github.com/MerlinKodo/gvisor@v0.0.0-20231110090155-957f62ecf90e/pkg/sentry/kernel/abstract_socket_namespace.go (about)

     1  // Copyright 2018 The gVisor Authors.
     2  //
     3  // Licensed under the Apache License, Version 2.0 (the "License");
     4  // you may not use this file except in compliance with the License.
     5  // You may obtain a copy of the License at
     6  //
     7  //     http://www.apache.org/licenses/LICENSE-2.0
     8  //
     9  // Unless required by applicable law or agreed to in writing, software
    10  // distributed under the License is distributed on an "AS IS" BASIS,
    11  // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    12  // See the License for the specific language governing permissions and
    13  // limitations under the License.
    14  
    15  package kernel
    16  
    17  import (
    18  	"fmt"
    19  
    20  	"github.com/MerlinKodo/gvisor/pkg/context"
    21  	"github.com/MerlinKodo/gvisor/pkg/refs"
    22  	"github.com/MerlinKodo/gvisor/pkg/sentry/socket/unix/transport"
    23  	"github.com/MerlinKodo/gvisor/pkg/sync"
    24  	"golang.org/x/sys/unix"
    25  )
    26  
    27  // +stateify savable
    28  type abstractEndpoint struct {
    29  	ep     transport.BoundEndpoint
    30  	socket refs.TryRefCounter
    31  	name   string
    32  	ns     *AbstractSocketNamespace
    33  }
    34  
    35  // AbstractSocketNamespace is used to implement the Linux abstract socket functionality.
    36  //
    37  // +stateify savable
    38  type AbstractSocketNamespace struct {
    39  	mu sync.Mutex `state:"nosave"`
    40  
    41  	// Keeps a mapping from name to endpoint. AbstractSocketNamespace does not hold
    42  	// any references on any sockets that it contains; when retrieving a socket,
    43  	// TryIncRef() must be called in case the socket is concurrently being
    44  	// destroyed. It is the responsibility of the socket to remove itself from the
    45  	// abstract socket namespace when it is destroyed.
    46  	endpoints map[string]abstractEndpoint
    47  }
    48  
    49  // NewAbstractSocketNamespace returns a new AbstractSocketNamespace.
    50  func NewAbstractSocketNamespace() *AbstractSocketNamespace {
    51  	return &AbstractSocketNamespace{
    52  		endpoints: make(map[string]abstractEndpoint),
    53  	}
    54  }
    55  
    56  // A boundEndpoint wraps a transport.BoundEndpoint to maintain a reference on
    57  // its backing socket.
    58  type boundEndpoint struct {
    59  	transport.BoundEndpoint
    60  	socket refs.TryRefCounter
    61  }
    62  
    63  // Release implements transport.BoundEndpoint.Release.
    64  func (e *boundEndpoint) Release(ctx context.Context) {
    65  	e.socket.DecRef(ctx)
    66  	e.BoundEndpoint.Release(ctx)
    67  }
    68  
    69  // BoundEndpoint retrieves the endpoint bound to the given name. The return
    70  // value is nil if no endpoint was bound.
    71  func (a *AbstractSocketNamespace) BoundEndpoint(name string) transport.BoundEndpoint {
    72  	a.mu.Lock()
    73  	defer a.mu.Unlock()
    74  
    75  	ep, ok := a.endpoints[name]
    76  	if !ok {
    77  		return nil
    78  	}
    79  
    80  	if !ep.socket.TryIncRef() {
    81  		// The socket has reached zero references and is being destroyed.
    82  		return nil
    83  	}
    84  
    85  	return &boundEndpoint{ep.ep, ep.socket}
    86  }
    87  
    88  // Bind binds the given socket.
    89  //
    90  // When the last reference managed by socket is dropped, ep may be removed from the
    91  // namespace.
    92  func (a *AbstractSocketNamespace) Bind(ctx context.Context, name string, ep transport.BoundEndpoint, socket refs.TryRefCounter) error {
    93  	a.mu.Lock()
    94  	defer a.mu.Unlock()
    95  
    96  	// Check if there is already a socket (which has not yet been destroyed) bound at name.
    97  	if ep, ok := a.endpoints[name]; ok {
    98  		if ep.socket.TryIncRef() {
    99  			ep.socket.DecRef(ctx)
   100  			return unix.EADDRINUSE
   101  		}
   102  	}
   103  
   104  	ae := abstractEndpoint{ep: ep, name: name, ns: a}
   105  	ae.socket = socket
   106  	a.endpoints[name] = ae
   107  	return nil
   108  }
   109  
   110  // Remove removes the specified socket at name from the abstract socket
   111  // namespace, if it has not yet been replaced.
   112  func (a *AbstractSocketNamespace) Remove(name string, socket refs.TryRefCounter) {
   113  	a.mu.Lock()
   114  	defer a.mu.Unlock()
   115  
   116  	ep, ok := a.endpoints[name]
   117  	if !ok {
   118  		// We never delete a map entry apart from a socket's destructor (although the
   119  		// map entry may be overwritten). Therefore, a socket should exist, even if it
   120  		// may not be the one we expect.
   121  		panic(fmt.Sprintf("expected socket to exist at '%s' in abstract socket namespace", name))
   122  	}
   123  
   124  	// A Bind() operation may race with callers of Remove(), e.g. in the
   125  	// following case:
   126  	//   socket1 reaches zero references and begins destruction
   127  	//   a.Bind("foo", ep, socket2) replaces socket1 with socket2
   128  	//   socket1's destructor calls a.Remove("foo", socket1)
   129  	//
   130  	// Therefore, we need to check that the socket at name is what we expect
   131  	// before modifying the map.
   132  	if ep.socket == socket {
   133  		delete(a.endpoints, name)
   134  	}
   135  }