decoder.go

     1  // Copyright 2009 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package gob
     6  
     7  import (
     8  	"bufio"
     9  	"errors"
    10  	"internal/saferio"
    11  	"io"
    12  	"reflect"
    13  	"sync"
    14  )
    15  
    16  // tooBig provides a sanity check for sizes; used in several places. Upper limit
    17  // of is 1GB on 32-bit systems, 8GB on 64-bit, allowing room to grow a little
    18  // without overflow.
    19  const tooBig = (1 << 30) << (^uint(0) >> 62)
    20  
    21  // A Decoder manages the receipt of type and data information read from the
    22  // remote side of a connection.  It is safe for concurrent use by multiple
    23  // goroutines.
    24  //
    25  // The Decoder does only basic sanity checking on decoded input sizes,
    26  // and its limits are not configurable. Take caution when decoding gob data
    27  // from untrusted sources.
    28  type Decoder struct {
    29  	mutex        sync.Mutex                              // each item must be received atomically
    30  	r            io.Reader                               // source of the data
    31  	buf          decBuffer                               // buffer for more efficient i/o from r
    32  	wireType     map[typeId]*wireType                    // map from remote ID to local description
    33  	decoderCache map[reflect.Type]map[typeId]**decEngine // cache of compiled engines
    34  	ignorerCache map[typeId]**decEngine                  // ditto for ignored objects
    35  	freeList     *decoderState                           // list of free decoderStates; avoids reallocation
    36  	countBuf     []byte                                  // used for decoding integers while parsing messages
    37  	err          error
    38  	// ignoreDepth tracks the depth of recursively parsed ignored fields
    39  	ignoreDepth int
    40  }
    41  
    42  // NewDecoder returns a new decoder that reads from the [io.Reader].
    43  // If r does not also implement [io.ByteReader], it will be wrapped in a
    44  // [bufio.Reader].
    45  func NewDecoder(r io.Reader) *Decoder {
    46  	dec := new(Decoder)
    47  	// We use the ability to read bytes as a plausible surrogate for buffering.
    48  	if _, ok := r.(io.ByteReader); !ok {
    49  		r = bufio.NewReader(r)
    50  	}
    51  	dec.r = r
    52  	dec.wireType = make(map[typeId]*wireType)
    53  	dec.decoderCache = make(map[reflect.Type]map[typeId]**decEngine)
    54  	dec.ignorerCache = make(map[typeId]**decEngine)
    55  	dec.countBuf = make([]byte, 9) // counts may be uint64s (unlikely!), require 9 bytes
    56  
    57  	return dec
    58  }
    59  
    60  // recvType loads the definition of a type.
    61  func (dec *Decoder) recvType(id typeId) {
    62  	// Have we already seen this type? That's an error
    63  	if id < firstUserId || dec.wireType[id] != nil {
    64  		dec.err = errors.New("gob: duplicate type received")
    65  		return
    66  	}
    67  
    68  	// Type:
    69  	wire := new(wireType)
    70  	dec.decodeValue(tWireType, reflect.ValueOf(wire))
    71  	if dec.err != nil {
    72  		return
    73  	}
    74  	// Remember we've seen this type.
    75  	dec.wireType[id] = wire
    76  }
    77  
    78  var errBadCount = errors.New("invalid message length")
    79  
    80  // recvMessage reads the next count-delimited item from the input. It is the converse
    81  // of Encoder.writeMessage. It returns false on EOF or other error reading the message.
    82  func (dec *Decoder) recvMessage() bool {
    83  	// Read a count.
    84  	nbytes, _, err := decodeUintReader(dec.r, dec.countBuf)
    85  	if err != nil {
    86  		dec.err = err
    87  		return false
    88  	}
    89  	if nbytes >= tooBig {
    90  		dec.err = errBadCount
    91  		return false
    92  	}
    93  	dec.readMessage(int(nbytes))
    94  	return dec.err == nil
    95  }
    96  
    97  // readMessage reads the next nbytes bytes from the input.
    98  func (dec *Decoder) readMessage(nbytes int) {
    99  	if dec.buf.Len() != 0 {
   100  		// The buffer should always be empty now.
   101  		panic("non-empty decoder buffer")
   102  	}
   103  	// Read the data
   104  	var buf []byte
   105  	buf, dec.err = saferio.ReadData(dec.r, uint64(nbytes))
   106  	dec.buf.SetBytes(buf)
   107  	if dec.err == io.EOF {
   108  		dec.err = io.ErrUnexpectedEOF
   109  	}
   110  }
   111  
   112  // toInt turns an encoded uint64 into an int, according to the marshaling rules.
   113  func toInt(x uint64) int64 {
   114  	i := int64(x >> 1)
   115  	if x&1 != 0 {
   116  		i = ^i
   117  	}
   118  	return i
   119  }
   120  
   121  func (dec *Decoder) nextInt() int64 {
   122  	n, _, err := decodeUintReader(&dec.buf, dec.countBuf)
   123  	if err != nil {
   124  		dec.err = err
   125  	}
   126  	return toInt(n)
   127  }
   128  
   129  func (dec *Decoder) nextUint() uint64 {
   130  	n, _, err := decodeUintReader(&dec.buf, dec.countBuf)
   131  	if err != nil {
   132  		dec.err = err
   133  	}
   134  	return n
   135  }
   136  
   137  // decodeTypeSequence parses:
   138  // TypeSequence
   139  //
   140  //	(TypeDefinition DelimitedTypeDefinition*)?
   141  //
   142  // and returns the type id of the next value. It returns -1 at
   143  // EOF.  Upon return, the remainder of dec.buf is the value to be
   144  // decoded. If this is an interface value, it can be ignored by
   145  // resetting that buffer.
   146  func (dec *Decoder) decodeTypeSequence(isInterface bool) typeId {
   147  	firstMessage := true
   148  	for dec.err == nil {
   149  		if dec.buf.Len() == 0 {
   150  			if !dec.recvMessage() {
   151  				// We can only return io.EOF if the input was empty.
   152  				// If we read one or more type spec messages,
   153  				// require a data item message to follow.
   154  				// If we hit an EOF before that, then give ErrUnexpectedEOF.
   155  				if !firstMessage && dec.err == io.EOF {
   156  					dec.err = io.ErrUnexpectedEOF
   157  				}
   158  				break
   159  			}
   160  		}
   161  		// Receive a type id.
   162  		id := typeId(dec.nextInt())
   163  		if id >= 0 {
   164  			// Value follows.
   165  			return id
   166  		}
   167  		// Type definition for (-id) follows.
   168  		dec.recvType(-id)
   169  		if dec.err != nil {
   170  			break
   171  		}
   172  		// When decoding an interface, after a type there may be a
   173  		// DelimitedValue still in the buffer. Skip its count.
   174  		// (Alternatively, the buffer is empty and the byte count
   175  		// will be absorbed by recvMessage.)
   176  		if dec.buf.Len() > 0 {
   177  			if !isInterface {
   178  				dec.err = errors.New("extra data in buffer")
   179  				break
   180  			}
   181  			dec.nextUint()
   182  		}
   183  		firstMessage = false
   184  	}
   185  	return -1
   186  }
   187  
   188  // Decode reads the next value from the input stream and stores
   189  // it in the data represented by the empty interface value.
   190  // If e is nil, the value will be discarded. Otherwise,
   191  // the value underlying e must be a pointer to the
   192  // correct type for the next data item received.
   193  // If the input is at EOF, Decode returns [io.EOF] and
   194  // does not modify e.
   195  func (dec *Decoder) Decode(e any) error {
   196  	if e == nil {
   197  		return dec.DecodeValue(reflect.Value{})
   198  	}
   199  	value := reflect.ValueOf(e)
   200  	// If e represents a value as opposed to a pointer, the answer won't
   201  	// get back to the caller. Make sure it's a pointer.
   202  	if value.Type().Kind() != reflect.Pointer {
   203  		dec.err = errors.New("gob: attempt to decode into a non-pointer")
   204  		return dec.err
   205  	}
   206  	return dec.DecodeValue(value)
   207  }
   208  
   209  // DecodeValue reads the next value from the input stream.
   210  // If v is the zero reflect.Value (v.Kind() == Invalid), DecodeValue discards the value.
   211  // Otherwise, it stores the value into v. In that case, v must represent
   212  // a non-nil pointer to data or be an assignable reflect.Value (v.CanSet())
   213  // If the input is at EOF, DecodeValue returns [io.EOF] and
   214  // does not modify v.
   215  func (dec *Decoder) DecodeValue(v reflect.Value) error {
   216  	if v.IsValid() {
   217  		if v.Kind() == reflect.Pointer && !v.IsNil() {
   218  			// That's okay, we'll store through the pointer.
   219  		} else if !v.CanSet() {
   220  			return errors.New("gob: DecodeValue of unassignable value")
   221  		}
   222  	}
   223  	// Make sure we're single-threaded through here.
   224  	dec.mutex.Lock()
   225  	defer dec.mutex.Unlock()
   226  
   227  	dec.buf.Reset() // In case data lingers from previous invocation.
   228  	dec.err = nil
   229  	id := dec.decodeTypeSequence(false)
   230  	if dec.err == nil {
   231  		dec.decodeValue(id, v)
   232  	}
   233  	return dec.err
   234  }
   235  
   236  // If debug.go is compiled into the program, debugFunc prints a human-readable
   237  // representation of the gob data read from r by calling that file's Debug function.
   238  // Otherwise it is nil.
   239  var debugFunc func(io.Reader)
   240
View as plain text