1 // Copyright 2014 The Rust Project Developers. See the COPYRIGHT
2 // file at the top-level directory of this distribution and at
3 // http://rust-lang.org/COPYRIGHT.
5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
8 // option. This file may not be copied, modified, or distributed
9 // except according to those terms.
11 //! Named pipes implementation for windows
13 //! If are unfortunate enough to be reading this code, I would like to first
14 //! apologize. This was my first encounter with windows named pipes, and it
15 //! didn't exactly turn out very cleanly. If you, too, are new to named pipes,
16 //! read on as I'll try to explain some fun things that I ran into.
18 //! # Unix pipes vs Named pipes
20 //! As with everything else, named pipes on windows are pretty different from
21 //! unix pipes on unix. On unix, you use one "server pipe" to accept new client
22 //! pipes. So long as this server pipe is active, new children pipes can
23 //! connect. On windows, you instead have a number of "server pipes", and each
24 //! of these server pipes can throughout their lifetime be attached to a client
25 //! or not. Once attached to a client, a server pipe may then disconnect at a
28 //! # Accepting clients
30 //! As with most other I/O interfaces, our Listener/Acceptor/Stream interfaces
31 //! are built around the unix flavors. This means that we have one "server
32 //! pipe" to which many clients can connect. In order to make this compatible
33 //! with the windows model, each connected client consumes ownership of a server
34 //! pipe, and then a new server pipe is created for the next client.
36 //! Note that the server pipes attached to clients are never given back to the
37 //! listener for recycling. This could possibly be implemented with a channel so
38 //! the listener half can re-use server pipes, but for now I err'd on the simple
39 //! side of things. Each stream accepted by a listener will destroy the server
40 //! pipe after the stream is dropped.
42 //! This model ends up having a small race or two, and you can find more details
43 //! on the `native_accept` method.
45 //! # Simultaneous reads and writes
47 //! In testing, I found that two simultaneous writes and two simultaneous reads
48 //! on a pipe ended up working out just fine, but problems were encountered when
49 //! a read was executed simultaneously with a write. After some googling around,
50 //! it sounded like named pipes just weren't built for this kind of interaction,
51 //! and the suggested solution was to use overlapped I/O.
53 //! I don't really know what overlapped I/O is, but my basic understanding after
54 //! reading about it is that you have an external Event which is used to signal
55 //! I/O completion, passed around in some OVERLAPPED structures. As to what this
56 //! is, I'm not exactly sure.
58 //! This problem implies that all named pipes are created with the
59 //! FILE_FLAG_OVERLAPPED option. This means that all of their I/O is
60 //! asynchronous. Each I/O operation has an associated OVERLAPPED structure, and
61 //! inside of this structure is a HANDLE from CreateEvent. After the I/O is
62 //! determined to be pending (may complete in the future), the
63 //! GetOverlappedResult function is used to block on the event, waiting for the
66 //! This scheme ended up working well enough. There were two snags that I ran
69 //! * Each UnixStream instance needs its own read/write events to wait on. These
70 //! can't be shared among clones of the same stream because the documentation
71 //! states that it unsets the event when the I/O is started (would possibly
72 //! corrupt other events simultaneously waiting). For convenience's sake,
73 //! these events are lazily initialized.
75 //! * Each server pipe needs to be created with FILE_FLAG_OVERLAPPED in addition
76 //! to all pipes created through `connect`. Notably this means that the
77 //! ConnectNamedPipe function is nonblocking, implying that the Listener needs
78 //! to have yet another event to do the actual blocking.
82 //! The conclusion here is that I probably don't know the best way to work with
83 //! windows named pipes, but the solution here seems to work well enough to get
84 //! the test suite passing (the suite is in libstd), and that's good enough for
93 use sync::{atomic, Arc, Mutex};
94 use io::{mod, IoError, IoResult};
96 use sys_common::{mod, eof};
98 use super::{c, os, timer, to_utf16, decode_error_detailed};
100 struct Event(libc::HANDLE);
103 fn new(manual_reset: bool, initial_state: bool) -> IoResult<Event> {
105 libc::CreateEventW(ptr::null_mut(),
106 manual_reset as libc::BOOL,
107 initial_state as libc::BOOL,
110 if event as uint == 0 {
111 Err(super::last_error())
117 fn handle(&self) -> libc::HANDLE { let Event(handle) = *self; handle }
120 impl Drop for Event {
122 unsafe { let _ = libc::CloseHandle(self.handle()); }
127 handle: libc::HANDLE,
129 read_closed: atomic::AtomicBool,
130 write_closed: atomic::AtomicBool,
134 fn new(handle: libc::HANDLE) -> Inner {
137 lock: Mutex::new(()),
138 read_closed: atomic::AtomicBool::new(false),
139 write_closed: atomic::AtomicBool::new(false),
144 impl Drop for Inner {
147 let _ = libc::FlushFileBuffers(self.handle);
148 let _ = libc::CloseHandle(self.handle);
153 unsafe fn pipe(name: *const u16, init: bool) -> libc::HANDLE {
154 libc::CreateNamedPipeW(
156 libc::PIPE_ACCESS_DUPLEX |
157 if init {libc::FILE_FLAG_FIRST_PIPE_INSTANCE} else {0} |
158 libc::FILE_FLAG_OVERLAPPED,
159 libc::PIPE_TYPE_BYTE | libc::PIPE_READMODE_BYTE |
161 libc::PIPE_UNLIMITED_INSTANCES,
169 pub fn await(handle: libc::HANDLE, deadline: u64,
170 events: &[libc::HANDLE]) -> IoResult<uint> {
171 use libc::consts::os::extra::{WAIT_FAILED, WAIT_TIMEOUT, WAIT_OBJECT_0};
173 // If we've got a timeout, use WaitForSingleObject in tandem with CancelIo
174 // to figure out if we should indeed get the result.
175 let ms = if deadline == 0 {
176 libc::INFINITE as u64
178 let now = timer::now();
179 if deadline < now {0} else {deadline - now}
182 c::WaitForMultipleObjects(events.len() as libc::DWORD,
188 WAIT_FAILED => Err(super::last_error()),
189 WAIT_TIMEOUT => unsafe {
190 let _ = c::CancelIo(handle);
191 Err(sys_common::timeout("operation timed out"))
193 n => Ok((n - WAIT_OBJECT_0) as uint)
197 fn epipe() -> IoError {
200 desc: "the pipe has ended",
205 ////////////////////////////////////////////////////////////////////////////////
207 ////////////////////////////////////////////////////////////////////////////////
209 pub struct UnixStream {
211 write: Option<Event>,
217 unsafe impl Send for UnixStream {}
218 unsafe impl Sync for UnixStream {}
221 fn try_connect(p: *const u16) -> Option<libc::HANDLE> {
222 // Note that most of this is lifted from the libuv implementation.
223 // The idea is that if we fail to open a pipe in read/write mode
224 // that we try afterwards in just read or just write
225 let mut result = unsafe {
227 libc::GENERIC_READ | libc::GENERIC_WRITE,
231 libc::FILE_FLAG_OVERLAPPED,
234 if result != libc::INVALID_HANDLE_VALUE {
238 let err = unsafe { libc::GetLastError() };
239 if err == libc::ERROR_ACCESS_DENIED as libc::DWORD {
242 libc::GENERIC_READ | libc::FILE_WRITE_ATTRIBUTES,
246 libc::FILE_FLAG_OVERLAPPED,
249 if result != libc::INVALID_HANDLE_VALUE {
253 let err = unsafe { libc::GetLastError() };
254 if err == libc::ERROR_ACCESS_DENIED as libc::DWORD {
257 libc::GENERIC_WRITE | libc::FILE_READ_ATTRIBUTES,
261 libc::FILE_FLAG_OVERLAPPED,
264 if result != libc::INVALID_HANDLE_VALUE {
271 pub fn connect(addr: &CString, timeout: Option<u64>) -> IoResult<UnixStream> {
272 let addr = try!(to_utf16(addr.as_str()));
273 let start = timer::now();
275 match UnixStream::try_connect(addr.as_ptr()) {
277 let inner = Inner::new(handle);
278 let mut mode = libc::PIPE_TYPE_BYTE |
279 libc::PIPE_READMODE_BYTE |
282 libc::SetNamedPipeHandleState(inner.handle,
288 Err(super::last_error())
291 inner: Arc::new(inner),
302 // On windows, if you fail to connect, you may need to call the
303 // `WaitNamedPipe` function, and this is indicated with an error
304 // code of ERROR_PIPE_BUSY.
305 let code = unsafe { libc::GetLastError() };
306 if code as int != libc::ERROR_PIPE_BUSY as int {
307 return Err(super::last_error())
312 let now = timer::now();
313 let timed_out = (now - start) >= timeout || unsafe {
314 let ms = (timeout - (now - start)) as libc::DWORD;
315 libc::WaitNamedPipeW(addr.as_ptr(), ms) == 0
318 return Err(sys_common::timeout("connect timed out"))
322 // An example I found on Microsoft's website used 20
323 // seconds, libuv uses 30 seconds, hence we make the
324 // obvious choice of waiting for 25 seconds.
326 if unsafe { libc::WaitNamedPipeW(addr.as_ptr(), 25000) } == 0 {
327 return Err(super::last_error())
334 pub fn handle(&self) -> libc::HANDLE { self.inner.handle }
336 fn read_closed(&self) -> bool {
337 self.inner.read_closed.load(atomic::SeqCst)
340 fn write_closed(&self) -> bool {
341 self.inner.write_closed.load(atomic::SeqCst)
344 fn cancel_io(&self) -> IoResult<()> {
345 match unsafe { c::CancelIoEx(self.handle(), ptr::null_mut()) } {
346 0 if os::errno() == libc::ERROR_NOT_FOUND as uint => {
349 0 => Err(super::last_error()),
354 pub fn read(&mut self, buf: &mut [u8]) -> IoResult<uint> {
355 if self.read.is_none() {
356 self.read = Some(try!(Event::new(true, false)));
359 let mut bytes_read = 0;
360 let mut overlapped: libc::OVERLAPPED = unsafe { mem::zeroed() };
361 overlapped.hEvent = self.read.as_ref().unwrap().handle();
363 // Pre-flight check to see if the reading half has been closed. This
364 // must be done before issuing the ReadFile request, but after we
367 // See comments in close_read() about why this lock is necessary.
368 let guard = unsafe { self.inner.lock.lock() };
369 if self.read_closed() {
373 // Issue a nonblocking requests, succeeding quickly if it happened to
376 libc::ReadFile(self.handle(),
377 buf.as_ptr() as libc::LPVOID,
378 buf.len() as libc::DWORD,
382 if ret != 0 { return Ok(bytes_read as uint) }
384 // If our errno doesn't say that the I/O is pending, then we hit some
385 // legitimate error and return immediately.
386 if os::errno() != libc::ERROR_IO_PENDING as uint {
387 return Err(super::last_error())
390 // Now that we've issued a successful nonblocking request, we need to
391 // wait for it to finish. This can all be done outside the lock because
392 // we'll see any invocation of CancelIoEx. We also call this in a loop
393 // because we're woken up if the writing half is closed, we just need to
394 // realize that the reading half wasn't closed and we go right back to
398 // Process a timeout if one is pending
399 let wait_succeeded = await(self.handle(), self.read_deadline,
400 &[overlapped.hEvent]);
403 libc::GetOverlappedResult(self.handle(),
408 // If we succeeded, or we failed for some reason other than
409 // CancelIoEx, return immediately
410 if ret != 0 { return Ok(bytes_read as uint) }
411 if os::errno() != libc::ERROR_OPERATION_ABORTED as uint {
412 return Err(super::last_error())
415 // If the reading half is now closed, then we're done. If we woke up
416 // because the writing half was closed, keep trying.
417 if wait_succeeded.is_err() {
418 return Err(sys_common::timeout("read timed out"))
420 if self.read_closed() {
426 pub fn write(&mut self, buf: &[u8]) -> IoResult<()> {
427 if self.write.is_none() {
428 self.write = Some(try!(Event::new(true, false)));
432 let mut overlapped: libc::OVERLAPPED = unsafe { mem::zeroed() };
433 overlapped.hEvent = self.write.as_ref().unwrap().handle();
435 while offset < buf.len() {
436 let mut bytes_written = 0;
438 // This sequence below is quite similar to the one found in read().
439 // Some careful looping is done to ensure that if close_write() is
440 // invoked we bail out early, and if close_read() is invoked we keep
441 // going after we woke up.
443 // See comments in close_read() about why this lock is necessary.
444 let guard = unsafe { self.inner.lock.lock() };
445 if self.write_closed() {
449 libc::WriteFile(self.handle(),
450 buf[offset..].as_ptr() as libc::LPVOID,
451 (buf.len() - offset) as libc::DWORD,
455 let err = os::errno();
459 if err != libc::ERROR_IO_PENDING as uint {
460 return Err(decode_error_detailed(err as i32))
462 // Process a timeout if one is pending
463 let wait_succeeded = await(self.handle(), self.write_deadline,
464 &[overlapped.hEvent]);
466 libc::GetOverlappedResult(self.handle(),
471 // If we weren't aborted, this was a legit error, if we were
472 // aborted, then check to see if the write half was actually
473 // closed or whether we woke up from the read half closing.
475 if os::errno() != libc::ERROR_OPERATION_ABORTED as uint {
476 return Err(super::last_error())
478 if !wait_succeeded.is_ok() {
479 let amt = offset + bytes_written as uint;
482 kind: io::ShortWrite(amt),
483 desc: "short write during write",
487 Err(sys_common::timeout("write timed out"))
490 if self.write_closed() {
496 offset += bytes_written as uint;
501 pub fn close_read(&mut self) -> IoResult<()> {
502 // On windows, there's no actual shutdown() method for pipes, so we're
503 // forced to emulate the behavior manually at the application level. To
504 // do this, we need to both cancel any pending requests, as well as
505 // prevent all future requests from succeeding. These two operations are
506 // not atomic with respect to one another, so we must use a lock to do
509 // The read() code looks like:
511 // 1. Make sure the pipe is still open
512 // 2. Submit a read request
513 // 3. Wait for the read request to finish
515 // The race this lock is preventing is if another thread invokes
516 // close_read() between steps 1 and 2. By atomically executing steps 1
517 // and 2 with a lock with respect to close_read(), we're guaranteed that
518 // no thread will erroneously sit in a read forever.
519 let _guard = unsafe { self.inner.lock.lock() };
520 self.inner.read_closed.store(true, atomic::SeqCst);
524 pub fn close_write(&mut self) -> IoResult<()> {
525 // see comments in close_read() for why this lock is necessary
526 let _guard = unsafe { self.inner.lock.lock() };
527 self.inner.write_closed.store(true, atomic::SeqCst);
531 pub fn set_timeout(&mut self, timeout: Option<u64>) {
532 let deadline = timeout.map(|a| timer::now() + a).unwrap_or(0);
533 self.read_deadline = deadline;
534 self.write_deadline = deadline;
536 pub fn set_read_timeout(&mut self, timeout: Option<u64>) {
537 self.read_deadline = timeout.map(|a| timer::now() + a).unwrap_or(0);
539 pub fn set_write_timeout(&mut self, timeout: Option<u64>) {
540 self.write_deadline = timeout.map(|a| timer::now() + a).unwrap_or(0);
544 impl Clone for UnixStream {
545 fn clone(&self) -> UnixStream {
547 inner: self.inner.clone(),
556 ////////////////////////////////////////////////////////////////////////////////
558 ////////////////////////////////////////////////////////////////////////////////
560 pub struct UnixListener {
561 handle: libc::HANDLE,
565 unsafe impl Send for UnixListener {}
566 unsafe impl Sync for UnixListener {}
569 pub fn bind(addr: &CString) -> IoResult<UnixListener> {
570 // Although we technically don't need the pipe until much later, we
571 // create the initial handle up front to test the validity of the name
573 let addr_v = try!(to_utf16(addr.as_str()));
574 let ret = unsafe { pipe(addr_v.as_ptr(), true) };
575 if ret == libc::INVALID_HANDLE_VALUE {
576 Err(super::last_error())
578 Ok(UnixListener { handle: ret, name: addr.clone() })
582 pub fn listen(self) -> IoResult<UnixAcceptor> {
585 event: try!(Event::new(true, false)),
587 inner: Arc::new(AcceptorState {
588 abort: try!(Event::new(true, false)),
589 closed: atomic::AtomicBool::new(false),
594 pub fn handle(&self) -> libc::HANDLE {
599 impl Drop for UnixListener {
601 unsafe { let _ = libc::CloseHandle(self.handle); }
605 pub struct UnixAcceptor {
606 inner: Arc<AcceptorState>,
607 listener: UnixListener,
612 unsafe impl Send for UnixAcceptor {}
613 unsafe impl Sync for UnixAcceptor {}
615 struct AcceptorState {
617 closed: atomic::AtomicBool,
620 unsafe impl Send for AcceptorState {}
621 unsafe impl Sync for AcceptorState {}
624 pub fn accept(&mut self) -> IoResult<UnixStream> {
625 // This function has some funky implementation details when working with
626 // unix pipes. On windows, each server named pipe handle can be
627 // connected to a one or zero clients. To the best of my knowledge, a
628 // named server is considered active and present if there exists at
629 // least one server named pipe for it.
631 // The model of this function is to take the current known server
632 // handle, connect a client to it, and then transfer ownership to the
633 // UnixStream instance. The next time accept() is invoked, it'll need a
634 // different server handle to connect a client to.
636 // Note that there is a possible race here. Once our server pipe is
637 // handed off to a `UnixStream` object, the stream could be closed,
638 // meaning that there would be no active server pipes, hence even though
639 // we have a valid `UnixAcceptor`, no one can connect to it. For this
640 // reason, we generate the next accept call's server pipe at the end of
641 // this function call.
643 // This provides us an invariant that we always have at least one server
644 // connection open at a time, meaning that all connects to this acceptor
645 // should succeed while this is active.
647 // The actual implementation of doing this is a little tricky. Once a
648 // server pipe is created, a client can connect to it at any time. I
649 // assume that which server a client connects to is nondeterministic, so
650 // we also need to guarantee that the only server able to be connected
651 // to is the one that we're calling ConnectNamedPipe on. This means that
652 // we have to create the second server pipe *after* we've already
653 // accepted a connection. In order to at least somewhat gracefully
654 // handle errors, this means that if the second server pipe creation
655 // fails that we disconnect the connected client and then just keep
656 // using the original server pipe.
657 let handle = self.listener.handle;
659 // If we've had an artificial call to close_accept, be sure to never
660 // proceed in accepting new clients in the future
661 if self.inner.closed.load(atomic::SeqCst) { return Err(eof()) }
663 let name = try!(to_utf16(self.listener.name.as_str()));
665 // Once we've got a "server handle", we need to wait for a client to
666 // connect. The ConnectNamedPipe function will block this thread until
667 // someone on the other end connects. This function can "fail" if a
668 // client connects after we created the pipe but before we got down
669 // here. Thanks windows.
670 let mut overlapped: libc::OVERLAPPED = unsafe { mem::zeroed() };
671 overlapped.hEvent = self.event.handle();
672 if unsafe { libc::ConnectNamedPipe(handle, &mut overlapped) == 0 } {
673 let mut err = unsafe { libc::GetLastError() };
675 if err == libc::ERROR_IO_PENDING as libc::DWORD {
676 // Process a timeout if one is pending
677 let wait_succeeded = await(handle, self.deadline,
678 &[self.inner.abort.handle(),
681 // This will block until the overlapped I/O is completed. The
682 // timeout was previously handled, so this will either block in
683 // the normal case or succeed very quickly in the timeout case.
685 let mut transfer = 0;
686 libc::GetOverlappedResult(handle,
692 if wait_succeeded.is_ok() {
693 err = unsafe { libc::GetLastError() };
695 return Err(sys_common::timeout("accept timed out"))
698 // we succeeded, bypass the check below
699 err = libc::ERROR_PIPE_CONNECTED as libc::DWORD;
702 if err != libc::ERROR_PIPE_CONNECTED as libc::DWORD {
703 return Err(super::last_error())
707 // Now that we've got a connected client to our handle, we need to
708 // create a second server pipe. If this fails, we disconnect the
709 // connected client and return an error (see comments above).
710 let new_handle = unsafe { pipe(name.as_ptr(), false) };
711 if new_handle == libc::INVALID_HANDLE_VALUE {
712 let ret = Err(super::last_error());
713 // If our disconnection fails, then there's not really a whole lot
714 // that we can do, so panic
715 let err = unsafe { libc::DisconnectNamedPipe(handle) };
719 self.listener.handle = new_handle;
722 // Transfer ownership of our handle into this stream
724 inner: Arc::new(Inner::new(handle)),
732 pub fn set_timeout(&mut self, timeout: Option<u64>) {
733 self.deadline = timeout.map(|i| i + timer::now()).unwrap_or(0);
736 pub fn close_accept(&mut self) -> IoResult<()> {
737 self.inner.closed.store(true, atomic::SeqCst);
739 c::SetEvent(self.inner.abort.handle())
742 Err(super::last_error())
748 pub fn handle(&self) -> libc::HANDLE {
749 self.listener.handle()
753 impl Clone for UnixAcceptor {
754 fn clone(&self) -> UnixAcceptor {
755 let name = to_utf16(self.listener.name.as_str()).ok().unwrap();
757 inner: self.inner.clone(),
758 event: Event::new(true, false).ok().unwrap(),
760 listener: UnixListener {
761 name: self.listener.name.clone(),
763 let p = pipe(name.as_ptr(), false) ;
764 assert!(p != libc::INVALID_HANDLE_VALUE as libc::HANDLE);