diff options
Diffstat (limited to 'pkg/sentry/socket/unix')
-rw-r--r-- | pkg/sentry/socket/unix/BUILD | 2 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/connectioned.go | 15 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/connectionless.go | 12 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/queue.go | 13 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/unix.go | 48 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/unix.go | 78 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/unix_vfs2.go | 38 |
7 files changed, 124 insertions, 82 deletions
diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index de2cc4bdf..061a689a9 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -21,6 +21,7 @@ go_library( "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", + "//pkg/sentry/fs/lock", "//pkg/sentry/fsimpl/sockfs", "//pkg/sentry/kernel", "//pkg/sentry/kernel/time", @@ -34,5 +35,6 @@ go_library( "//pkg/tcpip", "//pkg/usermem", "//pkg/waiter", + "//tools/go_marshal/marshal", ], ) diff --git a/pkg/sentry/socket/unix/transport/connectioned.go b/pkg/sentry/socket/unix/transport/connectioned.go index ce5b94ee7..c67b602f0 100644 --- a/pkg/sentry/socket/unix/transport/connectioned.go +++ b/pkg/sentry/socket/unix/transport/connectioned.go @@ -211,7 +211,7 @@ func (e *connectionedEndpoint) Listening() bool { // The socket will be a fresh state after a call to close and may be reused. // That is, close may be used to "unbind" or "disconnect" the socket in error // paths. -func (e *connectionedEndpoint) Close() { +func (e *connectionedEndpoint) Close(ctx context.Context) { e.Lock() var c ConnectedEndpoint var r Receiver @@ -233,7 +233,7 @@ func (e *connectionedEndpoint) Close() { case e.Listening(): close(e.acceptedChan) for n := range e.acceptedChan { - n.Close() + n.Close(ctx) } e.acceptedChan = nil e.path = "" @@ -241,18 +241,18 @@ func (e *connectionedEndpoint) Close() { e.Unlock() if c != nil { c.CloseNotify() - c.Release() + c.Release(ctx) } if r != nil { r.CloseNotify() - r.Release() + r.Release(ctx) } } // BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. func (e *connectionedEndpoint) BidirectionalConnect(ctx context.Context, ce ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *syserr.Error { if ce.Type() != e.stype { - return syserr.ErrConnectionRefused + return syserr.ErrWrongProtocolForSocket } // Check if ce is e to avoid a deadlock. @@ -340,7 +340,7 @@ func (e *connectionedEndpoint) BidirectionalConnect(ctx context.Context, ce Conn return nil default: // Busy; return ECONNREFUSED per spec. - ne.Close() + ne.Close(ctx) e.Unlock() ce.Unlock() return syserr.ErrConnectionRefused @@ -476,6 +476,9 @@ func (e *connectionedEndpoint) Readiness(mask waiter.EventMask) waiter.EventMask // State implements socket.Socket.State. func (e *connectionedEndpoint) State() uint32 { + e.Lock() + defer e.Unlock() + if e.Connected() { return linux.SS_CONNECTED } diff --git a/pkg/sentry/socket/unix/transport/connectionless.go b/pkg/sentry/socket/unix/transport/connectionless.go index 4b06d63ac..70ee8f9b8 100644 --- a/pkg/sentry/socket/unix/transport/connectionless.go +++ b/pkg/sentry/socket/unix/transport/connectionless.go @@ -54,10 +54,10 @@ func (e *connectionlessEndpoint) isBound() bool { // Close puts the endpoint in a closed state and frees all resources associated // with it. -func (e *connectionlessEndpoint) Close() { +func (e *connectionlessEndpoint) Close(ctx context.Context) { e.Lock() if e.connected != nil { - e.connected.Release() + e.connected.Release(ctx) e.connected = nil } @@ -71,7 +71,7 @@ func (e *connectionlessEndpoint) Close() { e.Unlock() r.CloseNotify() - r.Release() + r.Release(ctx) } // BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. @@ -108,10 +108,10 @@ func (e *connectionlessEndpoint) SendMsg(ctx context.Context, data [][]byte, c C if err != nil { return 0, syserr.ErrInvalidEndpointState } - defer connected.Release() + defer connected.Release(ctx) e.Lock() - n, notify, err := connected.Send(data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) + n, notify, err := connected.Send(ctx, data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) e.Unlock() if notify { @@ -135,7 +135,7 @@ func (e *connectionlessEndpoint) Connect(ctx context.Context, server BoundEndpoi e.Lock() if e.connected != nil { - e.connected.Release() + e.connected.Release(ctx) } e.connected = connected e.Unlock() diff --git a/pkg/sentry/socket/unix/transport/queue.go b/pkg/sentry/socket/unix/transport/queue.go index d8f3ad63d..ef6043e19 100644 --- a/pkg/sentry/socket/unix/transport/queue.go +++ b/pkg/sentry/socket/unix/transport/queue.go @@ -15,6 +15,7 @@ package transport import ( + "gvisor.dev/gvisor/pkg/context" "gvisor.dev/gvisor/pkg/refs" "gvisor.dev/gvisor/pkg/sync" "gvisor.dev/gvisor/pkg/syserr" @@ -57,10 +58,10 @@ func (q *queue) Close() { // Both the read and write queues must be notified after resetting: // q.ReaderQueue.Notify(waiter.EventIn) // q.WriterQueue.Notify(waiter.EventOut) -func (q *queue) Reset() { +func (q *queue) Reset(ctx context.Context) { q.mu.Lock() for cur := q.dataList.Front(); cur != nil; cur = cur.Next() { - cur.Release() + cur.Release(ctx) } q.dataList.Reset() q.used = 0 @@ -68,8 +69,8 @@ func (q *queue) Reset() { } // DecRef implements RefCounter.DecRef with destructor q.Reset. -func (q *queue) DecRef() { - q.DecRefWithDestructor(q.Reset) +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. } @@ -111,7 +112,7 @@ func (q *queue) IsWritable() bool { // // If notify is true, ReaderQueue.Notify must be called: // q.ReaderQueue.Notify(waiter.EventIn) -func (q *queue) Enqueue(data [][]byte, c ControlMessages, from tcpip.FullAddress, discardEmpty bool, truncate bool) (l int64, notify bool, err *syserr.Error) { +func (q *queue) Enqueue(ctx context.Context, data [][]byte, c ControlMessages, from tcpip.FullAddress, discardEmpty bool, truncate bool) (l int64, notify bool, err *syserr.Error) { q.mu.Lock() if q.closed { @@ -124,7 +125,7 @@ func (q *queue) Enqueue(data [][]byte, c ControlMessages, from tcpip.FullAddress } if discardEmpty && l == 0 { q.mu.Unlock() - c.Release() + c.Release(ctx) return 0, false, nil } diff --git a/pkg/sentry/socket/unix/transport/unix.go b/pkg/sentry/socket/unix/transport/unix.go index 2f1b127df..475d7177e 100644 --- a/pkg/sentry/socket/unix/transport/unix.go +++ b/pkg/sentry/socket/unix/transport/unix.go @@ -37,7 +37,7 @@ type RightsControlMessage interface { Clone() RightsControlMessage // Release releases any resources owned by the RightsControlMessage. - Release() + Release(ctx context.Context) } // A CredentialsControlMessage is a control message containing Unix credentials. @@ -74,9 +74,9 @@ func (c *ControlMessages) Clone() ControlMessages { } // Release releases both the credentials and the rights. -func (c *ControlMessages) Release() { +func (c *ControlMessages) Release(ctx context.Context) { if c.Rights != nil { - c.Rights.Release() + c.Rights.Release(ctx) } *c = ControlMessages{} } @@ -90,7 +90,7 @@ type Endpoint interface { // Close puts the endpoint in a closed state and frees all resources // associated with it. - Close() + Close(ctx context.Context) // RecvMsg reads data and a control message from the endpoint. This method // does not block if there is no data pending. @@ -252,7 +252,7 @@ type BoundEndpoint interface { // Release releases any resources held by the BoundEndpoint. It must be // called before dropping all references to a BoundEndpoint returned by a // function. - Release() + Release(ctx context.Context) } // message represents a message passed over a Unix domain socket. @@ -281,8 +281,8 @@ func (m *message) Length() int64 { } // Release releases any resources held by the message. -func (m *message) Release() { - m.Control.Release() +func (m *message) Release(ctx context.Context) { + m.Control.Release(ctx) } // Peek returns a copy of the message. @@ -304,7 +304,7 @@ type Receiver interface { // See Endpoint.RecvMsg for documentation on shared arguments. // // notify indicates if RecvNotify should be called. - Recv(data [][]byte, creds bool, numRights int, peek bool) (recvLen, msgLen int64, cm ControlMessages, CMTruncated bool, source tcpip.FullAddress, notify bool, err *syserr.Error) + Recv(ctx context.Context, data [][]byte, creds bool, numRights int, peek bool) (recvLen, msgLen int64, cm ControlMessages, CMTruncated bool, source tcpip.FullAddress, notify bool, err *syserr.Error) // RecvNotify notifies the Receiver of a successful Recv. This must not be // called while holding any endpoint locks. @@ -333,7 +333,7 @@ type Receiver interface { // Release releases any resources owned by the Receiver. It should be // called before droping all references to a Receiver. - Release() + Release(ctx context.Context) } // queueReceiver implements Receiver for datagram sockets. @@ -344,7 +344,7 @@ type queueReceiver struct { } // Recv implements Receiver.Recv. -func (q *queueReceiver) Recv(data [][]byte, creds bool, numRights int, peek bool) (int64, int64, ControlMessages, bool, tcpip.FullAddress, bool, *syserr.Error) { +func (q *queueReceiver) Recv(ctx context.Context, data [][]byte, creds bool, numRights int, peek bool) (int64, int64, ControlMessages, bool, tcpip.FullAddress, bool, *syserr.Error) { var m *message var notify bool var err *syserr.Error @@ -398,8 +398,8 @@ func (q *queueReceiver) RecvMaxQueueSize() int64 { } // Release implements Receiver.Release. -func (q *queueReceiver) Release() { - q.readQueue.DecRef() +func (q *queueReceiver) Release(ctx context.Context) { + q.readQueue.DecRef(ctx) } // streamQueueReceiver implements Receiver for stream sockets. @@ -456,7 +456,7 @@ func (q *streamQueueReceiver) RecvMaxQueueSize() int64 { } // Recv implements Receiver.Recv. -func (q *streamQueueReceiver) Recv(data [][]byte, wantCreds bool, numRights int, peek bool) (int64, int64, ControlMessages, bool, tcpip.FullAddress, bool, *syserr.Error) { +func (q *streamQueueReceiver) Recv(ctx context.Context, data [][]byte, wantCreds bool, numRights int, peek bool) (int64, int64, ControlMessages, bool, tcpip.FullAddress, bool, *syserr.Error) { q.mu.Lock() defer q.mu.Unlock() @@ -502,7 +502,7 @@ func (q *streamQueueReceiver) Recv(data [][]byte, wantCreds bool, numRights int, var cmTruncated bool if c.Rights != nil && numRights == 0 { - c.Rights.Release() + c.Rights.Release(ctx) c.Rights = nil cmTruncated = true } @@ -557,7 +557,7 @@ func (q *streamQueueReceiver) Recv(data [][]byte, wantCreds bool, numRights int, // Consume rights. if numRights == 0 { cmTruncated = true - q.control.Rights.Release() + q.control.Rights.Release(ctx) } else { c.Rights = q.control.Rights haveRights = true @@ -582,7 +582,7 @@ type ConnectedEndpoint interface { // // syserr.ErrWouldBlock can be returned along with a partial write if // the caller should block to send the rest of the data. - Send(data [][]byte, c ControlMessages, from tcpip.FullAddress) (n int64, notify bool, err *syserr.Error) + Send(ctx context.Context, data [][]byte, c ControlMessages, from tcpip.FullAddress) (n int64, notify bool, err *syserr.Error) // SendNotify notifies the ConnectedEndpoint of a successful Send. This // must not be called while holding any endpoint locks. @@ -616,7 +616,7 @@ type ConnectedEndpoint interface { // Release releases any resources owned by the ConnectedEndpoint. It should // be called before droping all references to a ConnectedEndpoint. - Release() + Release(ctx context.Context) // CloseUnread sets the fact that this end is closed with unread data to // the peer socket. @@ -654,7 +654,7 @@ func (e *connectedEndpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) } // Send implements ConnectedEndpoint.Send. -func (e *connectedEndpoint) Send(data [][]byte, c ControlMessages, from tcpip.FullAddress) (int64, bool, *syserr.Error) { +func (e *connectedEndpoint) Send(ctx context.Context, data [][]byte, c ControlMessages, from tcpip.FullAddress) (int64, bool, *syserr.Error) { discardEmpty := false truncate := false if e.endpoint.Type() == linux.SOCK_STREAM { @@ -669,7 +669,7 @@ func (e *connectedEndpoint) Send(data [][]byte, c ControlMessages, from tcpip.Fu truncate = true } - return e.writeQueue.Enqueue(data, c, from, discardEmpty, truncate) + return e.writeQueue.Enqueue(ctx, data, c, from, discardEmpty, truncate) } // SendNotify implements ConnectedEndpoint.SendNotify. @@ -707,8 +707,8 @@ func (e *connectedEndpoint) SendMaxQueueSize() int64 { } // Release implements ConnectedEndpoint.Release. -func (e *connectedEndpoint) Release() { - e.writeQueue.DecRef() +func (e *connectedEndpoint) Release(ctx context.Context) { + e.writeQueue.DecRef(ctx) } // CloseUnread implements ConnectedEndpoint.CloseUnread. @@ -798,7 +798,7 @@ func (e *baseEndpoint) RecvMsg(ctx context.Context, data [][]byte, creds bool, n return 0, 0, ControlMessages{}, false, syserr.ErrNotConnected } - recvLen, msgLen, cms, cmt, a, notify, err := e.receiver.Recv(data, creds, numRights, peek) + recvLen, msgLen, cms, cmt, a, notify, err := e.receiver.Recv(ctx, data, creds, numRights, peek) e.Unlock() if err != nil { return 0, 0, ControlMessages{}, false, err @@ -827,7 +827,7 @@ func (e *baseEndpoint) SendMsg(ctx context.Context, data [][]byte, c ControlMess return 0, syserr.ErrAlreadyConnected } - n, notify, err := e.connected.Send(data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) + n, notify, err := e.connected.Send(ctx, data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) e.Unlock() if notify { @@ -1001,6 +1001,6 @@ func (e *baseEndpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) { } // Release implements BoundEndpoint.Release. -func (*baseEndpoint) Release() { +func (*baseEndpoint) Release(context.Context) { // Binding a baseEndpoint doesn't take a reference. } diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 5b29e9d7f..2b8454edb 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -40,6 +40,7 @@ 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, @@ -61,7 +62,7 @@ type SocketOperations struct { // New creates a new unix socket. func New(ctx context.Context, endpoint transport.Endpoint, stype linux.SockType) *fs.File { dirent := socket.NewDirent(ctx, unixSocketDevice) - defer dirent.DecRef() + defer dirent.DecRef(ctx) return NewWithDirent(ctx, dirent, endpoint, stype, fs.FileFlags{Read: true, Write: true, NonSeekable: true}) } @@ -96,17 +97,17 @@ type socketOpsCommon struct { } // DecRef implements RefCounter.DecRef. -func (s *socketOpsCommon) DecRef() { - s.DecRefWithDestructor(func() { - s.ep.Close() +func (s *socketOpsCommon) DecRef(ctx context.Context) { + s.DecRefWithDestructor(ctx, func(context.Context) { + s.ep.Close(ctx) }) } // Release implemements fs.FileOperations.Release. -func (s *socketOpsCommon) 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() + s.DecRef(ctx) } func (s *socketOpsCommon) isPacket() bool { @@ -184,7 +185,7 @@ 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) (interface{}, *syserr.Error) { +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) } @@ -233,7 +234,7 @@ func (s *SocketOperations) Accept(t *kernel.Task, peerRequested bool, flags int, } ns := New(t, ep, s.stype) - defer ns.DecRef() + defer ns.DecRef(t) if flags&linux.SOCK_NONBLOCK != 0 { flags := ns.Flags() @@ -283,7 +284,7 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { if t.IsNetworkNamespaced() { return syserr.ErrInvalidEndpointState } - if err := t.AbstractSockets().Bind(p[1:], bep, s); err != nil { + if err := t.AbstractSockets().Bind(t, p[1:], bep, s); err != nil { // syserr.ErrPortInUse corresponds to EADDRINUSE. return syserr.ErrPortInUse } @@ -293,7 +294,7 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { var name string cwd := t.FSContext().WorkingDirectory() - defer cwd.DecRef() + defer cwd.DecRef(t) // Is there no slash at all? if !strings.Contains(p, "/") { @@ -301,7 +302,7 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { name = p } else { root := t.FSContext().RootDirectory() - defer root.DecRef() + defer root.DecRef(t) // Find the last path component, we know that something follows // that final slash, otherwise extractPath() would have failed. lastSlash := strings.LastIndex(p, "/") @@ -317,7 +318,7 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { // No path available. return syserr.ErrNoSuchFile } - defer d.DecRef() + defer d.DecRef(t) name = p[lastSlash+1:] } @@ -331,7 +332,7 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { if err != nil { return syserr.ErrPortInUse } - childDir.DecRef() + childDir.DecRef(t) } return nil @@ -377,9 +378,9 @@ func extractEndpoint(t *kernel.Task, sockaddr []byte) (transport.BoundEndpoint, FollowFinalSymlink: true, } ep, e := t.Kernel().VFS().BoundEndpointAt(t, t.Credentials(), &pop, &vfs.BoundEndpointOptions{path}) - root.DecRef() + root.DecRef(t) if relPath { - start.DecRef() + start.DecRef(t) } if e != nil { return nil, syserr.FromError(e) @@ -392,15 +393,15 @@ func extractEndpoint(t *kernel.Task, sockaddr []byte) (transport.BoundEndpoint, cwd := t.FSContext().WorkingDirectory() remainingTraversals := uint(fs.DefaultTraversalLimit) d, e := t.MountNamespace().FindInode(t, root, cwd, path, &remainingTraversals) - cwd.DecRef() - root.DecRef() + cwd.DecRef(t) + root.DecRef(t) if e != nil { return nil, syserr.FromError(e) } // Extract the endpoint if one is there. ep := d.Inode.BoundEndpoint(path) - d.DecRef() + d.DecRef(t) if ep == nil { // No socket! return nil, syserr.ErrConnectionRefused @@ -414,10 +415,21 @@ func (s *socketOpsCommon) Connect(t *kernel.Task, sockaddr []byte, blocking bool if err != nil { return err } - defer ep.Release() + defer ep.Release(t) // Connect the server endpoint. - return s.ep.Connect(t, ep) + err = s.ep.Connect(t, ep) + + if err == syserr.ErrWrongProtocolForSocket { + // Linux for abstract sockets returns ErrConnectionRefused + // instead of ErrWrongProtocolForSocket. + path, _ := extractPath(sockaddr) + if len(path) > 0 && path[0] == 0 { + err = syserr.ErrConnectionRefused + } + } + + return err } // Write implements fs.FileOperations.Write. @@ -448,15 +460,25 @@ func (s *socketOpsCommon) SendMsg(t *kernel.Task, src usermem.IOSequence, to []b To: nil, } if len(to) > 0 { - ep, err := extractEndpoint(t, to) - if err != nil { - return 0, err - } - defer ep.Release() - w.To = ep + switch s.stype { + case linux.SOCK_SEQPACKET: + to = nil + case linux.SOCK_STREAM: + if s.State() == linux.SS_CONNECTED { + return 0, syserr.ErrAlreadyConnected + } + return 0, syserr.ErrNotSupported + default: + ep, err := extractEndpoint(t, to) + if err != nil { + return 0, err + } + defer ep.Release(t) + w.To = ep - if ep.Passcred() && w.Control.Credentials == nil { - w.Control.Credentials = control.MakeCreds(t) + if ep.Passcred() && w.Control.Credentials == nil { + w.Control.Credentials = control.MakeCreds(t) + } } } diff --git a/pkg/sentry/socket/unix/unix_vfs2.go b/pkg/sentry/socket/unix/unix_vfs2.go index 45e109361..dfa25241a 100644 --- a/pkg/sentry/socket/unix/unix_vfs2.go +++ b/pkg/sentry/socket/unix/unix_vfs2.go @@ -19,6 +19,7 @@ import ( "gvisor.dev/gvisor/pkg/context" "gvisor.dev/gvisor/pkg/fspath" "gvisor.dev/gvisor/pkg/sentry/arch" + fslock "gvisor.dev/gvisor/pkg/sentry/fs/lock" "gvisor.dev/gvisor/pkg/sentry/fsimpl/sockfs" "gvisor.dev/gvisor/pkg/sentry/kernel" "gvisor.dev/gvisor/pkg/sentry/socket" @@ -31,6 +32,7 @@ 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, @@ -39,6 +41,7 @@ type SocketVFS2 struct { vfsfd vfs.FileDescription vfs.FileDescriptionDefaultImpl vfs.DentryMetadataFileDescriptionImpl + vfs.LockFD socketOpsCommon } @@ -51,7 +54,7 @@ func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType) mnt := t.Kernel().SocketMount() d := sockfs.NewDentry(t.Credentials(), mnt) - fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d) + fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d, &vfs.FileLocks{}) if err != nil { return nil, syserr.FromError(err) } @@ -60,7 +63,7 @@ func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType) // NewFileDescription creates and returns a socket file description // corresponding to the given mount and dentry. -func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint32, mnt *vfs.Mount, d *vfs.Dentry) (*vfs.FileDescription, error) { +func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint32, mnt *vfs.Mount, d *vfs.Dentry, locks *vfs.FileLocks) (*vfs.FileDescription, error) { // You can create AF_UNIX, SOCK_RAW sockets. They're the same as // SOCK_DGRAM and don't require CAP_NET_RAW. if stype == linux.SOCK_RAW { @@ -73,6 +76,7 @@ func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint3 stype: stype, }, } + sock.LockFD.Init(locks) vfsfd := &sock.vfsfd if err := vfsfd.Init(sock, flags, mnt, d, &vfs.FileDescriptionOptions{ DenyPRead: true, @@ -86,7 +90,7 @@ func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint3 // GetSockOpt implements the linux syscall getsockopt(2) for sockets backed by // a transport.Endpoint. -func (s *SocketVFS2) GetSockOpt(t *kernel.Task, level int, name int, outPtr usermem.Addr, outLen int) (interface{}, *syserr.Error) { +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) } @@ -132,7 +136,7 @@ func (s *SocketVFS2) Accept(t *kernel.Task, peerRequested bool, flags int, block if err != nil { return 0, nil, 0, err } - defer ns.DecRef() + defer ns.DecRef(t) if flags&linux.SOCK_NONBLOCK != 0 { ns.SetStatusFlags(t, t.Credentials(), linux.SOCK_NONBLOCK) @@ -179,19 +183,19 @@ func (s *SocketVFS2) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { if t.IsNetworkNamespaced() { return syserr.ErrInvalidEndpointState } - if err := t.AbstractSockets().Bind(p[1:], bep, s); err != nil { + if err := t.AbstractSockets().Bind(t, p[1:], bep, s); err != nil { // syserr.ErrPortInUse corresponds to EADDRINUSE. return syserr.ErrPortInUse } } else { path := fspath.Parse(p) root := t.FSContext().RootDirectoryVFS2() - defer root.DecRef() + defer root.DecRef(t) start := root relPath := !path.Absolute if relPath { start = t.FSContext().WorkingDirectoryVFS2() - defer start.DecRef() + defer start.DecRef(t) } pop := vfs.PathOperation{ Root: root, @@ -297,6 +301,16 @@ func (s *SocketVFS2) SetSockOpt(t *kernel.Task, level int, name int, optVal []by return netstack.SetSockOpt(t, s, s.ep, level, name, optVal) } +// LockPOSIX implements vfs.FileDescriptionImpl.LockPOSIX. +func (s *SocketVFS2) LockPOSIX(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, start, length uint64, whence int16, block fslock.Blocker) error { + return s.Locks().LockPOSIX(ctx, &s.vfsfd, uid, t, start, length, whence, block) +} + +// UnlockPOSIX implements vfs.FileDescriptionImpl.UnlockPOSIX. +func (s *SocketVFS2) UnlockPOSIX(ctx context.Context, uid fslock.UniqueID, start, length uint64, whence int16) error { + return s.Locks().UnlockPOSIX(ctx, &s.vfsfd, uid, start, length, whence) +} + // providerVFS2 is a unix domain socket provider for VFS2. type providerVFS2 struct{} @@ -319,7 +333,7 @@ func (*providerVFS2) Socket(t *kernel.Task, stype linux.SockType, protocol int) f, err := NewSockfsFile(t, ep, stype) if err != nil { - ep.Close() + ep.Close(t) return nil, err } return f, nil @@ -343,14 +357,14 @@ func (*providerVFS2) Pair(t *kernel.Task, stype linux.SockType, protocol int) (* ep1, ep2 := transport.NewPair(t, stype, t.Kernel()) s1, err := NewSockfsFile(t, ep1, stype) if err != nil { - ep1.Close() - ep2.Close() + ep1.Close(t) + ep2.Close(t) return nil, nil, err } s2, err := NewSockfsFile(t, ep2, stype) if err != nil { - s1.DecRef() - ep2.Close() + s1.DecRef(t) + ep2.Close(t) return nil, nil, err } |