// Copyright 2016 The Chromium Authors. All rights reserved. // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. #include "mojo/edk/system/ports/node.h" #include <string.h> #include <utility> #include "base/atomicops.h" #include "base/logging.h" #include "base/memory/ref_counted.h" #include "base/synchronization/lock.h" #include "mojo/edk/system/ports/node_delegate.h" namespace mojo { namespace edk { namespace ports { namespace { int DebugError(const char* message, int error_code) { CHECK(false) << "Oops: " << message; return error_code; } #define OOPS(x) DebugError(#x, x) bool CanAcceptMoreMessages(const Port* port) { // Have we already doled out the last message (i.e., do we expect to NOT // receive further messages)? uint64_t next_sequence_num = port->message_queue.next_sequence_num(); if (port->state == Port::kClosed) return false; if (port->peer_closed || port->remove_proxy_on_last_message) { if (port->last_sequence_num_to_receive == next_sequence_num - 1) return false; } return true; } } // namespace class Node::LockedPort { public: explicit LockedPort(Port* port) : port_(port) { port_->lock.AssertAcquired(); } Port* get() const { return port_; } Port* operator->() const { return port_; } private: Port* const port_; }; Node::Node(const NodeName& name, NodeDelegate* delegate) : name_(name), delegate_(delegate) { } Node::~Node() { if (!ports_.empty()) DLOG(WARNING) << "Unclean shutdown for node " << name_; } bool Node::CanShutdownCleanly(ShutdownPolicy policy) { base::AutoLock ports_lock(ports_lock_); if (policy == ShutdownPolicy::DONT_ALLOW_LOCAL_PORTS) { #if DCHECK_IS_ON() for (auto entry : ports_) { DVLOG(2) << "Port " << entry.first << " referencing node " << entry.second->peer_node_name << " is blocking shutdown of " << "node " << name_ << " (state=" << entry.second->state << ")"; } #endif return ports_.empty(); } DCHECK_EQ(policy, ShutdownPolicy::ALLOW_LOCAL_PORTS); // NOTE: This is not efficient, though it probably doesn't need to be since // relatively few ports should be open during shutdown and shutdown doesn't // need to be blazingly fast. bool can_shutdown = true; for (auto entry : ports_) { base::AutoLock lock(entry.second->lock); if (entry.second->peer_node_name != name_ && entry.second->state != Port::kReceiving) { can_shutdown = false; #if DCHECK_IS_ON() DVLOG(2) << "Port " << entry.first << " referencing node " << entry.second->peer_node_name << " is blocking shutdown of " << "node " << name_ << " (state=" << entry.second->state << ")"; #else // Exit early when not debugging. break; #endif } } return can_shutdown; } int Node::GetPort(const PortName& port_name, PortRef* port_ref) { scoped_refptr<Port> port = GetPort(port_name); if (!port) return ERROR_PORT_UNKNOWN; *port_ref = PortRef(port_name, std::move(port)); return OK; } int Node::CreateUninitializedPort(PortRef* port_ref) { PortName port_name; delegate_->GenerateRandomPortName(&port_name); scoped_refptr<Port> port(new Port(kInitialSequenceNum, kInitialSequenceNum)); int rv = AddPortWithName(port_name, port); if (rv != OK) return rv; *port_ref = PortRef(port_name, std::move(port)); return OK; } int Node::InitializePort(const PortRef& port_ref, const NodeName& peer_node_name, const PortName& peer_port_name) { Port* port = port_ref.port(); { base::AutoLock lock(port->lock); if (port->state != Port::kUninitialized) return ERROR_PORT_STATE_UNEXPECTED; port->state = Port::kReceiving; port->peer_node_name = peer_node_name; port->peer_port_name = peer_port_name; } delegate_->PortStatusChanged(port_ref); return OK; } int Node::CreatePortPair(PortRef* port0_ref, PortRef* port1_ref) { int rv; rv = CreateUninitializedPort(port0_ref); if (rv != OK) return rv; rv = CreateUninitializedPort(port1_ref); if (rv != OK) return rv; rv = InitializePort(*port0_ref, name_, port1_ref->name()); if (rv != OK) return rv; rv = InitializePort(*port1_ref, name_, port0_ref->name()); if (rv != OK) return rv; return OK; } int Node::SetUserData(const PortRef& port_ref, scoped_refptr<UserData> user_data) { Port* port = port_ref.port(); base::AutoLock lock(port->lock); if (port->state == Port::kClosed) return ERROR_PORT_STATE_UNEXPECTED; port->user_data = std::move(user_data); return OK; } int Node::GetUserData(const PortRef& port_ref, scoped_refptr<UserData>* user_data) { Port* port = port_ref.port(); base::AutoLock lock(port->lock); if (port->state == Port::kClosed) return ERROR_PORT_STATE_UNEXPECTED; *user_data = port->user_data; return OK; } int Node::ClosePort(const PortRef& port_ref) { std::deque<PortName> referenced_port_names; ObserveClosureEventData data; NodeName peer_node_name; PortName peer_port_name; Port* port = port_ref.port(); { // We may need to erase the port, which requires ports_lock_ to be held, // but ports_lock_ must be acquired before any individual port locks. base::AutoLock ports_lock(ports_lock_); base::AutoLock lock(port->lock); if (port->state == Port::kUninitialized) { // If the port was not yet initialized, there's nothing interesting to do. ErasePort_Locked(port_ref.name()); return OK; } if (port->state != Port::kReceiving) return ERROR_PORT_STATE_UNEXPECTED; port->state = Port::kClosed; // We pass along the sequence number of the last message sent from this // port to allow the peer to have the opportunity to consume all inbound // messages before notifying the embedder that this port is closed. data.last_sequence_num = port->next_sequence_num_to_send - 1; peer_node_name = port->peer_node_name; peer_port_name = port->peer_port_name; // If the port being closed still has unread messages, then we need to take // care to close those ports so as to avoid leaking memory. port->message_queue.GetReferencedPorts(&referenced_port_names); ErasePort_Locked(port_ref.name()); } DVLOG(2) << "Sending ObserveClosure from " << port_ref.name() << "@" << name_ << " to " << peer_port_name << "@" << peer_node_name; delegate_->ForwardMessage( peer_node_name, NewInternalMessage(peer_port_name, EventType::kObserveClosure, data)); for (const auto& name : referenced_port_names) { PortRef ref; if (GetPort(name, &ref) == OK) ClosePort(ref); } return OK; } int Node::GetStatus(const PortRef& port_ref, PortStatus* port_status) { Port* port = port_ref.port(); base::AutoLock lock(port->lock); if (port->state != Port::kReceiving) return ERROR_PORT_STATE_UNEXPECTED; port_status->has_messages = port->message_queue.HasNextMessage(); port_status->receiving_messages = CanAcceptMoreMessages(port); port_status->peer_closed = port->peer_closed; return OK; } int Node::GetMessage(const PortRef& port_ref, ScopedMessage* message, MessageFilter* filter) { *message = nullptr; DVLOG(4) << "GetMessage for " << port_ref.name() << "@" << name_; Port* port = port_ref.port(); { base::AutoLock lock(port->lock); // This could also be treated like the port being unknown since the // embedder should no longer be referring to a port that has been sent. if (port->state != Port::kReceiving) return ERROR_PORT_STATE_UNEXPECTED; // Let the embedder get messages until there are no more before reporting // that the peer closed its end. if (!CanAcceptMoreMessages(port)) return ERROR_PORT_PEER_CLOSED; port->message_queue.GetNextMessage(message, filter); } // Allow referenced ports to trigger PortStatusChanged calls. if (*message) { for (size_t i = 0; i < (*message)->num_ports(); ++i) { const PortName& new_port_name = (*message)->ports()[i]; scoped_refptr<Port> new_port = GetPort(new_port_name); DCHECK(new_port) << "Port " << new_port_name << "@" << name_ << " does not exist!"; base::AutoLock lock(new_port->lock); DCHECK(new_port->state == Port::kReceiving); new_port->message_queue.set_signalable(true); } } return OK; } int Node::SendMessage(const PortRef& port_ref, ScopedMessage message) { int rv = SendMessageInternal(port_ref, &message); if (rv != OK) { // If send failed, close all carried ports. Note that we're careful not to // close the sending port itself if it happened to be one of the encoded // ports (an invalid but possible condition.) for (size_t i = 0; i < message->num_ports(); ++i) { if (message->ports()[i] == port_ref.name()) continue; PortRef port; if (GetPort(message->ports()[i], &port) == OK) ClosePort(port); } } return rv; } int Node::AcceptMessage(ScopedMessage message) { const EventHeader* header = GetEventHeader(*message); switch (header->type) { case EventType::kUser: return OnUserMessage(std::move(message)); case EventType::kPortAccepted: return OnPortAccepted(header->port_name); case EventType::kObserveProxy: return OnObserveProxy( header->port_name, *GetEventData<ObserveProxyEventData>(*message)); case EventType::kObserveProxyAck: return OnObserveProxyAck( header->port_name, GetEventData<ObserveProxyAckEventData>(*message)->last_sequence_num); case EventType::kObserveClosure: return OnObserveClosure( header->port_name, GetEventData<ObserveClosureEventData>(*message)->last_sequence_num); case EventType::kMergePort: return OnMergePort(header->port_name, *GetEventData<MergePortEventData>(*message)); } return OOPS(ERROR_NOT_IMPLEMENTED); } int Node::MergePorts(const PortRef& port_ref, const NodeName& destination_node_name, const PortName& destination_port_name) { Port* port = port_ref.port(); MergePortEventData data; { base::AutoLock lock(port->lock); DVLOG(1) << "Sending MergePort from " << port_ref.name() << "@" << name_ << " to " << destination_port_name << "@" << destination_node_name; // Send the port-to-merge over to the destination node so it can be merged // into the port cycle atomically there. data.new_port_name = port_ref.name(); WillSendPort(LockedPort(port), destination_node_name, &data.new_port_name, &data.new_port_descriptor); } delegate_->ForwardMessage( destination_node_name, NewInternalMessage(destination_port_name, EventType::kMergePort, data)); return OK; } int Node::MergeLocalPorts(const PortRef& port0_ref, const PortRef& port1_ref) { Port* port0 = port0_ref.port(); Port* port1 = port1_ref.port(); int rv; { // |ports_lock_| must be held when acquiring overlapping port locks. base::AutoLock ports_lock(ports_lock_); base::AutoLock port0_lock(port0->lock); base::AutoLock port1_lock(port1->lock); DVLOG(1) << "Merging local ports " << port0_ref.name() << "@" << name_ << " and " << port1_ref.name() << "@" << name_; if (port0->state != Port::kReceiving || port1->state != Port::kReceiving) rv = ERROR_PORT_STATE_UNEXPECTED; else rv = MergePorts_Locked(port0_ref, port1_ref); } if (rv != OK) { ClosePort(port0_ref); ClosePort(port1_ref); } return rv; } int Node::LostConnectionToNode(const NodeName& node_name) { // We can no longer send events to the given node. We also can't expect any // PortAccepted events. DVLOG(1) << "Observing lost connection from node " << name_ << " to node " << node_name; DestroyAllPortsWithPeer(node_name, kInvalidPortName); return OK; } int Node::OnUserMessage(ScopedMessage message) { PortName port_name = GetEventHeader(*message)->port_name; const auto* event = GetEventData<UserEventData>(*message); #if DCHECK_IS_ON() std::ostringstream ports_buf; for (size_t i = 0; i < message->num_ports(); ++i) { if (i > 0) ports_buf << ","; ports_buf << message->ports()[i]; } DVLOG(4) << "AcceptMessage " << event->sequence_num << " [ports=" << ports_buf.str() << "] at " << port_name << "@" << name_; #endif scoped_refptr<Port> port = GetPort(port_name); // Even if this port does not exist, cannot receive anymore messages or is // buffering or proxying messages, we still need these ports to be bound to // this node. When the message is forwarded, these ports will get transferred // following the usual method. If the message cannot be accepted, then the // newly bound ports will simply be closed. for (size_t i = 0; i < message->num_ports(); ++i) { int rv = AcceptPort(message->ports()[i], GetPortDescriptors(event)[i]); if (rv != OK) return rv; } bool has_next_message = false; bool message_accepted = false; if (port) { // We may want to forward messages once the port lock is held, so we must // acquire |ports_lock_| first. base::AutoLock ports_lock(ports_lock_); base::AutoLock lock(port->lock); // Reject spurious messages if we've already received the last expected // message. if (CanAcceptMoreMessages(port.get())) { message_accepted = true; port->message_queue.AcceptMessage(std::move(message), &has_next_message); if (port->state == Port::kBuffering) { has_next_message = false; } else if (port->state == Port::kProxying) { has_next_message = false; // Forward messages. We forward messages in sequential order here so // that we maintain the message queue's notion of next sequence number. // That's useful for the proxy removal process as we can tell when this // port has seen all of the messages it is expected to see. int rv = ForwardMessages_Locked(LockedPort(port.get()), port_name); if (rv != OK) return rv; MaybeRemoveProxy_Locked(LockedPort(port.get()), port_name); } } } if (!message_accepted) { DVLOG(2) << "Message not accepted!\n"; // Close all newly accepted ports as they are effectively orphaned. for (size_t i = 0; i < message->num_ports(); ++i) { PortRef port_ref; if (GetPort(message->ports()[i], &port_ref) == OK) { ClosePort(port_ref); } else { DLOG(WARNING) << "Cannot close non-existent port!\n"; } } } else if (has_next_message) { PortRef port_ref(port_name, port); delegate_->PortStatusChanged(port_ref); } return OK; } int Node::OnPortAccepted(const PortName& port_name) { scoped_refptr<Port> port = GetPort(port_name); if (!port) return ERROR_PORT_UNKNOWN; DVLOG(2) << "PortAccepted at " << port_name << "@" << name_ << " pointing to " << port->peer_port_name << "@" << port->peer_node_name; return BeginProxying(PortRef(port_name, std::move(port))); } int Node::OnObserveProxy(const PortName& port_name, const ObserveProxyEventData& event) { if (port_name == kInvalidPortName) { // An ObserveProxy with an invalid target port name is a broadcast used to // inform ports when their peer (which was itself a proxy) has become // defunct due to unexpected node disconnection. // // Receiving ports affected by this treat it as equivalent to peer closure. // Proxies affected by this can be removed and will in turn broadcast their // own death with a similar message. CHECK_EQ(event.proxy_to_node_name, kInvalidNodeName); CHECK_EQ(event.proxy_to_port_name, kInvalidPortName); DestroyAllPortsWithPeer(event.proxy_node_name, event.proxy_port_name); return OK; } // The port may have already been closed locally, in which case the // ObserveClosure message will contain the last_sequence_num field. // We can then silently ignore this message. scoped_refptr<Port> port = GetPort(port_name); if (!port) { DVLOG(1) << "ObserveProxy: " << port_name << "@" << name_ << " not found"; return OK; } DVLOG(2) << "ObserveProxy at " << port_name << "@" << name_ << ", proxy at " << event.proxy_port_name << "@" << event.proxy_node_name << " pointing to " << event.proxy_to_port_name << "@" << event.proxy_to_node_name; { base::AutoLock lock(port->lock); if (port->peer_node_name == event.proxy_node_name && port->peer_port_name == event.proxy_port_name) { if (port->state == Port::kReceiving) { port->peer_node_name = event.proxy_to_node_name; port->peer_port_name = event.proxy_to_port_name; ObserveProxyAckEventData ack; ack.last_sequence_num = port->next_sequence_num_to_send - 1; delegate_->ForwardMessage( event.proxy_node_name, NewInternalMessage(event.proxy_port_name, EventType::kObserveProxyAck, ack)); } else { // As a proxy ourselves, we don't know how to honor the ObserveProxy // event or to populate the last_sequence_num field of ObserveProxyAck. // Afterall, another port could be sending messages to our peer now // that we've sent out our own ObserveProxy event. Instead, we will // send an ObserveProxyAck indicating that the ObserveProxy event // should be re-sent (last_sequence_num set to kInvalidSequenceNum). // However, this has to be done after we are removed as a proxy. // Otherwise, we might just find ourselves back here again, which // would be akin to a busy loop. DVLOG(2) << "Delaying ObserveProxyAck to " << event.proxy_port_name << "@" << event.proxy_node_name; ObserveProxyAckEventData ack; ack.last_sequence_num = kInvalidSequenceNum; port->send_on_proxy_removal.reset( new std::pair<NodeName, ScopedMessage>( event.proxy_node_name, NewInternalMessage(event.proxy_port_name, EventType::kObserveProxyAck, ack))); } } else { // Forward this event along to our peer. Eventually, it should find the // port referring to the proxy. delegate_->ForwardMessage( port->peer_node_name, NewInternalMessage(port->peer_port_name, EventType::kObserveProxy, event)); } } return OK; } int Node::OnObserveProxyAck(const PortName& port_name, uint64_t last_sequence_num) { DVLOG(2) << "ObserveProxyAck at " << port_name << "@" << name_ << " (last_sequence_num=" << last_sequence_num << ")"; scoped_refptr<Port> port = GetPort(port_name); if (!port) return ERROR_PORT_UNKNOWN; // The port may have observed closure first, so // this is not an "Oops". { base::AutoLock lock(port->lock); if (port->state != Port::kProxying) return OOPS(ERROR_PORT_STATE_UNEXPECTED); if (last_sequence_num == kInvalidSequenceNum) { // Send again. InitiateProxyRemoval(LockedPort(port.get()), port_name); return OK; } // We can now remove this port once we have received and forwarded the last // message addressed to this port. port->remove_proxy_on_last_message = true; port->last_sequence_num_to_receive = last_sequence_num; } TryRemoveProxy(PortRef(port_name, std::move(port))); return OK; } int Node::OnObserveClosure(const PortName& port_name, uint64_t last_sequence_num) { // OK if the port doesn't exist, as it may have been closed already. scoped_refptr<Port> port = GetPort(port_name); if (!port) return OK; // This message tells the port that it should no longer expect more messages // beyond last_sequence_num. This message is forwarded along until we reach // the receiving end, and this message serves as an equivalent to // ObserveProxyAck. bool notify_delegate = false; ObserveClosureEventData forwarded_data; NodeName peer_node_name; PortName peer_port_name; bool try_remove_proxy = false; { base::AutoLock lock(port->lock); port->peer_closed = true; port->last_sequence_num_to_receive = last_sequence_num; DVLOG(2) << "ObserveClosure at " << port_name << "@" << name_ << " (state=" << port->state << ") pointing to " << port->peer_port_name << "@" << port->peer_node_name << " (last_sequence_num=" << last_sequence_num << ")"; // We always forward ObserveClosure, even beyond the receiving port which // cares about it. This ensures that any dead-end proxies beyond that port // are notified to remove themselves. if (port->state == Port::kReceiving) { notify_delegate = true; // When forwarding along the other half of the port cycle, this will only // reach dead-end proxies. Tell them we've sent our last message so they // can go away. // // TODO: Repurposing ObserveClosure for this has the desired result but // may be semantically confusing since the forwarding port is not actually // closed. Consider replacing this with a new event type. forwarded_data.last_sequence_num = port->next_sequence_num_to_send - 1; } else { // We haven't yet reached the receiving peer of the closed port, so // forward the message along as-is. forwarded_data.last_sequence_num = last_sequence_num; // See about removing the port if it is a proxy as our peer won't be able // to participate in proxy removal. port->remove_proxy_on_last_message = true; if (port->state == Port::kProxying) try_remove_proxy = true; } DVLOG(2) << "Forwarding ObserveClosure from " << port_name << "@" << name_ << " to peer " << port->peer_port_name << "@" << port->peer_node_name << " (last_sequence_num=" << forwarded_data.last_sequence_num << ")"; peer_node_name = port->peer_node_name; peer_port_name = port->peer_port_name; } if (try_remove_proxy) TryRemoveProxy(PortRef(port_name, port)); delegate_->ForwardMessage( peer_node_name, NewInternalMessage(peer_port_name, EventType::kObserveClosure, forwarded_data)); if (notify_delegate) { PortRef port_ref(port_name, std::move(port)); delegate_->PortStatusChanged(port_ref); } return OK; } int Node::OnMergePort(const PortName& port_name, const MergePortEventData& event) { scoped_refptr<Port> port = GetPort(port_name); DVLOG(1) << "MergePort at " << port_name << "@" << name_ << " (state=" << (port ? port->state : -1) << ") merging with proxy " << event.new_port_name << "@" << name_ << " pointing to " << event.new_port_descriptor.peer_port_name << "@" << event.new_port_descriptor.peer_node_name << " referred by " << event.new_port_descriptor.referring_port_name << "@" << event.new_port_descriptor.referring_node_name; bool close_target_port = false; bool close_new_port = false; // Accept the new port. This is now the receiving end of the other port cycle // to be merged with ours. int rv = AcceptPort(event.new_port_name, event.new_port_descriptor); if (rv != OK) { close_target_port = true; } else if (port) { // BeginProxying_Locked may call MaybeRemoveProxy_Locked, which in turn // needs to hold |ports_lock_|. We also acquire multiple port locks within. base::AutoLock ports_lock(ports_lock_); base::AutoLock lock(port->lock); if (port->state != Port::kReceiving) { close_new_port = true; } else { scoped_refptr<Port> new_port = GetPort_Locked(event.new_port_name); base::AutoLock new_port_lock(new_port->lock); DCHECK(new_port->state == Port::kReceiving); // Both ports are locked. Now all we have to do is swap their peer // information and set them up as proxies. PortRef port0_ref(port_name, port); PortRef port1_ref(event.new_port_name, new_port); int rv = MergePorts_Locked(port0_ref, port1_ref); if (rv == OK) return rv; close_new_port = true; close_target_port = true; } } else { close_new_port = true; } if (close_target_port) { PortRef target_port; rv = GetPort(port_name, &target_port); DCHECK(rv == OK); ClosePort(target_port); } if (close_new_port) { PortRef new_port; rv = GetPort(event.new_port_name, &new_port); DCHECK(rv == OK); ClosePort(new_port); } return ERROR_PORT_STATE_UNEXPECTED; } int Node::AddPortWithName(const PortName& port_name, scoped_refptr<Port> port) { base::AutoLock lock(ports_lock_); if (!ports_.insert(std::make_pair(port_name, std::move(port))).second) return OOPS(ERROR_PORT_EXISTS); // Suggests a bad UUID generator. DVLOG(2) << "Created port " << port_name << "@" << name_; return OK; } void Node::ErasePort(const PortName& port_name) { base::AutoLock lock(ports_lock_); ErasePort_Locked(port_name); } void Node::ErasePort_Locked(const PortName& port_name) { ports_lock_.AssertAcquired(); ports_.erase(port_name); DVLOG(2) << "Deleted port " << port_name << "@" << name_; } scoped_refptr<Port> Node::GetPort(const PortName& port_name) { base::AutoLock lock(ports_lock_); return GetPort_Locked(port_name); } scoped_refptr<Port> Node::GetPort_Locked(const PortName& port_name) { ports_lock_.AssertAcquired(); auto iter = ports_.find(port_name); if (iter == ports_.end()) return nullptr; #if defined(OS_ANDROID) && defined(ARCH_CPU_ARM64) // Workaround for https://crbug.com/665869. base::subtle::MemoryBarrier(); #endif return iter->second; } int Node::SendMessageInternal(const PortRef& port_ref, ScopedMessage* message) { ScopedMessage& m = *message; for (size_t i = 0; i < m->num_ports(); ++i) { if (m->ports()[i] == port_ref.name()) return ERROR_PORT_CANNOT_SEND_SELF; } Port* port = port_ref.port(); NodeName peer_node_name; { // We must acquire |ports_lock_| before grabbing any port locks, because // WillSendMessage_Locked may need to lock multiple ports out of order. base::AutoLock ports_lock(ports_lock_); base::AutoLock lock(port->lock); if (port->state != Port::kReceiving) return ERROR_PORT_STATE_UNEXPECTED; if (port->peer_closed) return ERROR_PORT_PEER_CLOSED; int rv = WillSendMessage_Locked(LockedPort(port), port_ref.name(), m.get()); if (rv != OK) return rv; // Beyond this point there's no sense in returning anything but OK. Even if // message forwarding or acceptance fails, there's nothing the embedder can // do to recover. Assume that failure beyond this point must be treated as a // transport failure. peer_node_name = port->peer_node_name; } if (peer_node_name != name_) { delegate_->ForwardMessage(peer_node_name, std::move(m)); return OK; } int rv = AcceptMessage(std::move(m)); if (rv != OK) { // See comment above for why we don't return an error in this case. DVLOG(2) << "AcceptMessage failed: " << rv; } return OK; } int Node::MergePorts_Locked(const PortRef& port0_ref, const PortRef& port1_ref) { Port* port0 = port0_ref.port(); Port* port1 = port1_ref.port(); ports_lock_.AssertAcquired(); port0->lock.AssertAcquired(); port1->lock.AssertAcquired(); CHECK(port0->state == Port::kReceiving); CHECK(port1->state == Port::kReceiving); // Ports cannot be merged with their own receiving peer! if (port0->peer_node_name == name_ && port0->peer_port_name == port1_ref.name()) return ERROR_PORT_STATE_UNEXPECTED; if (port1->peer_node_name == name_ && port1->peer_port_name == port0_ref.name()) return ERROR_PORT_STATE_UNEXPECTED; // Only merge if both ports have never sent a message. if (port0->next_sequence_num_to_send == kInitialSequenceNum && port1->next_sequence_num_to_send == kInitialSequenceNum) { // Swap the ports' peer information and switch them both into buffering // (eventually proxying) mode. std::swap(port0->peer_node_name, port1->peer_node_name); std::swap(port0->peer_port_name, port1->peer_port_name); port0->state = Port::kBuffering; if (port0->peer_closed) port0->remove_proxy_on_last_message = true; port1->state = Port::kBuffering; if (port1->peer_closed) port1->remove_proxy_on_last_message = true; int rv1 = BeginProxying_Locked(LockedPort(port0), port0_ref.name()); int rv2 = BeginProxying_Locked(LockedPort(port1), port1_ref.name()); if (rv1 == OK && rv2 == OK) { // If either merged port had a closed peer, its new peer needs to be // informed of this. if (port1->peer_closed) { ObserveClosureEventData data; data.last_sequence_num = port0->last_sequence_num_to_receive; delegate_->ForwardMessage( port0->peer_node_name, NewInternalMessage(port0->peer_port_name, EventType::kObserveClosure, data)); } if (port0->peer_closed) { ObserveClosureEventData data; data.last_sequence_num = port1->last_sequence_num_to_receive; delegate_->ForwardMessage( port1->peer_node_name, NewInternalMessage(port1->peer_port_name, EventType::kObserveClosure, data)); } return OK; } // If either proxy failed to initialize (e.g. had undeliverable messages // or ended up in a bad state somehow), we keep the system in a consistent // state by undoing the peer swap. std::swap(port0->peer_node_name, port1->peer_node_name); std::swap(port0->peer_port_name, port1->peer_port_name); port0->remove_proxy_on_last_message = false; port1->remove_proxy_on_last_message = false; port0->state = Port::kReceiving; port1->state = Port::kReceiving; } return ERROR_PORT_STATE_UNEXPECTED; } void Node::WillSendPort(const LockedPort& port, const NodeName& to_node_name, PortName* port_name, PortDescriptor* port_descriptor) { port->lock.AssertAcquired(); PortName local_port_name = *port_name; PortName new_port_name; delegate_->GenerateRandomPortName(&new_port_name); // Make sure we don't send messages to the new peer until after we know it // exists. In the meantime, just buffer messages locally. DCHECK(port->state == Port::kReceiving); port->state = Port::kBuffering; // If we already know our peer is closed, we already know this proxy can // be removed once it receives and forwards its last expected message. if (port->peer_closed) port->remove_proxy_on_last_message = true; *port_name = new_port_name; port_descriptor->peer_node_name = port->peer_node_name; port_descriptor->peer_port_name = port->peer_port_name; port_descriptor->referring_node_name = name_; port_descriptor->referring_port_name = local_port_name; port_descriptor->next_sequence_num_to_send = port->next_sequence_num_to_send; port_descriptor->next_sequence_num_to_receive = port->message_queue.next_sequence_num(); port_descriptor->last_sequence_num_to_receive = port->last_sequence_num_to_receive; port_descriptor->peer_closed = port->peer_closed; memset(port_descriptor->padding, 0, sizeof(port_descriptor->padding)); // Configure the local port to point to the new port. port->peer_node_name = to_node_name; port->peer_port_name = new_port_name; } int Node::AcceptPort(const PortName& port_name, const PortDescriptor& port_descriptor) { scoped_refptr<Port> port = make_scoped_refptr( new Port(port_descriptor.next_sequence_num_to_send, port_descriptor.next_sequence_num_to_receive)); port->state = Port::kReceiving; port->peer_node_name = port_descriptor.peer_node_name; port->peer_port_name = port_descriptor.peer_port_name; port->last_sequence_num_to_receive = port_descriptor.last_sequence_num_to_receive; port->peer_closed = port_descriptor.peer_closed; DVLOG(2) << "Accepting port " << port_name << " [peer_closed=" << port->peer_closed << "; last_sequence_num_to_receive=" << port->last_sequence_num_to_receive << "]"; // A newly accepted port is not signalable until the message referencing the // new port finds its way to the consumer (see GetMessage). port->message_queue.set_signalable(false); int rv = AddPortWithName(port_name, std::move(port)); if (rv != OK) return rv; // Allow referring port to forward messages. delegate_->ForwardMessage( port_descriptor.referring_node_name, NewInternalMessage(port_descriptor.referring_port_name, EventType::kPortAccepted)); return OK; } int Node::WillSendMessage_Locked(const LockedPort& port, const PortName& port_name, Message* message) { ports_lock_.AssertAcquired(); port->lock.AssertAcquired(); DCHECK(message); // Messages may already have a sequence number if they're being forwarded // by a proxy. Otherwise, use the next outgoing sequence number. uint64_t* sequence_num = &GetMutableEventData<UserEventData>(message)->sequence_num; if (*sequence_num == 0) *sequence_num = port->next_sequence_num_to_send++; #if DCHECK_IS_ON() std::ostringstream ports_buf; for (size_t i = 0; i < message->num_ports(); ++i) { if (i > 0) ports_buf << ","; ports_buf << message->ports()[i]; } #endif if (message->num_ports() > 0) { // Note: Another thread could be trying to send the same ports, so we need // to ensure that they are ours to send before we mutate their state. std::vector<scoped_refptr<Port>> ports; ports.resize(message->num_ports()); { for (size_t i = 0; i < message->num_ports(); ++i) { ports[i] = GetPort_Locked(message->ports()[i]); DCHECK(ports[i]); ports[i]->lock.Acquire(); int error = OK; if (ports[i]->state != Port::kReceiving) error = ERROR_PORT_STATE_UNEXPECTED; else if (message->ports()[i] == port->peer_port_name) error = ERROR_PORT_CANNOT_SEND_PEER; if (error != OK) { // Oops, we cannot send this port. for (size_t j = 0; j <= i; ++j) ports[i]->lock.Release(); // Backpedal on the sequence number. port->next_sequence_num_to_send--; return error; } } } PortDescriptor* port_descriptors = GetMutablePortDescriptors(GetMutableEventData<UserEventData>(message)); for (size_t i = 0; i < message->num_ports(); ++i) { WillSendPort(LockedPort(ports[i].get()), port->peer_node_name, message->mutable_ports() + i, port_descriptors + i); } for (size_t i = 0; i < message->num_ports(); ++i) ports[i]->lock.Release(); } #if DCHECK_IS_ON() DVLOG(4) << "Sending message " << GetEventData<UserEventData>(*message)->sequence_num << " [ports=" << ports_buf.str() << "]" << " from " << port_name << "@" << name_ << " to " << port->peer_port_name << "@" << port->peer_node_name; #endif GetMutableEventHeader(message)->port_name = port->peer_port_name; return OK; } int Node::BeginProxying_Locked(const LockedPort& port, const PortName& port_name) { ports_lock_.AssertAcquired(); port->lock.AssertAcquired(); if (port->state != Port::kBuffering) return OOPS(ERROR_PORT_STATE_UNEXPECTED); port->state = Port::kProxying; int rv = ForwardMessages_Locked(LockedPort(port), port_name); if (rv != OK) return rv; // We may have observed closure while buffering. In that case, we can advance // to removing the proxy without sending out an ObserveProxy message. We // already know the last expected message, etc. if (port->remove_proxy_on_last_message) { MaybeRemoveProxy_Locked(LockedPort(port), port_name); // Make sure we propagate closure to our current peer. ObserveClosureEventData data; data.last_sequence_num = port->last_sequence_num_to_receive; delegate_->ForwardMessage( port->peer_node_name, NewInternalMessage(port->peer_port_name, EventType::kObserveClosure, data)); } else { InitiateProxyRemoval(LockedPort(port), port_name); } return OK; } int Node::BeginProxying(PortRef port_ref) { Port* port = port_ref.port(); { base::AutoLock ports_lock(ports_lock_); base::AutoLock lock(port->lock); if (port->state != Port::kBuffering) return OOPS(ERROR_PORT_STATE_UNEXPECTED); port->state = Port::kProxying; int rv = ForwardMessages_Locked(LockedPort(port), port_ref.name()); if (rv != OK) return rv; } bool should_remove; NodeName peer_node_name; ScopedMessage closure_message; { base::AutoLock lock(port->lock); if (port->state != Port::kProxying) return OOPS(ERROR_PORT_STATE_UNEXPECTED); should_remove = port->remove_proxy_on_last_message; if (should_remove) { // Make sure we propagate closure to our current peer. ObserveClosureEventData data; data.last_sequence_num = port->last_sequence_num_to_receive; peer_node_name = port->peer_node_name; closure_message = NewInternalMessage(port->peer_port_name, EventType::kObserveClosure, data); } else { InitiateProxyRemoval(LockedPort(port), port_ref.name()); } } if (should_remove) { TryRemoveProxy(port_ref); delegate_->ForwardMessage(peer_node_name, std::move(closure_message)); } return OK; } int Node::ForwardMessages_Locked(const LockedPort& port, const PortName &port_name) { ports_lock_.AssertAcquired(); port->lock.AssertAcquired(); for (;;) { ScopedMessage message; port->message_queue.GetNextMessage(&message, nullptr); if (!message) break; int rv = WillSendMessage_Locked(LockedPort(port), port_name, message.get()); if (rv != OK) return rv; delegate_->ForwardMessage(port->peer_node_name, std::move(message)); } return OK; } void Node::InitiateProxyRemoval(const LockedPort& port, const PortName& port_name) { port->lock.AssertAcquired(); // To remove this node, we start by notifying the connected graph that we are // a proxy. This allows whatever port is referencing this node to skip it. // Eventually, this node will receive ObserveProxyAck (or ObserveClosure if // the peer was closed in the meantime). ObserveProxyEventData data; data.proxy_node_name = name_; data.proxy_port_name = port_name; data.proxy_to_node_name = port->peer_node_name; data.proxy_to_port_name = port->peer_port_name; delegate_->ForwardMessage( port->peer_node_name, NewInternalMessage(port->peer_port_name, EventType::kObserveProxy, data)); } void Node::MaybeRemoveProxy_Locked(const LockedPort& port, const PortName& port_name) { // |ports_lock_| must be held so we can potentilaly ErasePort_Locked(). ports_lock_.AssertAcquired(); port->lock.AssertAcquired(); DCHECK(port->state == Port::kProxying); // Make sure we have seen ObserveProxyAck before removing the port. if (!port->remove_proxy_on_last_message) return; if (!CanAcceptMoreMessages(port.get())) { // This proxy port is done. We can now remove it! ErasePort_Locked(port_name); if (port->send_on_proxy_removal) { NodeName to_node = port->send_on_proxy_removal->first; ScopedMessage& message = port->send_on_proxy_removal->second; delegate_->ForwardMessage(to_node, std::move(message)); port->send_on_proxy_removal.reset(); } } else { DVLOG(2) << "Cannot remove port " << port_name << "@" << name_ << " now; waiting for more messages"; } } void Node::TryRemoveProxy(PortRef port_ref) { Port* port = port_ref.port(); bool should_erase = false; ScopedMessage msg; NodeName to_node; { base::AutoLock lock(port->lock); // Port already removed. Nothing to do. if (port->state == Port::kClosed) return; DCHECK(port->state == Port::kProxying); // Make sure we have seen ObserveProxyAck before removing the port. if (!port->remove_proxy_on_last_message) return; if (!CanAcceptMoreMessages(port)) { // This proxy port is done. We can now remove it! should_erase = true; if (port->send_on_proxy_removal) { to_node = port->send_on_proxy_removal->first; msg = std::move(port->send_on_proxy_removal->second); port->send_on_proxy_removal.reset(); } } else { DVLOG(2) << "Cannot remove port " << port_ref.name() << "@" << name_ << " now; waiting for more messages"; } } if (should_erase) ErasePort(port_ref.name()); if (msg) delegate_->ForwardMessage(to_node, std::move(msg)); } void Node::DestroyAllPortsWithPeer(const NodeName& node_name, const PortName& port_name) { // Wipes out all ports whose peer node matches |node_name| and whose peer port // matches |port_name|. If |port_name| is |kInvalidPortName|, only the peer // node is matched. std::vector<PortRef> ports_to_notify; std::vector<PortName> dead_proxies_to_broadcast; std::deque<PortName> referenced_port_names; { base::AutoLock ports_lock(ports_lock_); for (auto iter = ports_.begin(); iter != ports_.end(); ++iter) { Port* port = iter->second.get(); { base::AutoLock port_lock(port->lock); if (port->peer_node_name == node_name && (port_name == kInvalidPortName || port->peer_port_name == port_name)) { if (!port->peer_closed) { // Treat this as immediate peer closure. It's an exceptional // condition akin to a broken pipe, so we don't care about losing // messages. port->peer_closed = true; port->last_sequence_num_to_receive = port->message_queue.next_sequence_num() - 1; if (port->state == Port::kReceiving) ports_to_notify.push_back(PortRef(iter->first, port)); } // We don't expect to forward any further messages, and we don't // expect to receive a Port{Accepted,Rejected} event. Because we're // a proxy with no active peer, we cannot use the normal proxy removal // procedure of forward-propagating an ObserveProxy. Instead we // broadcast our own death so it can be back-propagated. This is // inefficient but rare. if (port->state != Port::kReceiving) { dead_proxies_to_broadcast.push_back(iter->first); iter->second->message_queue.GetReferencedPorts( &referenced_port_names); } } } } for (const auto& proxy_name : dead_proxies_to_broadcast) { ports_.erase(proxy_name); DVLOG(2) << "Forcibly deleted port " << proxy_name << "@" << name_; } } // Wake up any receiving ports who have just observed simulated peer closure. for (const auto& port : ports_to_notify) delegate_->PortStatusChanged(port); for (const auto& proxy_name : dead_proxies_to_broadcast) { // Broadcast an event signifying that this proxy is no longer functioning. ObserveProxyEventData event; event.proxy_node_name = name_; event.proxy_port_name = proxy_name; event.proxy_to_node_name = kInvalidNodeName; event.proxy_to_port_name = kInvalidPortName; delegate_->BroadcastMessage(NewInternalMessage( kInvalidPortName, EventType::kObserveProxy, event)); // Also process death locally since the port that points this closed one // could be on the current node. // Note: Although this is recursive, only a single port is involved which // limits the expected branching to 1. DestroyAllPortsWithPeer(name_, proxy_name); } // Close any ports referenced by the closed proxies. for (const auto& name : referenced_port_names) { PortRef ref; if (GetPort(name, &ref) == OK) ClosePort(ref); } } ScopedMessage Node::NewInternalMessage_Helper(const PortName& port_name, const EventType& type, const void* data, size_t num_data_bytes) { ScopedMessage message; delegate_->AllocMessage(sizeof(EventHeader) + num_data_bytes, &message); EventHeader* header = GetMutableEventHeader(message.get()); header->port_name = port_name; header->type = type; header->padding = 0; if (num_data_bytes) memcpy(header + 1, data, num_data_bytes); return message; } } // namespace ports } // namespace edk } // namespace mojo