Run Format

Source file src/runtime/netpoll_solaris.go

Documentation: runtime

  // Copyright 2014 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  package runtime
  import "unsafe"
  // Solaris runtime-integrated network poller.
  // Solaris uses event ports for scalable network I/O. Event
  // ports are level-triggered, unlike epoll and kqueue which
  // can be configured in both level-triggered and edge-triggered
  // mode. Level triggering means we have to keep track of a few things
  // ourselves. After we receive an event for a file descriptor,
  // it's our responsibility to ask again to be notified for future
  // events for that descriptor. When doing this we must keep track of
  // what kind of events the goroutines are currently interested in,
  // for example a fd may be open both for reading and writing.
  // A description of the high level operation of this code
  // follows. Networking code will get a file descriptor by some means
  // and will register it with the netpolling mechanism by a code path
  // that eventually calls runtime·netpollopen. runtime·netpollopen
  // calls port_associate with an empty event set. That means that we
  // will not receive any events at this point. The association needs
  // to be done at this early point because we need to process the I/O
  // readiness notification at some point in the future. If I/O becomes
  // ready when nobody is listening, when we finally care about it,
  // nobody will tell us anymore.
  // Beside calling runtime·netpollopen, the networking code paths
  // will call runtime·netpollarm each time goroutines are interested
  // in doing network I/O. Because now we know what kind of I/O we
  // are interested in (reading/writing), we can call port_associate
  // passing the correct type of event set (POLLIN/POLLOUT). As we made
  // sure to have already associated the file descriptor with the port,
  // when we now call port_associate, we will unblock the main poller
  // loop (in runtime·netpoll) right away if the socket is actually
  // ready for I/O.
  // The main poller loop runs in its own thread waiting for events
  // using port_getn. When an event happens, it will tell the scheduler
  // about it using runtime·netpollready. Besides doing this, it must
  // also re-associate the events that were not part of this current
  // notification with the file descriptor. Failing to do this would
  // mean each notification will prevent concurrent code using the
  // same file descriptor in parallel.
  // The logic dealing with re-associations is encapsulated in
  // runtime·netpollupdate. This function takes care to associate the
  // descriptor only with the subset of events that were previously
  // part of the association, except the one that just happened. We
  // can't re-associate with that right away, because event ports
  // are level triggered so it would cause a busy loop. Instead, that
  // association is effected only by the runtime·netpollarm code path,
  // when Go code actually asks for I/O.
  // The open and arming mechanisms are serialized using the lock
  // inside PollDesc. This is required because the netpoll loop runs
  // asynchronously in respect to other Go code and by the time we get
  // to call port_associate to update the association in the loop, the
  // file descriptor might have been closed and reopened already. The
  // lock allows runtime·netpollupdate to be called synchronously from
  // the loop thread while preventing other threads operating to the
  // same PollDesc, so once we unblock in the main loop, until we loop
  // again we know for sure we are always talking about the same file
  // descriptor and can safely access the data we want (the event set).
  //go:cgo_import_dynamic libc_port_create port_create "libc.so"
  //go:cgo_import_dynamic libc_port_associate port_associate "libc.so"
  //go:cgo_import_dynamic libc_port_dissociate port_dissociate "libc.so"
  //go:cgo_import_dynamic libc_port_getn port_getn "libc.so"
  //go:linkname libc_port_create libc_port_create
  //go:linkname libc_port_associate libc_port_associate
  //go:linkname libc_port_dissociate libc_port_dissociate
  //go:linkname libc_port_getn libc_port_getn
  var (
  	libc_port_getn libcFunc
  func errno() int32 {
  	return *getg().m.perrno
  func fcntl(fd, cmd int32, arg uintptr) int32 {
  	return int32(sysvicall3(&libc_fcntl, uintptr(fd), uintptr(cmd), arg))
  func port_create() int32 {
  	return int32(sysvicall0(&libc_port_create))
  func port_associate(port, source int32, object uintptr, events uint32, user uintptr) int32 {
  	return int32(sysvicall5(&libc_port_associate, uintptr(port), uintptr(source), object, uintptr(events), user))
  func port_dissociate(port, source int32, object uintptr) int32 {
  	return int32(sysvicall3(&libc_port_dissociate, uintptr(port), uintptr(source), object))
  func port_getn(port int32, evs *portevent, max uint32, nget *uint32, timeout *timespec) int32 {
  	return int32(sysvicall5(&libc_port_getn, uintptr(port), uintptr(unsafe.Pointer(evs)), uintptr(max), uintptr(unsafe.Pointer(nget)), uintptr(unsafe.Pointer(timeout))))
  var portfd int32 = -1
  func netpollinit() {
  	portfd = port_create()
  	if portfd >= 0 {
  		fcntl(portfd, _F_SETFD, _FD_CLOEXEC)
  	print("runtime: port_create failed (errno=", errno(), ")\n")
  	throw("runtime: netpollinit failed")
  func netpolldescriptor() uintptr {
  	return uintptr(portfd)
  func netpollopen(fd uintptr, pd *pollDesc) int32 {
  	// We don't register for any specific type of events yet, that's
  	// netpollarm's job. We merely ensure we call port_associate before
  	// asynchronous connect/accept completes, so when we actually want
  	// to do any I/O, the call to port_associate (from netpollarm,
  	// with the interested event set) will unblock port_getn right away
  	// because of the I/O readiness notification.
  	pd.user = 0
  	r := port_associate(portfd, _PORT_SOURCE_FD, fd, 0, uintptr(unsafe.Pointer(pd)))
  	return r
  func netpollclose(fd uintptr) int32 {
  	return port_dissociate(portfd, _PORT_SOURCE_FD, fd)
  // Updates the association with a new set of interested events. After
  // this call, port_getn will return one and only one event for that
  // particular descriptor, so this function needs to be called again.
  func netpollupdate(pd *pollDesc, set, clear uint32) {
  	if pd.closing {
  	old := pd.user
  	events := (old & ^clear) | set
  	if old == events {
  	if events != 0 && port_associate(portfd, _PORT_SOURCE_FD, pd.fd, events, uintptr(unsafe.Pointer(pd))) != 0 {
  		print("runtime: port_associate failed (errno=", errno(), ")\n")
  		throw("runtime: netpollupdate failed")
  	pd.user = events
  // subscribe the fd to the port such that port_getn will return one event.
  func netpollarm(pd *pollDesc, mode int) {
  	switch mode {
  	case 'r':
  		netpollupdate(pd, _POLLIN, 0)
  	case 'w':
  		netpollupdate(pd, _POLLOUT, 0)
  		throw("runtime: bad mode")
  // polls for ready network connections
  // returns list of goroutines that become runnable
  func netpoll(block bool) *g {
  	if portfd == -1 {
  		return nil
  	var wait *timespec
  	var zero timespec
  	if !block {
  		wait = &zero
  	var events [128]portevent
  	var n uint32 = 1
  	if port_getn(portfd, &events[0], uint32(len(events)), &n, wait) < 0 {
  		if e := errno(); e != _EINTR {
  			print("runtime: port_getn on fd ", portfd, " failed (errno=", e, ")\n")
  			throw("runtime: netpoll failed")
  		goto retry
  	var gp guintptr
  	for i := 0; i < int(n); i++ {
  		ev := &events[i]
  		if ev.portev_events == 0 {
  		pd := (*pollDesc)(unsafe.Pointer(ev.portev_user))
  		var mode, clear int32
  		if (ev.portev_events & (_POLLIN | _POLLHUP | _POLLERR)) != 0 {
  			mode += 'r'
  			clear |= _POLLIN
  		if (ev.portev_events & (_POLLOUT | _POLLHUP | _POLLERR)) != 0 {
  			mode += 'w'
  			clear |= _POLLOUT
  		// To effect edge-triggered events, we need to be sure to
  		// update our association with whatever events were not
  		// set with the event. For example if we are registered
  		// for POLLIN|POLLOUT, and we get POLLIN, besides waking
  		// the goroutine interested in POLLIN we have to not forget
  		// about the one interested in POLLOUT.
  		if clear != 0 {
  			netpollupdate(pd, 0, uint32(clear))
  		if mode != 0 {
  			netpollready(&gp, pd, mode)
  	if block && gp == 0 {
  		goto retry
  	return gp.ptr()

View as plain text