github.com/SagerNet/gvisor@v0.0.0-20210707092255-7731c139d75c/pkg/sentry/socket/hostinet/socket_vfs2.go (about)

     1  // Copyright 2020 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 hostinet
    16  
    17  import (
    18  	"golang.org/x/sys/unix"
    19  	"github.com/SagerNet/gvisor/pkg/abi/linux"
    20  	"github.com/SagerNet/gvisor/pkg/context"
    21  	"github.com/SagerNet/gvisor/pkg/errors/linuxerr"
    22  	"github.com/SagerNet/gvisor/pkg/fdnotifier"
    23  	"github.com/SagerNet/gvisor/pkg/sentry/arch"
    24  	"github.com/SagerNet/gvisor/pkg/sentry/fsimpl/sockfs"
    25  	"github.com/SagerNet/gvisor/pkg/sentry/hostfd"
    26  	"github.com/SagerNet/gvisor/pkg/sentry/kernel"
    27  	"github.com/SagerNet/gvisor/pkg/sentry/socket"
    28  	"github.com/SagerNet/gvisor/pkg/sentry/vfs"
    29  	"github.com/SagerNet/gvisor/pkg/syserr"
    30  	"github.com/SagerNet/gvisor/pkg/syserror"
    31  	"github.com/SagerNet/gvisor/pkg/usermem"
    32  	"github.com/SagerNet/gvisor/pkg/waiter"
    33  )
    34  
    35  // +stateify savable
    36  type socketVFS2 struct {
    37  	vfsfd vfs.FileDescription
    38  	vfs.FileDescriptionDefaultImpl
    39  	vfs.LockFD
    40  
    41  	// We store metadata for hostinet sockets internally. Technically, we should
    42  	// access metadata (e.g. through stat, chmod) on the host for correctness,
    43  	// but this is not very useful for inet socket fds, which do not belong to a
    44  	// concrete file anyway.
    45  	vfs.DentryMetadataFileDescriptionImpl
    46  
    47  	socketOpsCommon
    48  }
    49  
    50  var _ = socket.SocketVFS2(&socketVFS2{})
    51  
    52  func newVFS2Socket(t *kernel.Task, family int, stype linux.SockType, protocol int, fd int, flags uint32) (*vfs.FileDescription, *syserr.Error) {
    53  	mnt := t.Kernel().SocketMount()
    54  	d := sockfs.NewDentry(t, mnt)
    55  	defer d.DecRef(t)
    56  
    57  	s := &socketVFS2{
    58  		socketOpsCommon: socketOpsCommon{
    59  			family:   family,
    60  			stype:    stype,
    61  			protocol: protocol,
    62  			fd:       fd,
    63  		},
    64  	}
    65  	s.LockFD.Init(&vfs.FileLocks{})
    66  	if err := fdnotifier.AddFD(int32(fd), &s.queue); err != nil {
    67  		return nil, syserr.FromError(err)
    68  	}
    69  	vfsfd := &s.vfsfd
    70  	if err := vfsfd.Init(s, linux.O_RDWR|(flags&linux.O_NONBLOCK), mnt, d, &vfs.FileDescriptionOptions{
    71  		DenyPRead:         true,
    72  		DenyPWrite:        true,
    73  		UseDentryMetadata: true,
    74  	}); err != nil {
    75  		fdnotifier.RemoveFD(int32(s.fd))
    76  		return nil, syserr.FromError(err)
    77  	}
    78  	return vfsfd, nil
    79  }
    80  
    81  // Release implements vfs.FileDescriptionImpl.Release.
    82  func (s *socketVFS2) Release(ctx context.Context) {
    83  	kernel.KernelFromContext(ctx).DeleteSocketVFS2(&s.vfsfd)
    84  	s.socketOpsCommon.Release(ctx)
    85  }
    86  
    87  // Readiness implements waiter.Waitable.Readiness.
    88  func (s *socketVFS2) Readiness(mask waiter.EventMask) waiter.EventMask {
    89  	return s.socketOpsCommon.Readiness(mask)
    90  }
    91  
    92  // EventRegister implements waiter.Waitable.EventRegister.
    93  func (s *socketVFS2) EventRegister(e *waiter.Entry, mask waiter.EventMask) {
    94  	s.socketOpsCommon.EventRegister(e, mask)
    95  }
    96  
    97  // EventUnregister implements waiter.Waitable.EventUnregister.
    98  func (s *socketVFS2) EventUnregister(e *waiter.Entry) {
    99  	s.socketOpsCommon.EventUnregister(e)
   100  }
   101  
   102  // Ioctl implements vfs.FileDescriptionImpl.
   103  func (s *socketVFS2) Ioctl(ctx context.Context, uio usermem.IO, args arch.SyscallArguments) (uintptr, error) {
   104  	return ioctl(ctx, s.fd, uio, args)
   105  }
   106  
   107  // PRead implements vfs.FileDescriptionImpl.PRead.
   108  func (s *socketVFS2) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error) {
   109  	return 0, linuxerr.ESPIPE
   110  }
   111  
   112  // Read implements vfs.FileDescriptionImpl.
   113  func (s *socketVFS2) Read(ctx context.Context, dst usermem.IOSequence, opts vfs.ReadOptions) (int64, error) {
   114  	// All flags other than RWF_NOWAIT should be ignored.
   115  	// TODO(github.com/SagerNet/issue/2601): Support RWF_NOWAIT.
   116  	if opts.Flags != 0 {
   117  		return 0, syserror.EOPNOTSUPP
   118  	}
   119  
   120  	reader := hostfd.GetReadWriterAt(int32(s.fd), -1, opts.Flags)
   121  	n, err := dst.CopyOutFrom(ctx, reader)
   122  	hostfd.PutReadWriterAt(reader)
   123  	return int64(n), err
   124  }
   125  
   126  // PWrite implements vfs.FileDescriptionImpl.
   127  func (s *socketVFS2) PWrite(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.WriteOptions) (int64, error) {
   128  	return 0, linuxerr.ESPIPE
   129  }
   130  
   131  // Write implements vfs.FileDescriptionImpl.
   132  func (s *socketVFS2) Write(ctx context.Context, src usermem.IOSequence, opts vfs.WriteOptions) (int64, error) {
   133  	// All flags other than RWF_NOWAIT should be ignored.
   134  	// TODO(github.com/SagerNet/issue/2601): Support RWF_NOWAIT.
   135  	if opts.Flags != 0 {
   136  		return 0, syserror.EOPNOTSUPP
   137  	}
   138  
   139  	writer := hostfd.GetReadWriterAt(int32(s.fd), -1, opts.Flags)
   140  	n, err := src.CopyInTo(ctx, writer)
   141  	hostfd.PutReadWriterAt(writer)
   142  	return int64(n), err
   143  }
   144  
   145  type socketProviderVFS2 struct {
   146  	family int
   147  }
   148  
   149  // Socket implements socket.ProviderVFS2.Socket.
   150  func (p *socketProviderVFS2) Socket(t *kernel.Task, stypeflags linux.SockType, protocol int) (*vfs.FileDescription, *syserr.Error) {
   151  	// Check that we are using the host network stack.
   152  	stack := t.NetworkContext()
   153  	if stack == nil {
   154  		return nil, nil
   155  	}
   156  	if _, ok := stack.(*Stack); !ok {
   157  		return nil, nil
   158  	}
   159  
   160  	// Only accept TCP and UDP.
   161  	stype := stypeflags & linux.SOCK_TYPE_MASK
   162  	switch stype {
   163  	case unix.SOCK_STREAM:
   164  		switch protocol {
   165  		case 0, unix.IPPROTO_TCP:
   166  			// ok
   167  		default:
   168  			return nil, nil
   169  		}
   170  	case unix.SOCK_DGRAM:
   171  		switch protocol {
   172  		case 0, unix.IPPROTO_UDP:
   173  			// ok
   174  		default:
   175  			return nil, nil
   176  		}
   177  	default:
   178  		return nil, nil
   179  	}
   180  
   181  	// Conservatively ignore all flags specified by the application and add
   182  	// SOCK_NONBLOCK since socketOperations requires it. Pass a protocol of 0
   183  	// to simplify the syscall filters, since 0 and IPPROTO_* are equivalent.
   184  	fd, err := unix.Socket(p.family, int(stype)|unix.SOCK_NONBLOCK|unix.SOCK_CLOEXEC, 0)
   185  	if err != nil {
   186  		return nil, syserr.FromError(err)
   187  	}
   188  	return newVFS2Socket(t, p.family, stype, protocol, fd, uint32(stypeflags&unix.SOCK_NONBLOCK))
   189  }
   190  
   191  // Pair implements socket.Provider.Pair.
   192  func (p *socketProviderVFS2) Pair(t *kernel.Task, stype linux.SockType, protocol int) (*vfs.FileDescription, *vfs.FileDescription, *syserr.Error) {
   193  	// Not supported by AF_INET/AF_INET6.
   194  	return nil, nil, nil
   195  }