caddy/listen_unix.go

243 lines
7.1 KiB
Go
Raw Normal View History

// Copyright 2015 Matthew Holt and The Caddy Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// Even though the filename ends in _unix.go, we still have to specify the
// build constraint here, because the filename convention only works for
// literal GOOS values, and "unix" is a shortcut unique to build tags.
//go:build unix && !solaris
package caddy
import (
"context"
"errors"
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
"io"
"io/fs"
"net"
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
"os"
"sync/atomic"
"syscall"
"go.uber.org/zap"
"golang.org/x/sys/unix"
)
// reuseUnixSocket copies and reuses the unix domain socket (UDS) if we already
// have it open; if not, unlink it so we can have it. No-op if not a unix network.
func reuseUnixSocket(network, addr string) (any, error) {
if !IsUnixNetwork(network) {
return nil, nil
}
socketKey := listenerKey(network, addr)
socket, exists := unixSockets[socketKey]
if exists {
// make copy of file descriptor
socketFile, err := socket.File() // does dup() deep down
if err != nil {
return nil, err
}
// use copied fd to make new Listener or PacketConn, then replace
// it in the map so that future copies always come from the most
// recent fd (as the previous ones will be closed, and we'd get
// "use of closed network connection" errors) -- note that we
// preserve the *pointer* to the counter (not just the value) so
// that all socket wrappers will refer to the same value
switch unixSocket := socket.(type) {
case *unixListener:
ln, err := net.FileListener(socketFile)
if err != nil {
return nil, err
}
atomic.AddInt32(unixSocket.count, 1)
unixSockets[socketKey] = &unixListener{ln.(*net.UnixListener), socketKey, unixSocket.count}
case *unixConn:
pc, err := net.FilePacketConn(socketFile)
if err != nil {
return nil, err
}
atomic.AddInt32(unixSocket.count, 1)
unixSockets[socketKey] = &unixConn{pc.(*net.UnixConn), addr, socketKey, unixSocket.count}
}
return unixSockets[socketKey], nil
}
// from what I can tell after some quick research, it's quite common for programs to
// leave their socket file behind after they close, so the typical pattern is to
// unlink it before you bind to it -- this is often crucial if the last program using
// it was killed forcefully without a chance to clean up the socket, but there is a
// race, as the comment in net.UnixListener.close() explains... oh well, I guess?
if err := syscall.Unlink(addr); err != nil && !errors.Is(err, fs.ErrNotExist) {
return nil, err
}
return nil, nil
}
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
func listenReusable(ctx context.Context, lnKey string, network, address string, config net.ListenConfig) (any, error) {
// wrap any Control function set by the user so we can also add our reusePort control without clobbering theirs
oldControl := config.Control
config.Control = func(network, address string, c syscall.RawConn) error {
if oldControl != nil {
if err := oldControl(network, address, c); err != nil {
return err
}
}
return reusePort(network, address, c)
}
// even though SO_REUSEPORT lets us bind the socket multiple times,
// we still put it in the listenerPool so we can count how many
// configs are using this socket; necessary to ensure we can know
// whether to enforce shutdown delays, for example (see #5393).
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
var ln io.Closer
var err error
switch network {
case "udp", "udp4", "udp6", "unixgram":
ln, err = config.ListenPacket(ctx, network, address)
default:
ln, err = config.Listen(ctx, network, address)
}
if err == nil {
listenerPool.LoadOrStore(lnKey, nil)
}
2023-06-22 07:47:23 +08:00
// if new listener is a unix socket, make sure we can reuse it later
// (we do our own "unlink on close" -- not required, but more tidy)
one := int32(1)
if unix, ok := ln.(*net.UnixListener); ok {
unix.SetUnlinkOnClose(false)
ln = &unixListener{unix, lnKey, &one}
unixSockets[lnKey] = ln.(*unixListener)
}
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
// TODO: Not 100% sure this is necessary, but we do this for net.UnixListener in listen_unix.go, so...
if unix, ok := ln.(*net.UnixConn); ok {
ln = &unixConn{unix, address, lnKey, &one}
unixSockets[lnKey] = ln.(*unixConn)
}
// lightly wrap the listener so that when it is closed,
// we can decrement the usage pool counter
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
switch specificLn := ln.(type) {
case net.Listener:
return deleteListener{specificLn, lnKey}, err
case net.PacketConn:
return deletePacketConn{specificLn, lnKey}, err
}
// other types, I guess we just return them directly
return ln, err
}
// reusePort sets SO_REUSEPORT. Ineffective for unix sockets.
func reusePort(network, address string, conn syscall.RawConn) error {
if IsUnixNetwork(network) {
return nil
}
return conn.Control(func(descriptor uintptr) {
if err := unix.SetsockoptInt(int(descriptor), unix.SOL_SOCKET, unixSOREUSEPORT, 1); err != nil {
Log().Error("setting SO_REUSEPORT",
zap.String("network", network),
zap.String("address", address),
zap.Uintptr("descriptor", descriptor),
zap.Error(err))
}
})
}
2023-06-22 07:59:54 +08:00
type unixListener struct {
*net.UnixListener
mapKey string
count *int32 // accessed atomically
}
func (uln *unixListener) Close() error {
newCount := atomic.AddInt32(uln.count, -1)
if newCount == 0 {
defer func() {
addr := uln.Addr().String()
unixSocketsMu.Lock()
delete(unixSockets, uln.mapKey)
unixSocketsMu.Unlock()
_ = syscall.Unlink(addr)
}()
}
return uln.UnixListener.Close()
}
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
type unixConn struct {
*net.UnixConn
filename string
mapKey string
count *int32 // accessed atomically
}
func (uc *unixConn) Close() error {
newCount := atomic.AddInt32(uc.count, -1)
if newCount == 0 {
defer func() {
unixSocketsMu.Lock()
delete(unixSockets, uc.mapKey)
unixSocketsMu.Unlock()
_ = syscall.Unlink(uc.filename)
}()
}
return uc.UnixConn.Close()
}
func (uc *unixConn) Unwrap() net.PacketConn {
return uc.UnixConn
}
// unixSockets keeps track of the currently-active unix sockets
// so we can transfer their FDs gracefully during reloads.
var unixSockets = make(map[string]interface {
File() (*os.File, error)
})
// deleteListener is a type that simply deletes itself
// from the listenerPool when it closes. It is used
// solely for the purpose of reference counting (i.e.
// counting how many configs are using a given socket).
type deleteListener struct {
net.Listener
lnKey string
}
func (dl deleteListener) Close() error {
_, _ = listenerPool.Delete(dl.lnKey)
return dl.Listener.Close()
}
core: Apply SO_REUSEPORT to UDP sockets (#5725) * core: Apply SO_REUSEPORT to UDP sockets For some reason, 10 months ago when I implemented SO_REUSEPORT for TCP, I didn't realize, or forgot, that it can be used for UDP too. It is a much better solution than using deadline hacks to reuse a socket, at least for TCP. Then https://github.com/mholt/caddy-l4/issues/132 was posted, in which we see that UDP servers never actually stopped when the L4 app was stopped. I verified this using this command: $ nc -u 127.0.0.1 55353 combined with POSTing configs to the /load admin endpoint (which alternated between an echo server and a proxy server so I could tell which config was being used). I refactored the code to use SO_REUSEPORT for UDP, but of course we still need graceful reloads on all platforms, not just Unix, so I also implemented a deadline hack similar to what we used for TCP before. That implementation for TCP was not perfect, possibly having a logical (not data) race condition; but for UDP so far it seems to be working. Verified the same way I verified that SO_REUSEPORT works. I think this code is slightly cleaner and I'm fairly confident this code is effective. * Check error * Fix return * Fix var name * implement Unwrap interface and clean up * move unix packet conn to platform specific file * implement Unwrap for unix packet conn * Move sharedPacketConn into proper file * Fix Windows * move sharedPacketConn and fakeClosePacketConn to proper file --------- Co-authored-by: Weidi Deng <weidi_deng@icloud.com>
2023-10-17 12:17:32 +08:00
// deletePacketConn is like deleteListener, but
// for net.PacketConns.
type deletePacketConn struct {
net.PacketConn
lnKey string
}
func (dl deletePacketConn) Close() error {
_, _ = listenerPool.Delete(dl.lnKey)
return dl.PacketConn.Close()
}
func (dl deletePacketConn) Unwrap() net.PacketConn {
return dl.PacketConn
}