer, advancing the
421 /// buffer's internal cursor, returning how many bytes were read.
422 ///
423 /// Receives any pending data from the socket but does not wait for new data
424 /// to arrive. On success, returns the number of bytes read. Because
425 /// `try_read_buf()` is non-blocking, the buffer does not have to be stored by
426 /// the async task and can exist entirely on the stack.
427 ///
428 /// Usually, [`readable()`] or [`ready()`] is used with this function.
429 ///
430 /// [`readable()`]: UnixStream::readable()
431 /// [`ready()`]: UnixStream::ready()
432 ///
433 /// # Return
434 ///
435 /// If data is successfully read, `Ok(n)` is returned, where `n` is the
436 /// number of bytes read. `Ok(0)` indicates the stream's read half is closed
437 /// and will no longer yield data. If the stream is not ready to read data
438 /// `Err(io::ErrorKind::WouldBlock)` is returned.
439 ///
440 /// # Examples
441 ///
442 /// ```no_run
443 /// use tokio::net::UnixStream;
444 /// use std::error::Error;
445 /// use std::io;
446 ///
447 /// #[tokio::main]
448 /// async fn main() -> Result<(), Box<dyn Error>> {
449 /// // Connect to a peer
450 /// let dir = tempfile::tempdir().unwrap();
451 /// let bind_path = dir.path().join("bind_path");
452 /// let stream = UnixStream::connect(bind_path).await?;
453 ///
454 /// loop {
455 /// // Wait for the socket to be readable
456 /// stream.readable().await?;
457 ///
458 /// let mut buf = Vec::with_capacity(4096);
459 ///
460 /// // Try to read data, this may still fail with `WouldBlock`
461 /// // if the readiness event is a false positive.
462 /// match stream.try_read_buf(&mut buf) {
463 /// Ok(0) => break,
464 /// Ok(n) => {
465 /// println!("read {} bytes", n);
466 /// }
467 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
468 /// continue;
469 /// }
470 /// Err(e) => {
471 /// return Err(e.into());
472 /// }
473 /// }
474 /// }
475 ///
476 /// Ok(())
477 /// }
478 /// ```
479 pub fn try_read_buf<B: BufMut>(&self, buf: &mut B) -> io::Result<usize> {
480 self.io.registration().try_io(Interest::READABLE, || {
481 use std::io::Read;
482
483 let dst = buf.chunk_mut();
484 let dst =
485 unsafe { &mut *(dst as *mut _ as *mut [std::mem::MaybeUninit<u8>] as *mut [u8]) };
486
487 // Safety: We trust `UnixStream::read` to have filled up `n` bytes in the
488 // buffer.
489 let n = (&*self.io).read(dst)?;
490
491 unsafe {
492 buf.advance_mut(n);
493 }
494
495 Ok(n)
496 })
497 }
498 }
499
500 /// Waits for the socket to become writable.
501 ///
502 /// This function is equivalent to `ready(Interest::WRITABLE)` and is usually
503 /// paired with `try_write()`.
504 ///
505 /// # Cancel safety
506 ///
507 /// This method is cancel safe. Once a readiness event occurs, the method
508 /// will continue to return immediately until the readiness event is
509 /// consumed by an attempt to write that fails with `WouldBlock` or
510 /// `Poll::Pending`.
511 ///
512 /// # Examples
513 ///
514 /// ```no_run
515 /// use tokio::net::UnixStream;
516 /// use std::error::Error;
517 /// use std::io;
518 ///
519 /// #[tokio::main]
520 /// async fn main() -> Result<(), Box<dyn Error>> {
521 /// // Connect to a peer
522 /// let dir = tempfile::tempdir().unwrap();
523 /// let bind_path = dir.path().join("bind_path");
524 /// let stream = UnixStream::connect(bind_path).await?;
525 ///
526 /// loop {
527 /// // Wait for the socket to be writable
528 /// stream.writable().await?;
529 ///
530 /// // Try to write data, this may still fail with `WouldBlock`
531 /// // if the readiness event is a false positive.
532 /// match stream.try_write(b"hello world") {
533 /// Ok(n) => {
534 /// break;
535 /// }
536 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
537 /// continue;
538 /// }
539 /// Err(e) => {
540 /// return Err(e.into());
541 /// }
542 /// }
543 /// }
544 ///
545 /// Ok(())
546 /// }
547 /// ```
548 pub async fn writable(&self) -> io::Result<()> {
549 self.ready(Interest::WRITABLE).await?;
550 Ok(())
551 }
552
553 /// Polls for write readiness.
554 ///
555 /// If the unix stream is not currently ready for writing, this method will
556 /// store a clone of the `Waker` from the provided `Context`. When the unix
557 /// stream becomes ready for writing, `Waker::wake` will be called on the
558 /// waker.
559 ///
560 /// Note that on multiple calls to `poll_write_ready` or `poll_write`, only
561 /// the `Waker` from the `Context` passed to the most recent call is
562 /// scheduled to receive a wakeup. (However, `poll_read_ready` retains a
563 /// second, independent waker.)
564 ///
565 /// This function is intended for cases where creating and pinning a future
566 /// via [`writable`] is not feasible. Where possible, using [`writable`] is
567 /// preferred, as this supports polling from multiple tasks at once.
568 ///
569 /// # Return value
570 ///
571 /// The function returns:
572 ///
573 /// * `Poll::Pending` if the unix stream is not ready for writing.
574 /// * `Poll::Ready(Ok(()))` if the unix stream is ready for writing.
575 /// * `Poll::Ready(Err(e))` if an error is encountered.
576 ///
577 /// # Errors
578 ///
579 /// This function may encounter any standard I/O error except `WouldBlock`.
580 ///
581 /// [`writable`]: method@Self::writable
582 pub fn poll_write_ready(&self, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
583 self.io.registration().poll_write_ready(cx).map_ok(|_| ())
584 }
585
586 /// Tries to write a buffer to the stream, returning how many bytes were
587 /// written.
588 ///
589 /// The function will attempt to write the entire contents of `buf`, but
590 /// only part of the buffer may be written.
591 ///
592 /// This function is usually paired with `writable()`.
593 ///
594 /// # Return
595 ///
596 /// If data is successfully written, `Ok(n)` is returned, where `n` is the
597 /// number of bytes written. If the stream is not ready to write data,
598 /// `Err(io::ErrorKind::WouldBlock)` is returned.
599 ///
600 /// # Examples
601 ///
602 /// ```no_run
603 /// use tokio::net::UnixStream;
604 /// use std::error::Error;
605 /// use std::io;
606 ///
607 /// #[tokio::main]
608 /// async fn main() -> Result<(), Box<dyn Error>> {
609 /// // Connect to a peer
610 /// let dir = tempfile::tempdir().unwrap();
611 /// let bind_path = dir.path().join("bind_path");
612 /// let stream = UnixStream::connect(bind_path).await?;
613 ///
614 /// loop {
615 /// // Wait for the socket to be writable
616 /// stream.writable().await?;
617 ///
618 /// // Try to write data, this may still fail with `WouldBlock`
619 /// // if the readiness event is a false positive.
620 /// match stream.try_write(b"hello world") {
621 /// Ok(n) => {
622 /// break;
623 /// }
624 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
625 /// continue;
626 /// }
627 /// Err(e) => {
628 /// return Err(e.into());
629 /// }
630 /// }
631 /// }
632 ///
633 /// Ok(())
634 /// }
635 /// ```
636 pub fn try_write(&self, buf: &[u8]) -> io::Result<usize> {
637 self.io
638 .registration()
639 .try_io(Interest::WRITABLE, || (&*self.io).write(buf))
640 }
641
642 /// Tries to write several buffers to the stream, returning how many bytes
643 /// were written.
644 ///
645 /// Data is written from each buffer in order, with the final buffer read
646 /// from possible being only partially consumed. This method behaves
647 /// equivalently to a single call to [`try_write()`] with concatenated
648 /// buffers.
649 ///
650 /// This function is usually paired with `writable()`.
651 ///
652 /// [`try_write()`]: UnixStream::try_write()
653 ///
654 /// # Return
655 ///
656 /// If data is successfully written, `Ok(n)` is returned, where `n` is the
657 /// number of bytes written. If the stream is not ready to write data,
658 /// `Err(io::ErrorKind::WouldBlock)` is returned.
659 ///
660 /// # Examples
661 ///
662 /// ```no_run
663 /// use tokio::net::UnixStream;
664 /// use std::error::Error;
665 /// use std::io;
666 ///
667 /// #[tokio::main]
668 /// async fn main() -> Result<(), Box<dyn Error>> {
669 /// // Connect to a peer
670 /// let dir = tempfile::tempdir().unwrap();
671 /// let bind_path = dir.path().join("bind_path");
672 /// let stream = UnixStream::connect(bind_path).await?;
673 ///
674 /// let bufs = [io::IoSlice::new(b"hello "), io::IoSlice::new(b"world")];
675 ///
676 /// loop {
677 /// // Wait for the socket to be writable
678 /// stream.writable().await?;
679 ///
680 /// // Try to write data, this may still fail with `WouldBlock`
681 /// // if the readiness event is a false positive.
682 /// match stream.try_write_vectored(&bufs) {
683 /// Ok(n) => {
684 /// break;
685 /// }
686 /// Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
687 /// continue;
688 /// }
689 /// Err(e) => {
690 /// return Err(e.into());
691 /// }
692 /// }
693 /// }
694 ///
695 /// Ok(())
696 /// }
697 /// ```
698 pub fn try_write_vectored(&self, buf: &[io::IoSlice<'_>]) -> io::Result<usize> {
699 self.io
700 .registration()
701 .try_io(Interest::WRITABLE, || (&*self.io).write_vectored(buf))
702 }
703
704 /// Tries to read or write from the socket using a user-provided IO operation.
705 ///
706 /// If the socket is ready, the provided closure is called. The closure
707 /// should attempt to perform IO operation on the socket by manually
708 /// calling the appropriate syscall. If the operation fails because the
709 /// socket is not actually ready, then the closure should return a
710 /// `WouldBlock` error and the readiness flag is cleared. The return value
711 /// of the closure is then returned by `try_io`.
712 ///
713 /// If the socket is not ready, then the closure is not called
714 /// and a `WouldBlock` error is returned.
715 ///
716 /// The closure should only return a `WouldBlock` error if it has performed
717 /// an IO operation on the socket that failed due to the socket not being
718 /// ready. Returning a `WouldBlock` error in any other situation will
719 /// incorrectly clear the readiness flag, which can cause the socket to
720 /// behave incorrectly.
721 ///
722 /// The closure should not perform the IO operation using any of the methods
723 /// defined on the Tokio `UnixStream` type, as this will mess with the
724 /// readiness flag and can cause the socket to behave incorrectly.
725 ///
726 /// This method is not intended to be used with combined interests.
727 /// The closure should perform only one type of IO operation, so it should not
728 /// require more than one ready state. This method may panic or sleep forever
729 /// if it is called with a combined interest.
730 ///
731 /// Usually, [`readable()`], [`writable()`] or [`ready()`] is used with this function.
732 ///
733 /// [`readable()`]: UnixStream::readable()
734 /// [`writable()`]: UnixStream::writable()
735 /// [`ready()`]: UnixStream::ready()
736 pub fn try_io<R>(
737 &self,
738 interest: Interest,
739 f: impl FnOnce() -> io::Result<R>,
740 ) -> io::Result<R> {
741 self.io
742 .registration()
743 .try_io(interest, || self.io.try_io(f))
744 }
745
746 /// Reads or writes from the socket using a user-provided IO operation.
747 ///
748 /// The readiness of the socket is awaited and when the socket is ready,
749 /// the provided closure is called. The closure should attempt to perform
750 /// IO operation on the socket by manually calling the appropriate syscall.
751 /// If the operation fails because the socket is not actually ready,
752 /// then the closure should return a `WouldBlock` error. In such case the
753 /// readiness flag is cleared and the socket readiness is awaited again.
754 /// This loop is repeated until the closure returns an `Ok` or an error
755 /// other than `WouldBlock`.
756 ///
757 /// The closure should only return a `WouldBlock` error if it has performed
758 /// an IO operation on the socket that failed due to the socket not being
759 /// ready. Returning a `WouldBlock` error in any other situation will
760 /// incorrectly clear the readiness flag, which can cause the socket to
761 /// behave incorrectly.
762 ///
763 /// The closure should not perform the IO operation using any of the methods
764 /// defined on the Tokio `UnixStream` type, as this will mess with the
765 /// readiness flag and can cause the socket to behave incorrectly.
766 ///
767 /// This method is not intended to be used with combined interests.
768 /// The closure should perform only one type of IO operation, so it should not
769 /// require more than one ready state. This method may panic or sleep forever
770 /// if it is called with a combined interest.
771 pub async fn async_io<R>(
772 &self,
773 interest: Interest,
774 mut f: impl FnMut() -> io::Result<R>,
775 ) -> io::Result<R> {
776 self.io
777 .registration()
778 .async_io(interest, || self.io.try_io(&mut f))
779 .await
780 }
781
782 /// Creates new [`UnixStream`] from a [`std::os::unix::net::UnixStream`].
783 ///
784 /// This function is intended to be used to wrap a `UnixStream` from the
785 /// standard library in the Tokio equivalent.
786 ///
787 /// # Notes
788 ///
789 /// The caller is responsible for ensuring that the stream is in
790 /// non-blocking mode. Otherwise all I/O operations on the stream
791 /// will block the thread, which will cause unexpected behavior.
792 /// Non-blocking mode can be set using [`set_nonblocking`].
793 ///
794 /// [`set_nonblocking`]: std::os::unix::net::UnixStream::set_nonblocking
795 ///
796 /// # Examples
797 ///
798 /// ```no_run
799 /// use tokio::net::UnixStream;
800 /// use std::os::unix::net::UnixStream as StdUnixStream;
801 /// # use std::error::Error;
802 ///
803 /// # async fn dox() -> Result<(), Box<dyn Error>> {
804 /// let std_stream = StdUnixStream::connect("/path/to/the/socket")?;
805 /// std_stream.set_nonblocking(true)?;
806 /// let stream = UnixStream::from_std(std_stream)?;
807 /// # Ok(())
808 /// # }
809 /// ```
810 ///
811 /// # Panics
812 ///
813 /// This function panics if it is not called from within a runtime with
814 /// IO enabled.
815 ///
816 /// The runtime is usually set implicitly when this function is called
817 /// from a future driven by a tokio runtime, otherwise runtime can be set
818 /// explicitly with [`Runtime::enter`](crate::runtime::Runtime::enter) function.
819 #[track_caller]
820 pub fn from_std(stream: net::UnixStream) -> io::Result<UnixStream> {
821 let stream = mio::net::UnixStream::from_std(stream);
822 let io = PollEvented::new(stream)?;
823
824 Ok(UnixStream { io })
825 }
826
827 /// Turns a [`tokio::net::UnixStream`] into a [`std::os::unix::net::UnixStream`].
828 ///
829 /// The returned [`std::os::unix::net::UnixStream`] will have nonblocking
830 /// mode set as `true`. Use [`set_nonblocking`] to change the blocking
831 /// mode if needed.
832 ///
833 /// # Examples
834 ///
835 /// ```
836 /// use std::error::Error;
837 /// use std::io::Read;
838 /// use tokio::net::UnixListener;
839 /// # use tokio::net::UnixStream;
840 /// # use tokio::io::AsyncWriteExt;
841 ///
842 /// #[tokio::main]
843 /// async fn main() -> Result<(), Box<dyn Error>> {
844 /// let dir = tempfile::tempdir().unwrap();
845 /// let bind_path = dir.path().join("bind_path");
846 ///
847 /// let mut data = [0u8; 12];
848 /// let listener = UnixListener::bind(&bind_path)?;
849 /// # let handle = tokio::spawn(async {
850 /// # let mut stream = UnixStream::connect(bind_path).await.unwrap();
851 /// # stream.write(b"Hello world!").await.unwrap();
852 /// # });
853 /// let (tokio_unix_stream, _) = listener.accept().await?;
854 /// let mut std_unix_stream = tokio_unix_stream.into_std()?;
855 /// # handle.await.expect("The task being joined has panicked");
856 /// std_unix_stream.set_nonblocking(false)?;
857 /// std_unix_stream.read_exact(&mut data)?;
858 /// # assert_eq!(b"Hello world!", &data);
859 /// Ok(())
860 /// }
861 /// ```
862 /// [`tokio::net::UnixStream`]: UnixStream
863 /// [`std::os::unix::net::UnixStream`]: std::os::unix::net::UnixStream
864 /// [`set_nonblocking`]: fn@std::os::unix::net::UnixStream::set_nonblocking
865 pub fn into_std(self) -> io::Result<std::os::unix::net::UnixStream> {
866 self.io
867 .into_inner()
868 .map(IntoRawFd::into_raw_fd)
869 .map(|raw_fd| unsafe { std::os::unix::net::UnixStream::from_raw_fd(raw_fd) })
870 }
871
872 /// Creates an unnamed pair of connected sockets.
873 ///
874 /// This function will create a pair of interconnected Unix sockets for
875 /// communicating back and forth between one another. Each socket will
876 /// be associated with the default event loop's handle.
877 pub fn pair() -> io::Result<(UnixStream, UnixStream)> {
878 let (a, b) = mio::net::UnixStream::pair()?;
879 let a = UnixStream::new(a)?;
880 let b = UnixStream::new(b)?;
881
882 Ok((a, b))
883 }
884
885 pub(crate) fn new(stream: mio::net::UnixStream) -> io::Result<UnixStream> {
886 let io = PollEvented::new(stream)?;
887 Ok(UnixStream { io })
888 }
889
890 /// Returns the socket address of the local half of this connection.
891 ///
892 /// # Examples
893 ///
894 /// ```no_run
895 /// use tokio::net::UnixStream;
896 ///
897 /// # async fn dox() -> Result<(), Box<dyn std::error::Error>> {
898 /// let dir = tempfile::tempdir().unwrap();
899 /// let bind_path = dir.path().join("bind_path");
900 /// let stream = UnixStream::connect(bind_path).await?;
901 ///
902 /// println!("{:?}", stream.local_addr()?);
903 /// # Ok(())
904 /// # }
905 /// ```
906 pub fn local_addr(&self) -> io::Result<SocketAddr> {
907 self.io.local_addr().map(SocketAddr)
908 }
909
910 /// Returns the socket address of the remote half of this connection.
911 ///
912 /// # Examples
913 ///
914 /// ```no_run
915 /// use tokio::net::UnixStream;
916 ///
917 /// # async fn dox() -> Result<(), Box<dyn std::error::Error>> {
918 /// let dir = tempfile::tempdir().unwrap();
919 /// let bind_path = dir.path().join("bind_path");
920 /// let stream = UnixStream::connect(bind_path).await?;
921 ///
922 /// println!("{:?}", stream.peer_addr()?);
923 /// # Ok(())
924 /// # }
925 /// ```
926 pub fn peer_addr(&self) -> io::Result<SocketAddr> {
927 self.io.peer_addr().map(SocketAddr)
928 }
929
930 /// Returns effective credentials of the process which called `connect` or `pair`.
931 pub fn peer_cred(&self) -> io::Result<UCred> {
932 ucred::get_peer_cred(self)
933 }
934
935 /// Returns the value of the `SO_ERROR` option.
936 pub fn take_error(&self) -> io::Result<Option<io::Error>> {
937 self.io.take_error()
938 }
939
940 /// Shuts down the read, write, or both halves of this connection.
941 ///
942 /// This function will cause all pending and future I/O calls on the
943 /// specified portions to immediately return with an appropriate value
944 /// (see the documentation of `Shutdown`).
945 pub(super) fn shutdown_std(&self, how: Shutdown) -> io::Result<()> {
946 self.io.shutdown(how)
947 }
948
949 // These lifetime markers also appear in the generated documentation, and make
950 // it more clear that this is a *borrowed* split.
951 #[allow(clippy::needless_lifetimes)]
952 /// Splits a `UnixStream` into a read half and a write half, which can be used
953 /// to read and write the stream concurrently.
954 ///
955 /// This method is more efficient than [`into_split`], but the halves cannot be
956 /// moved into independently spawned tasks.
957 ///
958 /// [`into_split`]: Self::into_split()
959 pub fn split<'a>(&'a mut self) -> (ReadHalf<'a>, WriteHalf<'a>) {
960 split(self)
961 }
962
963 /// Splits a `UnixStream` into a read half and a write half, which can be used
964 /// to read and write the stream concurrently.
965 ///
966 /// Unlike [`split`], the owned halves can be moved to separate tasks, however
967 /// this comes at the cost of a heap allocation.
968 ///
969 /// **Note:** Dropping the write half will shut down the write half of the
970 /// stream. This is equivalent to calling [`shutdown()`] on the `UnixStream`.
971 ///
972 /// [`split`]: Self::split()
973 /// [`shutdown()`]: fn@crate::io::AsyncWriteExt::shutdown
974 pub fn into_split(self) -> (OwnedReadHalf, OwnedWriteHalf) {
975 split_owned(self)
976 }
977}
978
979impl TryFrom<net::UnixStream> for UnixStream {
980 type Error = io::Error;
981
982 /// Consumes stream, returning the tokio I/O object.
983 ///
984 /// This is equivalent to
985 /// [`UnixStream::from_std(stream)`](UnixStream::from_std).
986 fn try_from(stream: net::UnixStream) -> io::Result<Self> {
987 Self::from_std(stream)
988 }
989}
990
991impl AsyncRead for UnixStream {
992 fn poll_read(
993 self: Pin<&mut Self>,
994 cx: &mut Context<'_>,
995 buf: &mut ReadBuf<'_>,
996 ) -> Poll<io::Result<()>> {
997 self.poll_read_priv(cx, buf)
998 }
999}
1000
1001impl AsyncWrite for UnixStream {
1002 fn poll_write(
1003 self: Pin<&mut Self>,
1004 cx: &mut Context<'_>,
1005 buf: &[u8],
1006 ) -> Poll<io::Result<usize>> {
1007 self.poll_write_priv(cx, buf)
1008 }
1009
1010 fn poll_write_vectored(
1011 self: Pin<&mut Self>,
1012 cx: &mut Context<'_>,
1013 bufs: &[io::IoSlice<'_>],
1014 ) -> Poll<io::Result<usize>> {
1015 self.poll_write_vectored_priv(cx, bufs)
1016 }
1017
1018 fn is_write_vectored(&self) -> bool {
1019 true
1020 }
1021
1022 fn poll_flush(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<io::Result<()>> {
1023 Poll::Ready(Ok(()))
1024 }
1025
1026 fn poll_shutdown(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<io::Result<()>> {
1027 self.shutdown_std(std::net::Shutdown::Write)?;
1028 Poll::Ready(Ok(()))
1029 }
1030}
1031
1032impl UnixStream {
1033 // == Poll IO functions that takes `&self` ==
1034 //
1035 // To read or write without mutable access to the `UnixStream`, combine the
1036 // `poll_read_ready` or `poll_write_ready` methods with the `try_read` or
1037 // `try_write` methods.
1038
1039 pub(crate) fn poll_read_priv(
1040 &self,
1041 cx: &mut Context<'_>,
1042 buf: &mut ReadBuf<'_>,
1043 ) -> Poll<io::Result<()>> {
1044 // Safety: `UnixStream::read` correctly handles reads into uninitialized memory
1045 unsafe { self.io.poll_read(cx, buf) }
1046 }
1047
1048 pub(crate) fn poll_write_priv(
1049 &self,
1050 cx: &mut Context<'_>,
1051 buf: &[u8],
1052 ) -> Poll<io::Result<usize>> {
1053 self.io.poll_write(cx, buf)
1054 }
1055
1056 pub(super) fn poll_write_vectored_priv(
1057 &self,
1058 cx: &mut Context<'_>,
1059 bufs: &[io::IoSlice<'_>],
1060 ) -> Poll<io::Result<usize>> {
1061 self.io.poll_write_vectored(cx, bufs)
1062 }
1063}
1064
1065impl fmt::Debug for UnixStream {
1066 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1067 self.io.fmt(f)
1068 }
1069}
1070
1071impl AsRawFd for UnixStream {
1072 fn as_raw_fd(&self) -> RawFd {
1073 self.io.as_raw_fd()
1074 }
1075}
1076
1077impl AsFd for UnixStream {
1078 fn as_fd(&self) -> BorrowedFd<'_> {
1079 unsafe { BorrowedFd::borrow_raw(self.as_raw_fd()) }
1080 }
1081}