rust/src/libstd/io/net/unix.rs

350 lines
9.5 KiB
Rust
Raw Normal View History

// Copyright 2013 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
2013-10-15 19:44:08 -07:00
/*!
Named pipes
This module contains the ability to communicate over named pipes with
synchronous I/O. On windows, this corresponds to talking over a Named Pipe,
while on Unix it corresponds to UNIX domain sockets.
These pipes are similar to TCP in the sense that you can have both a stream to a
server and a server itself. The server provided accepts other `UnixStream`
instances as clients.
*/
use prelude::*;
2013-10-15 19:44:08 -07:00
use c_str::ToCStr;
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
use clone::Clone;
use rt::rtio::{IoFactory, LocalIo, RtioUnixListener};
use rt::rtio::{RtioUnixAcceptor, RtioPipe};
2013-11-10 22:46:32 -08:00
use io::pipe::PipeStream;
use io::{Listener, Acceptor, Reader, Writer, IoResult};
2013-10-15 19:44:08 -07:00
/// A stream which communicates over a named pipe.
pub struct UnixStream {
priv obj: PipeStream,
}
impl UnixStream {
fn new(obj: ~RtioPipe) -> UnixStream {
UnixStream { obj: PipeStream::new(obj) }
2013-10-15 19:44:08 -07:00
}
/// Connect to a pipe named by `path`. This will attempt to open a
/// connection to the underlying socket.
///
/// The returned stream will be closed when the object falls out of scope.
///
/// # Example
///
/// ```rust
/// # #[allow(unused_must_use)];
/// use std::io::net::unix::UnixStream;
2013-10-15 19:44:08 -07:00
///
/// let server = Path::new("path/to/my/socket");
/// let mut stream = UnixStream::connect(&server);
/// stream.write([1, 2, 3]);
/// ```
pub fn connect<P: ToCStr>(path: &P) -> IoResult<UnixStream> {
LocalIo::maybe_raise(|io| {
io.unix_connect(&path.to_c_str()).map(UnixStream::new)
})
}
}
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
impl Clone for UnixStream {
fn clone(&self) -> UnixStream {
UnixStream { obj: self.obj.clone() }
}
}
impl Reader for UnixStream {
fn read(&mut self, buf: &mut [u8]) -> IoResult<uint> { self.obj.read(buf) }
}
impl Writer for UnixStream {
fn write(&mut self, buf: &[u8]) -> IoResult<()> { self.obj.write(buf) }
}
2013-10-15 19:44:08 -07:00
pub struct UnixListener {
priv obj: ~RtioUnixListener,
2013-10-15 19:44:08 -07:00
}
impl UnixListener {
2013-10-15 19:44:08 -07:00
/// Creates a new listener, ready to receive incoming connections on the
/// specified socket. The server will be named by `path`.
///
/// This listener will be closed when it falls out of scope.
///
/// # Example
///
/// ```
2014-02-14 23:44:22 -08:00
/// # fn main() {}
/// # fn foo() {
/// # #[allow(unused_must_use)];
/// use std::io::net::unix::UnixListener;
2014-02-14 23:44:22 -08:00
/// use std::io::{Listener, Acceptor};
2013-10-15 19:44:08 -07:00
///
2014-02-14 23:44:22 -08:00
/// let server = Path::new("/path/to/my/socket");
/// let stream = UnixListener::bind(&server);
/// for mut client in stream.listen().incoming() {
/// client.write([1, 2, 3, 4]);
/// }
2014-02-14 23:44:22 -08:00
/// # }
/// ```
pub fn bind<P: ToCStr>(path: &P) -> IoResult<UnixListener> {
LocalIo::maybe_raise(|io| {
io.unix_bind(&path.to_c_str()).map(|s| UnixListener { obj: s })
})
}
}
impl Listener<UnixStream, UnixAcceptor> for UnixListener {
fn listen(self) -> IoResult<UnixAcceptor> {
self.obj.listen().map(|obj| UnixAcceptor { obj: obj })
2013-10-15 19:44:08 -07:00
}
}
2013-10-15 19:44:08 -07:00
pub struct UnixAcceptor {
priv obj: ~RtioUnixAcceptor,
2013-10-15 19:44:08 -07:00
}
impl Acceptor<UnixStream> for UnixAcceptor {
fn accept(&mut self) -> IoResult<UnixStream> {
self.obj.accept().map(UnixStream::new)
2013-10-15 19:44:08 -07:00
}
}
#[cfg(test)]
mod tests {
use prelude::*;
use super::*;
2013-11-10 22:46:32 -08:00
use io::*;
use io::test::*;
2013-10-15 19:44:08 -07:00
pub fn smalltest(server: proc(UnixStream), client: proc(UnixStream)) {
let path1 = next_test_unix();
let path2 = path1.clone();
let (port, chan) = Chan::new();
2013-10-15 19:44:08 -07:00
spawn(proc() {
2013-12-05 18:19:06 -08:00
port.recv();
2013-12-15 18:17:43 -08:00
client(UnixStream::connect(&path2).unwrap());
});
let mut acceptor = UnixListener::bind(&path1).listen();
chan.send(());
server(acceptor.accept().unwrap());
2013-10-15 19:44:08 -07:00
}
iotest!(fn bind_error() {
let path = "path/to/nowhere";
match UnixListener::bind(&path) {
2014-01-30 14:10:53 -08:00
Ok(..) => fail!(),
Err(e) => {
assert!(e.kind == PermissionDenied || e.kind == FileNotFound ||
e.kind == InvalidInput);
}
2014-01-30 14:10:53 -08:00
}
})
iotest!(fn connect_error() {
let path = if cfg!(windows) {
r"\\.\pipe\this_should_not_exist_ever"
} else {
"path/to/nowhere"
};
match UnixStream::connect(&path) {
2014-01-30 14:10:53 -08:00
Ok(..) => fail!(),
Err(e) => {
assert!(e.kind == FileNotFound || e.kind == OtherIoError);
}
2014-01-30 14:10:53 -08:00
}
})
2013-10-15 19:44:08 -07:00
iotest!(fn smoke() {
smalltest(proc(mut server) {
2013-10-15 19:44:08 -07:00
let mut buf = [0];
2014-01-30 14:10:53 -08:00
server.read(buf).unwrap();
2013-10-15 19:44:08 -07:00
assert!(buf[0] == 99);
}, proc(mut client) {
2014-01-30 14:10:53 -08:00
client.write([99]).unwrap();
2013-10-15 19:44:08 -07:00
})
})
2013-10-15 19:44:08 -07:00
iotest!(fn read_eof() {
smalltest(proc(mut server) {
2013-10-15 19:44:08 -07:00
let mut buf = [0];
2014-01-30 14:10:53 -08:00
assert!(server.read(buf).is_err());
assert!(server.read(buf).is_err());
}, proc(_client) {
2013-10-15 19:44:08 -07:00
// drop the client
})
})
2013-10-15 19:44:08 -07:00
iotest!(fn write_begone() {
smalltest(proc(mut server) {
2013-10-15 19:44:08 -07:00
let buf = [0];
2014-01-30 14:10:53 -08:00
loop {
match server.write(buf) {
Ok(..) => {}
Err(e) => {
assert!(e.kind == BrokenPipe ||
e.kind == NotConnected ||
e.kind == ConnectionReset,
2014-01-30 14:10:53 -08:00
"unknown error {:?}", e);
break;
}
}
2013-10-15 19:44:08 -07:00
}
}, proc(_client) {
2013-10-15 19:44:08 -07:00
// drop the client
})
})
2013-10-15 19:44:08 -07:00
iotest!(fn accept_lots() {
let times = 10;
let path1 = next_test_unix();
let path2 = path1.clone();
let (port, chan) = Chan::new();
2013-10-15 19:44:08 -07:00
spawn(proc() {
2013-12-05 18:19:06 -08:00
port.recv();
for _ in range(0, times) {
2013-12-05 18:19:06 -08:00
let mut stream = UnixStream::connect(&path2);
match stream.write([100]) {
Ok(..) => {}
Err(e) => fail!("failed write: {}", e)
}
}
});
let mut acceptor = match UnixListener::bind(&path1).listen() {
Ok(a) => a,
Err(e) => fail!("failed listen: {}", e),
};
chan.send(());
for _ in range(0, times) {
let mut client = acceptor.accept();
let mut buf = [0];
match client.read(buf) {
Ok(..) => {}
Err(e) => fail!("failed read/accept: {}", e),
}
assert_eq!(buf[0], 100);
}
})
2013-10-15 19:44:08 -07:00
#[cfg(unix)]
iotest!(fn path_exists() {
let path = next_test_unix();
let _acceptor = UnixListener::bind(&path).listen();
assert!(path.exists());
})
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
iotest!(fn unix_clone_smoke() {
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
let addr = next_test_unix();
let mut acceptor = UnixListener::bind(&addr).listen();
spawn(proc() {
let mut s = UnixStream::connect(&addr);
let mut buf = [0, 0];
debug!("client reading");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
assert_eq!(s.read(buf), Ok(1));
assert_eq!(buf[0], 1);
debug!("client writing");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
s.write([2]).unwrap();
debug!("client dropping");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
});
let mut s1 = acceptor.accept().unwrap();
let s2 = s1.clone();
let (p1, c1) = Chan::new();
let (p2, c2) = Chan::new();
spawn(proc() {
let mut s2 = s2;
p1.recv();
debug!("writer writing");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
s2.write([1]).unwrap();
debug!("writer done");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
c2.send(());
});
c1.send(());
let mut buf = [0, 0];
debug!("reader reading");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
assert_eq!(s1.read(buf), Ok(1));
debug!("reader done");
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
p2.recv();
})
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
iotest!(fn unix_clone_two_read() {
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
let addr = next_test_unix();
let mut acceptor = UnixListener::bind(&addr).listen();
2014-02-11 16:33:34 -08:00
let (p, c) = Chan::new();
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
let c2 = c.clone();
spawn(proc() {
let mut s = UnixStream::connect(&addr);
s.write([1]).unwrap();
p.recv();
s.write([2]).unwrap();
p.recv();
});
let mut s1 = acceptor.accept().unwrap();
let s2 = s1.clone();
let (p, done) = Chan::new();
spawn(proc() {
let mut s2 = s2;
let mut buf = [0, 0];
s2.read(buf).unwrap();
c2.send(());
done.send(());
});
let mut buf = [0, 0];
s1.read(buf).unwrap();
c.send(());
p.recv();
})
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
iotest!(fn unix_clone_two_write() {
Implement clone() for TCP/UDP/Unix sockets This is part of the overall strategy I would like to take when approaching issue #11165. The only two I/O objects that reasonably want to be "split" are the network stream objects. Everything else can be "split" by just creating another version. The initial idea I had was the literally split the object into a reader and a writer half, but that would just introduce lots of clutter with extra interfaces that were a little unnnecssary, or it would return a ~Reader and a ~Writer which means you couldn't access things like the remote peer name or local socket name. The solution I found to be nicer was to just clone the stream itself. The clone is just a clone of the handle, nothing fancy going on at the kernel level. Conceptually I found this very easy to wrap my head around (everything else supports clone()), and it solved the "split" problem at the same time. The cloning support is pretty specific per platform/lib combination: * native/win32 - uses some specific WSA apis to clone the SOCKET handle * native/unix - uses dup() to get another file descriptor * green/all - This is where things get interesting. When we support full clones of a handle, this implies that we're allowing simultaneous writes and reads to happen. It turns out that libuv doesn't support two simultaneous reads or writes of the same object. It does support *one* read and *one* write at the same time, however. Some extra infrastructure was added to just block concurrent writers/readers until the previous read/write operation was completed. I've added tests to the tcp/unix modules to make sure that this functionality is supported everywhere.
2014-01-22 19:32:16 -08:00
let addr = next_test_unix();
let mut acceptor = UnixListener::bind(&addr).listen();
spawn(proc() {
let mut s = UnixStream::connect(&addr);
let mut buf = [0, 1];
s.read(buf).unwrap();
s.read(buf).unwrap();
});
let mut s1 = acceptor.accept().unwrap();
let s2 = s1.clone();
let (p, done) = Chan::new();
spawn(proc() {
let mut s2 = s2;
s2.write([1]).unwrap();
done.send(());
});
s1.write([2]).unwrap();
p.recv();
})
}