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

377 lines
11 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.
#[allow(missing_doc)];
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;
2013-06-19 17:23:55 -07:00
use result::{Ok, Err};
2013-11-10 22:46:32 -08:00
use io::net::ip::SocketAddr;
use io::{Reader, Writer, IoResult};
use rt::rtio::{RtioSocket, RtioUdpSocket, IoFactory, LocalIo};
2013-06-19 17:23:55 -07:00
pub struct UdpSocket {
priv obj: ~RtioUdpSocket
}
2013-06-19 17:23:55 -07:00
impl UdpSocket {
pub fn bind(addr: SocketAddr) -> IoResult<UdpSocket> {
LocalIo::maybe_raise(|io| {
io.udp_bind(addr).map(|s| UdpSocket { obj: s })
})
2013-06-19 17:23:55 -07:00
}
pub fn recvfrom(&mut self, buf: &mut [u8]) -> IoResult<(uint, SocketAddr)> {
self.obj.recvfrom(buf)
2013-06-19 17:23:55 -07:00
}
pub fn sendto(&mut self, buf: &[u8], dst: SocketAddr) -> IoResult<()> {
self.obj.sendto(buf, dst)
2013-06-19 17:23:55 -07:00
}
pub fn connect(self, other: SocketAddr) -> UdpStream {
UdpStream { socket: self, connected_to: other }
2013-06-19 17:23:55 -07:00
}
pub fn socket_name(&mut self) -> IoResult<SocketAddr> {
self.obj.socket_name()
}
2013-06-19 17:23:55 -07:00
}
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 UdpSocket {
/// Creates a new handle to this UDP socket, allowing for simultaneous reads
/// and writes of the socket.
///
/// The underlying UDP socket will not be closed until all handles to the
/// socket have been deallocated. Two concurrent reads will not receive the
/// same data. Instead, the first read will receive the first packet
/// received, and the second read will receive the second packet.
fn clone(&self) -> UdpSocket {
UdpSocket { obj: self.obj.clone() }
}
}
2013-06-17 12:34:58 -07:00
pub struct UdpStream {
priv socket: UdpSocket,
priv connected_to: SocketAddr
2013-06-17 12:34:58 -07:00
}
impl UdpStream {
pub fn as_socket<T>(&mut self, f: |&mut UdpSocket| -> T) -> T {
f(&mut self.socket)
}
2013-06-17 12:34:58 -07:00
pub fn disconnect(self) -> UdpSocket { self.socket }
}
impl Reader for UdpStream {
fn read(&mut self, buf: &mut [u8]) -> IoResult<uint> {
let peer = self.connected_to;
self.as_socket(|sock| {
2013-06-25 11:43:40 -07:00
match sock.recvfrom(buf) {
Ok((_nread, src)) if src != peer => Ok(0),
Ok((nread, _src)) => Ok(nread),
Err(e) => Err(e),
2013-06-25 11:43:40 -07:00
}
})
2013-06-19 17:23:55 -07:00
}
}
impl Writer for UdpStream {
fn write(&mut self, buf: &[u8]) -> IoResult<()> {
let connected_to = self.connected_to;
self.as_socket(|sock| sock.sendto(buf, connected_to))
}
}
2013-06-25 11:43:40 -07:00
#[cfg(test)]
mod test {
use super::*;
2014-01-05 13:05:02 -05:00
use io::net::ip::{SocketAddr};
2013-06-25 11:43:40 -07:00
// FIXME #11530 this fails on android because tests are run as root
2013-12-28 16:40:15 -08:00
iotest!(fn bind_error() {
2014-01-30 14:10:53 -08:00
let addr = SocketAddr { ip: Ipv4Addr(0, 0, 0, 0), port: 1 };
match UdpSocket::bind(addr) {
Ok(..) => fail!(),
Err(e) => assert_eq!(e.kind, PermissionDenied),
}
} #[ignore(cfg(windows))] #[ignore(cfg(target_os = "android"))])
2013-06-25 11:43:40 -07:00
2013-12-28 16:40:15 -08:00
iotest!(fn socket_smoke_test_ip4() {
let server_ip = next_test_ip4();
let client_ip = next_test_ip4();
let (port, chan) = Chan::new();
2013-12-28 16:40:15 -08:00
let (port2, chan2) = Chan::new();
2013-06-25 11:43:40 -07:00
spawn(proc() {
2013-12-05 18:19:06 -08:00
match UdpSocket::bind(client_ip) {
2014-01-30 14:10:53 -08:00
Ok(ref mut client) => {
2013-12-05 18:19:06 -08:00
port.recv();
2014-01-30 14:10:53 -08:00
client.sendto([99], server_ip).unwrap()
2013-06-25 11:43:40 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-06-25 11:43:40 -07:00
}
2013-12-28 16:40:15 -08:00
chan2.send(());
});
2013-06-25 11:43:40 -07:00
match UdpSocket::bind(server_ip) {
2014-01-30 14:10:53 -08:00
Ok(ref mut server) => {
chan.send(());
let mut buf = [0];
match server.recvfrom(buf) {
2014-01-30 14:10:53 -08:00
Ok((nread, src)) => {
assert_eq!(nread, 1);
assert_eq!(buf[0], 99);
assert_eq!(src, client_ip);
2013-07-02 16:40:57 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-07-02 16:40:57 -07:00
}
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
}
2013-12-28 16:40:15 -08:00
port2.recv();
})
2013-07-02 16:40:57 -07:00
2013-12-28 16:40:15 -08:00
iotest!(fn socket_smoke_test_ip6() {
let server_ip = next_test_ip6();
let client_ip = next_test_ip6();
let (port, chan) = Chan::<()>::new();
spawn(proc() {
2013-12-05 18:19:06 -08:00
match UdpSocket::bind(client_ip) {
2014-01-30 14:10:53 -08:00
Ok(ref mut client) => {
2013-12-05 18:19:06 -08:00
port.recv();
2014-01-30 14:10:53 -08:00
client.sendto([99], server_ip).unwrap()
2013-07-02 16:40:57 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-07-02 16:40:57 -07:00
}
});
2013-07-02 16:40:57 -07:00
match UdpSocket::bind(server_ip) {
2014-01-30 14:10:53 -08:00
Ok(ref mut server) => {
chan.send(());
let mut buf = [0];
match server.recvfrom(buf) {
2014-01-30 14:10:53 -08:00
Ok((nread, src)) => {
assert_eq!(nread, 1);
assert_eq!(buf[0], 99);
assert_eq!(src, client_ip);
2013-07-02 16:40:57 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-07-02 16:40:57 -07:00
}
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
}
2013-12-28 16:40:15 -08:00
})
2013-12-28 16:40:15 -08:00
iotest!(fn stream_smoke_test_ip4() {
let server_ip = next_test_ip4();
let client_ip = next_test_ip4();
let (port, chan) = Chan::new();
2013-12-28 16:40:15 -08:00
let (port2, chan2) = Chan::new();
2013-07-02 16:40:57 -07:00
spawn(proc() {
2013-12-05 18:19:06 -08:00
match UdpSocket::bind(client_ip) {
2014-01-30 14:10:53 -08:00
Ok(client) => {
2013-12-05 18:19:06 -08:00
let client = ~client;
let mut stream = client.connect(server_ip);
port.recv();
2014-01-30 14:10:53 -08:00
stream.write([99]).unwrap();
2013-07-02 16:40:57 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-07-02 16:40:57 -07:00
}
2013-12-28 16:40:15 -08:00
chan2.send(());
});
2013-07-02 16:40:57 -07:00
match UdpSocket::bind(server_ip) {
2014-01-30 14:10:53 -08:00
Ok(server) => {
let server = ~server;
let mut stream = server.connect(client_ip);
chan.send(());
let mut buf = [0];
match stream.read(buf) {
2014-01-30 14:10:53 -08:00
Ok(nread) => {
assert_eq!(nread, 1);
assert_eq!(buf[0], 99);
2013-06-25 11:43:40 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-06-25 11:43:40 -07:00
}
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
}
2013-12-28 16:40:15 -08:00
port2.recv();
})
2013-12-28 16:40:15 -08:00
iotest!(fn stream_smoke_test_ip6() {
let server_ip = next_test_ip6();
let client_ip = next_test_ip6();
let (port, chan) = Chan::new();
2013-12-28 16:40:15 -08:00
let (port2, chan2) = Chan::new();
2013-06-25 11:43:40 -07:00
spawn(proc() {
2013-12-05 18:19:06 -08:00
match UdpSocket::bind(client_ip) {
2014-01-30 14:10:53 -08:00
Ok(client) => {
2013-12-05 18:19:06 -08:00
let client = ~client;
let mut stream = client.connect(server_ip);
port.recv();
2014-01-30 14:10:53 -08:00
stream.write([99]).unwrap();
2013-06-25 11:43:40 -07:00
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
2013-06-25 11:43:40 -07:00
}
2013-12-28 16:40:15 -08:00
chan2.send(());
});
match UdpSocket::bind(server_ip) {
2014-01-30 14:10:53 -08:00
Ok(server) => {
let server = ~server;
let mut stream = server.connect(client_ip);
chan.send(());
let mut buf = [0];
match stream.read(buf) {
2014-01-30 14:10:53 -08:00
Ok(nread) => {
assert_eq!(nread, 1);
assert_eq!(buf[0], 99);
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
}
}
2014-01-30 14:10:53 -08:00
Err(..) => fail!()
}
2013-12-28 16:40:15 -08:00
port2.recv();
})
2013-07-26 05:02:53 -04:00
2013-12-28 16:40:15 -08:00
pub fn socket_name(addr: SocketAddr) {
let server = UdpSocket::bind(addr);
2013-07-26 05:02:53 -04:00
2014-01-30 14:10:53 -08:00
assert!(server.is_ok());
let mut server = server.unwrap();
2013-07-26 05:02:53 -04:00
// Make sure socket_name gives
// us the socket we binded to.
let so_name = server.socket_name();
2014-01-30 14:10:53 -08:00
assert!(so_name.is_ok());
assert_eq!(addr, so_name.unwrap());
2013-07-26 05:02:53 -04:00
}
2013-12-28 16:40:15 -08:00
iotest!(fn socket_name_ip4() {
2013-07-26 05:02:53 -04:00
socket_name(next_test_ip4());
2013-12-28 16:40:15 -08:00
})
2013-07-26 05:02:53 -04:00
2013-12-28 16:40:15 -08:00
iotest!(fn socket_name_ip6() {
2013-07-26 05:02:53 -04:00
socket_name(next_test_ip6());
2013-12-28 16:40:15 -08:00
})
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 udp_clone_smoke() {
let addr1 = next_test_ip4();
let addr2 = next_test_ip4();
let mut sock1 = UdpSocket::bind(addr1).unwrap();
let sock2 = UdpSocket::bind(addr2).unwrap();
spawn(proc() {
let mut sock2 = sock2;
let mut buf = [0, 0];
assert_eq!(sock2.recvfrom(buf), Ok((1, addr1)));
assert_eq!(buf[0], 1);
sock2.sendto([2], addr1).unwrap();
});
let sock3 = sock1.clone();
let (p1, c1) = Chan::new();
let (p2, c2) = Chan::new();
spawn(proc() {
let mut sock3 = sock3;
p1.recv();
sock3.sendto([1], addr2).unwrap();
c2.send(());
});
c1.send(());
let mut buf = [0, 0];
assert_eq!(sock1.recvfrom(buf), Ok((1, addr2)));
p2.recv();
})
iotest!(fn udp_clone_two_read() {
let addr1 = next_test_ip4();
let addr2 = next_test_ip4();
let mut sock1 = UdpSocket::bind(addr1).unwrap();
let sock2 = UdpSocket::bind(addr2).unwrap();
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 sock2 = sock2;
sock2.sendto([1], addr1).unwrap();
p.recv();
sock2.sendto([2], addr1).unwrap();
p.recv();
});
let sock3 = sock1.clone();
let (p, done) = Chan::new();
spawn(proc() {
let mut sock3 = sock3;
let mut buf = [0, 0];
sock3.recvfrom(buf).unwrap();
c2.send(());
done.send(());
});
let mut buf = [0, 0];
sock1.recvfrom(buf).unwrap();
c.send(());
p.recv();
})
iotest!(fn udp_clone_two_write() {
let addr1 = next_test_ip4();
let addr2 = next_test_ip4();
let mut sock1 = UdpSocket::bind(addr1).unwrap();
let sock2 = UdpSocket::bind(addr2).unwrap();
2014-02-11 16:33:34 -08:00
let (p, c) = Chan::new();
let (serv_port, serv_chan) = 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
spawn(proc() {
let mut sock2 = sock2;
let mut buf = [0, 1];
p.recv();
match sock2.recvfrom(buf) {
Ok(..) => {}
Err(e) => fail!("failed receive: {}", e),
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
}
serv_chan.send(());
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 sock3 = sock1.clone();
let (p, done) = Chan::new();
let c2 = c.clone();
spawn(proc() {
let mut sock3 = sock3;
match sock3.sendto([1], addr2) {
Ok(..) => { let _ = c2.try_send(()); }
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
Err(..) => {}
}
done.send(());
});
match sock1.sendto([2], addr2) {
Ok(..) => { let _ = c.try_send(()); }
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
Err(..) => {}
}
drop(c);
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
p.recv();
serv_port.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
})
2013-06-25 11:43:40 -07:00
}