diff options
Diffstat (limited to 'pkg/sentry/socket/unix')
-rw-r--r-- | pkg/sentry/socket/unix/BUILD | 21 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/BUILD | 13 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/connectioned.go | 22 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/connectionless.go | 6 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/queue.go | 13 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/unix.go | 61 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/unix.go | 71 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/unix_vfs2.go | 60 |
8 files changed, 181 insertions, 86 deletions
diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index cb953e4dc..cce0acc33 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -7,10 +7,21 @@ go_template_instance( name = "socket_refs", out = "socket_refs.go", package = "unix", - prefix = "socketOpsCommon", - template = "//pkg/refs_vfs2:refs_template", + prefix = "socketOperations", + template = "//pkg/refsvfs2:refs_template", types = { - "T": "socketOpsCommon", + "T": "SocketOperations", + }, +) + +go_template_instance( + name = "socket_vfs2_refs", + out = "socket_vfs2_refs.go", + package = "unix", + prefix = "socketVFS2", + template = "//pkg/refsvfs2:refs_template", + types = { + "T": "SocketVFS2", }, ) @@ -20,6 +31,7 @@ go_library( "device.go", "io.go", "socket_refs.go", + "socket_vfs2_refs.go", "unix.go", "unix_vfs2.go", ], @@ -29,7 +41,9 @@ go_library( "//pkg/context", "//pkg/fspath", "//pkg/log", + "//pkg/marshal", "//pkg/refs", + "//pkg/refsvfs2", "//pkg/safemem", "//pkg/sentry/arch", "//pkg/sentry/device", @@ -49,6 +63,5 @@ go_library( "//pkg/tcpip", "//pkg/usermem", "//pkg/waiter", - "//tools/go_marshal/marshal", ], ) diff --git a/pkg/sentry/socket/unix/transport/BUILD b/pkg/sentry/socket/unix/transport/BUILD index c708b6030..3ebbd28b0 100644 --- a/pkg/sentry/socket/unix/transport/BUILD +++ b/pkg/sentry/socket/unix/transport/BUILD @@ -15,6 +15,17 @@ go_template_instance( }, ) +go_template_instance( + name = "queue_refs", + out = "queue_refs.go", + package = "transport", + prefix = "queue", + template = "//pkg/refsvfs2:refs_template", + types = { + "T": "queue", + }, +) + go_library( name = "transport", srcs = [ @@ -22,6 +33,7 @@ go_library( "connectioned_state.go", "connectionless.go", "queue.go", + "queue_refs.go", "transport_message_list.go", "unix.go", ], @@ -32,6 +44,7 @@ go_library( "//pkg/ilist", "//pkg/log", "//pkg/refs", + "//pkg/refsvfs2", "//pkg/sync", "//pkg/syserr", "//pkg/tcpip", diff --git a/pkg/sentry/socket/unix/transport/connectioned.go b/pkg/sentry/socket/unix/transport/connectioned.go index c67b602f0..aa4f3c04d 100644 --- a/pkg/sentry/socket/unix/transport/connectioned.go +++ b/pkg/sentry/socket/unix/transport/connectioned.go @@ -142,9 +142,9 @@ func NewPair(ctx context.Context, stype linux.SockType, uid UniqueIDProvider) (E } q1 := &queue{ReaderQueue: a.Queue, WriterQueue: b.Queue, limit: initialLimit} - q1.EnableLeakCheck("transport.queue") + q1.EnableLeakCheck() q2 := &queue{ReaderQueue: b.Queue, WriterQueue: a.Queue, limit: initialLimit} - q2.EnableLeakCheck("transport.queue") + q2.EnableLeakCheck() if stype == linux.SOCK_STREAM { a.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{q1}} @@ -300,14 +300,14 @@ func (e *connectionedEndpoint) BidirectionalConnect(ctx context.Context, ce Conn } readQueue := &queue{ReaderQueue: ce.WaiterQueue(), WriterQueue: ne.Queue, limit: initialLimit} - readQueue.EnableLeakCheck("transport.queue") + readQueue.EnableLeakCheck() ne.connected = &connectedEndpoint{ endpoint: ce, writeQueue: readQueue, } writeQueue := &queue{ReaderQueue: ne.Queue, WriterQueue: ce.WaiterQueue(), limit: initialLimit} - writeQueue.EnableLeakCheck("transport.queue") + writeQueue.EnableLeakCheck() if e.stype == linux.SOCK_STREAM { ne.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{readQueue: writeQueue}} } else { @@ -391,7 +391,7 @@ func (e *connectionedEndpoint) Listen(backlog int) *syserr.Error { } // Accept accepts a new connection. -func (e *connectionedEndpoint) Accept() (Endpoint, *syserr.Error) { +func (e *connectionedEndpoint) Accept(peerAddr *tcpip.FullAddress) (Endpoint, *syserr.Error) { e.Lock() defer e.Unlock() @@ -401,6 +401,18 @@ func (e *connectionedEndpoint) Accept() (Endpoint, *syserr.Error) { select { case ne := <-e.acceptedChan: + if peerAddr != nil { + ne.Lock() + c := ne.connected + ne.Unlock() + if c != nil { + addr, err := c.GetLocalAddress() + if err != nil { + return nil, syserr.TranslateNetstackError(err) + } + *peerAddr = addr + } + } return ne, nil default: diff --git a/pkg/sentry/socket/unix/transport/connectionless.go b/pkg/sentry/socket/unix/transport/connectionless.go index 70ee8f9b8..f8aacca13 100644 --- a/pkg/sentry/socket/unix/transport/connectionless.go +++ b/pkg/sentry/socket/unix/transport/connectionless.go @@ -42,7 +42,7 @@ var ( func NewConnectionless(ctx context.Context) Endpoint { ep := &connectionlessEndpoint{baseEndpoint{Queue: &waiter.Queue{}}} q := queue{ReaderQueue: ep.Queue, WriterQueue: &waiter.Queue{}, limit: initialLimit} - q.EnableLeakCheck("transport.queue") + q.EnableLeakCheck() ep.receiver = &queueReceiver{readQueue: &q} return ep } @@ -144,12 +144,12 @@ func (e *connectionlessEndpoint) Connect(ctx context.Context, server BoundEndpoi } // Listen starts listening on the connection. -func (e *connectionlessEndpoint) Listen(int) *syserr.Error { +func (*connectionlessEndpoint) Listen(int) *syserr.Error { return syserr.ErrNotSupported } // Accept accepts a new connection. -func (e *connectionlessEndpoint) Accept() (Endpoint, *syserr.Error) { +func (*connectionlessEndpoint) Accept(*tcpip.FullAddress) (Endpoint, *syserr.Error) { return nil, syserr.ErrNotSupported } diff --git a/pkg/sentry/socket/unix/transport/queue.go b/pkg/sentry/socket/unix/transport/queue.go index ef6043e19..342def28f 100644 --- a/pkg/sentry/socket/unix/transport/queue.go +++ b/pkg/sentry/socket/unix/transport/queue.go @@ -16,7 +16,6 @@ package transport import ( "gvisor.dev/gvisor/pkg/context" - "gvisor.dev/gvisor/pkg/refs" "gvisor.dev/gvisor/pkg/sync" "gvisor.dev/gvisor/pkg/syserr" "gvisor.dev/gvisor/pkg/tcpip" @@ -28,7 +27,7 @@ import ( // // +stateify savable type queue struct { - refs.AtomicRefCount + queueRefs ReaderQueue *waiter.Queue WriterQueue *waiter.Queue @@ -68,11 +67,13 @@ func (q *queue) Reset(ctx context.Context) { q.mu.Unlock() } -// DecRef implements RefCounter.DecRef with destructor q.Reset. +// DecRef implements RefCounter.DecRef. func (q *queue) DecRef(ctx context.Context) { - q.DecRefWithDestructor(ctx, q.Reset) - // We don't need to notify after resetting because no one cares about - // this queue after all references have been dropped. + q.queueRefs.DecRef(func() { + // We don't need to notify after resetting because no one cares about + // this queue after all references have been dropped. + q.Reset(ctx) + }) } // IsReadable determines if q is currently readable. diff --git a/pkg/sentry/socket/unix/transport/unix.go b/pkg/sentry/socket/unix/transport/unix.go index 475d7177e..b648273a4 100644 --- a/pkg/sentry/socket/unix/transport/unix.go +++ b/pkg/sentry/socket/unix/transport/unix.go @@ -32,6 +32,8 @@ import ( const initialLimit = 16 * 1024 // A RightsControlMessage is a control message containing FDs. +// +// +stateify savable type RightsControlMessage interface { // Clone returns a copy of the RightsControlMessage. Clone() RightsControlMessage @@ -151,7 +153,10 @@ type Endpoint interface { // block if no new connections are available. // // The returned Queue is the wait queue for the newly created endpoint. - Accept() (Endpoint, *syserr.Error) + // + // peerAddr if not nil will be populated with the address of the connected + // peer on a successful accept. + Accept(peerAddr *tcpip.FullAddress) (Endpoint, *syserr.Error) // Bind binds the endpoint to a specific local address and port. // Specifying a NIC is optional. @@ -172,9 +177,8 @@ type Endpoint interface { // connected. GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) - // SetSockOpt sets a socket option. opt should be one of the tcpip.*Option - // types. - SetSockOpt(opt interface{}) *tcpip.Error + // SetSockOpt sets a socket option. + SetSockOpt(opt tcpip.SettableSocketOption) *tcpip.Error // SetSockOptBool sets a socket option for simple cases when a value has // the int type. @@ -184,9 +188,8 @@ type Endpoint interface { // the int type. SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error - // GetSockOpt gets a socket option. opt should be a pointer to one of the - // tcpip.*Option types. - GetSockOpt(opt interface{}) *tcpip.Error + // GetSockOpt gets a socket option. + GetSockOpt(opt tcpip.GettableSocketOption) *tcpip.Error // GetSockOptBool gets a socket option for simple cases when a return // value has the int type. @@ -199,6 +202,9 @@ type Endpoint interface { // State returns the current state of the socket, as represented by Linux in // procfs. State() uint32 + + // LastError implements tcpip.Endpoint.LastError. + LastError() *tcpip.Error } // A Credentialer is a socket or endpoint that supports the SO_PASSCRED socket @@ -332,7 +338,7 @@ type Receiver interface { RecvMaxQueueSize() int64 // Release releases any resources owned by the Receiver. It should be - // called before droping all references to a Receiver. + // called before dropping all references to a Receiver. Release(ctx context.Context) } @@ -483,7 +489,7 @@ func (q *streamQueueReceiver) Recv(ctx context.Context, data [][]byte, wantCreds c := q.control.Clone() // Don't consume data since we are peeking. - copied, data, _ = vecCopy(data, q.buffer) + copied, _, _ = vecCopy(data, q.buffer) return copied, copied, c, false, q.addr, notify, nil } @@ -568,6 +574,12 @@ func (q *streamQueueReceiver) Recv(ctx context.Context, data [][]byte, wantCreds return copied, copied, c, cmTruncated, q.addr, notify, nil } +// Release implements Receiver.Release. +func (q *streamQueueReceiver) Release(ctx context.Context) { + q.queueReceiver.Release(ctx) + q.control.Release(ctx) +} + // A ConnectedEndpoint is an Endpoint that can be used to send Messages. type ConnectedEndpoint interface { // Passcred implements Endpoint.Passcred. @@ -615,7 +627,7 @@ type ConnectedEndpoint interface { SendMaxQueueSize() int64 // Release releases any resources owned by the ConnectedEndpoint. It should - // be called before droping all references to a ConnectedEndpoint. + // be called before dropping all references to a ConnectedEndpoint. Release(ctx context.Context) // CloseUnread sets the fact that this end is closed with unread data to @@ -742,6 +754,9 @@ type baseEndpoint struct { // path is not empty if the endpoint has been bound, // or may be used if the endpoint is connected. path string + + // linger is used for SO_LINGER socket option. + linger tcpip.LingerOption } // EventRegister implements waiter.Waitable.EventRegister. @@ -837,8 +852,14 @@ func (e *baseEndpoint) SendMsg(ctx context.Context, data [][]byte, c ControlMess return n, err } -// SetSockOpt sets a socket option. Currently not supported. -func (e *baseEndpoint) SetSockOpt(opt interface{}) *tcpip.Error { +// SetSockOpt sets a socket option. +func (e *baseEndpoint) SetSockOpt(opt tcpip.SettableSocketOption) *tcpip.Error { + switch v := opt.(type) { + case *tcpip.LingerOption: + e.Lock() + e.linger = *v + e.Unlock() + } return nil } @@ -866,7 +887,7 @@ func (e *baseEndpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error { func (e *baseEndpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) { switch opt { - case tcpip.KeepaliveEnabledOption: + case tcpip.KeepaliveEnabledOption, tcpip.AcceptConnOption: return false, nil case tcpip.PasscredOption: @@ -940,9 +961,12 @@ func (e *baseEndpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) { } // GetSockOpt implements tcpip.Endpoint.GetSockOpt. -func (e *baseEndpoint) GetSockOpt(opt interface{}) *tcpip.Error { - switch opt.(type) { - case tcpip.ErrorOption: +func (e *baseEndpoint) GetSockOpt(opt tcpip.GettableSocketOption) *tcpip.Error { + switch o := opt.(type) { + case *tcpip.LingerOption: + e.Lock() + *o = e.linger + e.Unlock() return nil default: @@ -951,6 +975,11 @@ func (e *baseEndpoint) GetSockOpt(opt interface{}) *tcpip.Error { } } +// LastError implements Endpoint.LastError. +func (*baseEndpoint) LastError() *tcpip.Error { + return nil +} + // Shutdown closes the read and/or write end of the endpoint connection to its // peer. func (e *baseEndpoint) Shutdown(flags tcpip.ShutdownFlags) *syserr.Error { diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index b7e8e4325..adad485a9 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -24,6 +24,7 @@ import ( "gvisor.dev/gvisor/pkg/abi/linux" "gvisor.dev/gvisor/pkg/context" "gvisor.dev/gvisor/pkg/fspath" + "gvisor.dev/gvisor/pkg/marshal" "gvisor.dev/gvisor/pkg/sentry/arch" "gvisor.dev/gvisor/pkg/sentry/fs" "gvisor.dev/gvisor/pkg/sentry/fs/fsutil" @@ -39,7 +40,6 @@ import ( "gvisor.dev/gvisor/pkg/tcpip" "gvisor.dev/gvisor/pkg/usermem" "gvisor.dev/gvisor/pkg/waiter" - "gvisor.dev/gvisor/tools/go_marshal/marshal" ) // SocketOperations is a Unix socket. It is similar to a netstack socket, @@ -55,6 +55,7 @@ type SocketOperations struct { fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + socketOperationsRefs socketOpsCommon } @@ -80,15 +81,30 @@ func NewWithDirent(ctx context.Context, d *fs.Dirent, ep transport.Endpoint, sty }, } s.EnableLeakCheck() - return fs.NewFile(ctx, d, flags, &s) } +// DecRef implements RefCounter.DecRef. +func (s *SocketOperations) DecRef(ctx context.Context) { + s.socketOperationsRefs.DecRef(func() { + s.ep.Close(ctx) + if s.abstractNamespace != nil { + s.abstractNamespace.Remove(s.abstractName, s) + } + }) +} + +// Release implemements fs.FileOperations.Release. +func (s *SocketOperations) Release(ctx context.Context) { + // Release only decrements a reference on s because s may be referenced in + // the abstract socket namespace. + s.DecRef(ctx) +} + // socketOpsCommon contains the socket operations common to VFS1 and VFS2. // // +stateify savable type socketOpsCommon struct { - socketOpsCommonRefs socket.SendReceiveTimeout ep transport.Endpoint @@ -101,23 +117,6 @@ type socketOpsCommon struct { abstractNamespace *kernel.AbstractSocketNamespace } -// DecRef implements RefCounter.DecRef. -func (s *socketOpsCommon) DecRef(ctx context.Context) { - s.socketOpsCommonRefs.DecRef(func() { - s.ep.Close(ctx) - if s.abstractNamespace != nil { - s.abstractNamespace.Remove(s.abstractName, s) - } - }) -} - -// Release implemements fs.FileOperations.Release. -func (s *socketOpsCommon) Release(ctx context.Context) { - // Release only decrements a reference on s because s may be referenced in - // the abstract socket namespace. - s.DecRef(ctx) -} - func (s *socketOpsCommon) isPacket() bool { switch s.stype { case linux.SOCK_DGRAM, linux.SOCK_SEQPACKET: @@ -194,7 +193,7 @@ func (s *SocketOperations) Ioctl(ctx context.Context, _ *fs.File, io usermem.IO, // GetSockOpt implements the linux syscall getsockopt(2) for sockets backed by // a transport.Endpoint. func (s *SocketOperations) GetSockOpt(t *kernel.Task, level, name int, outPtr usermem.Addr, outLen int) (marshal.Marshallable, *syserr.Error) { - return netstack.GetSockOpt(t, s, s.ep, linux.AF_UNIX, s.ep.Type(), level, name, outLen) + return netstack.GetSockOpt(t, s, s.ep, linux.AF_UNIX, s.ep.Type(), level, name, outPtr, outLen) } // Listen implements the linux syscall listen(2) for sockets backed by @@ -205,7 +204,7 @@ func (s *socketOpsCommon) Listen(t *kernel.Task, backlog int) *syserr.Error { // blockingAccept implements a blocking version of accept(2), that is, if no // connections are ready to be accept, it will block until one becomes ready. -func (s *SocketOperations) blockingAccept(t *kernel.Task) (transport.Endpoint, *syserr.Error) { +func (s *SocketOperations) blockingAccept(t *kernel.Task, peerAddr *tcpip.FullAddress) (transport.Endpoint, *syserr.Error) { // Register for notifications. e, ch := waiter.NewChannelEntry(nil) s.EventRegister(&e, waiter.EventIn) @@ -214,7 +213,7 @@ func (s *SocketOperations) blockingAccept(t *kernel.Task) (transport.Endpoint, * // Try to accept the connection; if it fails, then wait until we get a // notification. for { - if ep, err := s.ep.Accept(); err != syserr.ErrWouldBlock { + if ep, err := s.ep.Accept(peerAddr); err != syserr.ErrWouldBlock { return ep, err } @@ -227,15 +226,18 @@ func (s *SocketOperations) blockingAccept(t *kernel.Task) (transport.Endpoint, * // Accept implements the linux syscall accept(2) for sockets backed by // a transport.Endpoint. func (s *SocketOperations) Accept(t *kernel.Task, peerRequested bool, flags int, blocking bool) (int32, linux.SockAddr, uint32, *syserr.Error) { - // Issue the accept request to get the new endpoint. - ep, err := s.ep.Accept() + var peerAddr *tcpip.FullAddress + if peerRequested { + peerAddr = &tcpip.FullAddress{} + } + ep, err := s.ep.Accept(peerAddr) if err != nil { if err != syserr.ErrWouldBlock || !blocking { return 0, nil, 0, err } var err *syserr.Error - ep, err = s.blockingAccept(t) + ep, err = s.blockingAccept(t, peerAddr) if err != nil { return 0, nil, 0, err } @@ -252,13 +254,8 @@ func (s *SocketOperations) Accept(t *kernel.Task, peerRequested bool, flags int, var addr linux.SockAddr var addrLen uint32 - if peerRequested { - // Get address of the peer. - var err *syserr.Error - addr, addrLen, err = ns.FileOperations.(*SocketOperations).GetPeerName(t) - if err != nil { - return 0, nil, 0, err - } + if peerAddr != nil { + addr, addrLen = netstack.ConvertAddress(linux.AF_UNIX, *peerAddr) } fd, e := t.NewFDFrom(0, ns, kernel.FDFlags{ @@ -575,13 +572,17 @@ func (s *SocketOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOS if dst.NumBytes() == 0 { return 0, nil } - return dst.CopyOutFrom(ctx, &EndpointReader{ + r := &EndpointReader{ Ctx: ctx, Endpoint: s.ep, NumRights: 0, Peek: false, From: nil, - }) + } + n, err := dst.CopyOutFrom(ctx, r) + // Drop control messages. + r.Control.Release(ctx) + return n, err } // RecvMsg implements the linux syscall recvmsg(2) for sockets backed by diff --git a/pkg/sentry/socket/unix/unix_vfs2.go b/pkg/sentry/socket/unix/unix_vfs2.go index d066ef8ab..7a78444dc 100644 --- a/pkg/sentry/socket/unix/unix_vfs2.go +++ b/pkg/sentry/socket/unix/unix_vfs2.go @@ -18,6 +18,7 @@ import ( "gvisor.dev/gvisor/pkg/abi/linux" "gvisor.dev/gvisor/pkg/context" "gvisor.dev/gvisor/pkg/fspath" + "gvisor.dev/gvisor/pkg/marshal" "gvisor.dev/gvisor/pkg/sentry/arch" fslock "gvisor.dev/gvisor/pkg/sentry/fs/lock" "gvisor.dev/gvisor/pkg/sentry/fsimpl/sockfs" @@ -32,17 +33,19 @@ import ( "gvisor.dev/gvisor/pkg/tcpip" "gvisor.dev/gvisor/pkg/usermem" "gvisor.dev/gvisor/pkg/waiter" - "gvisor.dev/gvisor/tools/go_marshal/marshal" ) // SocketVFS2 implements socket.SocketVFS2 (and by extension, // vfs.FileDescriptionImpl) for Unix sockets. +// +// +stateify savable type SocketVFS2 struct { vfsfd vfs.FileDescription vfs.FileDescriptionDefaultImpl vfs.DentryMetadataFileDescriptionImpl vfs.LockFD + socketVFS2Refs socketOpsCommon } @@ -52,7 +55,8 @@ var _ = socket.SocketVFS2(&SocketVFS2{}) // returns a corresponding file description. func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType) (*vfs.FileDescription, *syserr.Error) { mnt := t.Kernel().SocketMount() - d := sockfs.NewDentry(t.Credentials(), mnt) + d := sockfs.NewDentry(t, mnt) + defer d.DecRef(t) fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d, &vfs.FileLocks{}) if err != nil { @@ -76,6 +80,7 @@ func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint3 stype: stype, }, } + sock.EnableLeakCheck() sock.LockFD.Init(locks) vfsfd := &sock.vfsfd if err := vfsfd.Init(sock, flags, mnt, d, &vfs.FileDescriptionOptions{ @@ -88,15 +93,34 @@ func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint3 return vfsfd, nil } +// DecRef implements RefCounter.DecRef. +func (s *SocketVFS2) DecRef(ctx context.Context) { + s.socketVFS2Refs.DecRef(func() { + t := kernel.TaskFromContext(ctx) + t.Kernel().DeleteSocketVFS2(&s.vfsfd) + s.ep.Close(ctx) + if s.abstractNamespace != nil { + s.abstractNamespace.Remove(s.abstractName, s) + } + }) +} + +// Release implements vfs.FileDescriptionImpl.Release. +func (s *SocketVFS2) Release(ctx context.Context) { + // Release only decrements a reference on s because s may be referenced in + // the abstract socket namespace. + s.DecRef(ctx) +} + // GetSockOpt implements the linux syscall getsockopt(2) for sockets backed by // a transport.Endpoint. func (s *SocketVFS2) GetSockOpt(t *kernel.Task, level, name int, outPtr usermem.Addr, outLen int) (marshal.Marshallable, *syserr.Error) { - return netstack.GetSockOpt(t, s, s.ep, linux.AF_UNIX, s.ep.Type(), level, name, outLen) + return netstack.GetSockOpt(t, s, s.ep, linux.AF_UNIX, s.ep.Type(), level, name, outPtr, outLen) } // blockingAccept implements a blocking version of accept(2), that is, if no // connections are ready to be accept, it will block until one becomes ready. -func (s *SocketVFS2) blockingAccept(t *kernel.Task) (transport.Endpoint, *syserr.Error) { +func (s *SocketVFS2) blockingAccept(t *kernel.Task, peerAddr *tcpip.FullAddress) (transport.Endpoint, *syserr.Error) { // Register for notifications. e, ch := waiter.NewChannelEntry(nil) s.socketOpsCommon.EventRegister(&e, waiter.EventIn) @@ -105,7 +129,7 @@ func (s *SocketVFS2) blockingAccept(t *kernel.Task) (transport.Endpoint, *syserr // Try to accept the connection; if it fails, then wait until we get a // notification. for { - if ep, err := s.ep.Accept(); err != syserr.ErrWouldBlock { + if ep, err := s.ep.Accept(peerAddr); err != syserr.ErrWouldBlock { return ep, err } @@ -118,15 +142,18 @@ func (s *SocketVFS2) blockingAccept(t *kernel.Task) (transport.Endpoint, *syserr // Accept implements the linux syscall accept(2) for sockets backed by // a transport.Endpoint. func (s *SocketVFS2) Accept(t *kernel.Task, peerRequested bool, flags int, blocking bool) (int32, linux.SockAddr, uint32, *syserr.Error) { - // Issue the accept request to get the new endpoint. - ep, err := s.ep.Accept() + var peerAddr *tcpip.FullAddress + if peerRequested { + peerAddr = &tcpip.FullAddress{} + } + ep, err := s.ep.Accept(peerAddr) if err != nil { if err != syserr.ErrWouldBlock || !blocking { return 0, nil, 0, err } var err *syserr.Error - ep, err = s.blockingAccept(t) + ep, err = s.blockingAccept(t, peerAddr) if err != nil { return 0, nil, 0, err } @@ -144,13 +171,8 @@ func (s *SocketVFS2) Accept(t *kernel.Task, peerRequested bool, flags int, block var addr linux.SockAddr var addrLen uint32 - if peerRequested { - // Get address of the peer. - var err *syserr.Error - addr, addrLen, err = ns.Impl().(*SocketVFS2).GetPeerName(t) - if err != nil { - return 0, nil, 0, err - } + if peerAddr != nil { + addr, addrLen = netstack.ConvertAddress(linux.AF_UNIX, *peerAddr) } fd, e := t.NewFDFromVFS2(0, ns, kernel.FDFlags{ @@ -246,13 +268,17 @@ func (s *SocketVFS2) Read(ctx context.Context, dst usermem.IOSequence, opts vfs. if dst.NumBytes() == 0 { return 0, nil } - return dst.CopyOutFrom(ctx, &EndpointReader{ + r := &EndpointReader{ Ctx: ctx, Endpoint: s.ep, NumRights: 0, Peek: false, From: nil, - }) + } + n, err := dst.CopyOutFrom(ctx, r) + // Drop control messages. + r.Control.Release(ctx) + return n, err } // PWrite implements vfs.FileDescriptionImpl. |