ext_rd.c

Go to the documentation of this file.

/****************************************************************************
*                                                                           *
*                       Certificate Attribute Read Routines                 *
*                        Copyright Peter Gutmann 1996-2008                  *
*                                                                           *
****************************************************************************/

#if defined( INC_ALL )
  #include "cert.h"
  #include "certattr.h"
  #include "asn1.h"
  #include "asn1_ext.h"
#else
  #include "cert/cert.h"
  #include "cert/certattr.h"
  #include "enc_dec/asn1.h"
  #include "enc_dec/asn1_ext.h"
#endif /* Compiler-specific includes */

/* Define the following to print a trace of the certificate fields being 
   parsed, useful for debugging broken certificates */

#if !defined( NDEBUG ) && 0
  #if defined( _MSC_VER )
    #pragma warning( disable: 4127 )    /* 'stackPos' may be hardcoded */
  #endif /* VC++ */
  #define TRACE_FIELDTYPE( attributeInfoPtr, stackPos ) \
          { \
          int i; \
          \
          DEBUG_PRINT(( "%4d:", stell( stream ) )); \
          for( i = 0; i < stackPos; i++ ) \
              DEBUG_PRINT(( "  " )); \
          if( ( attributeInfoPtr ) != NULL && \
              ( attributeInfoPtr )->description != NULL ) \
              { \
              DEBUG_PRINT(( ( attributeInfoPtr )->description )); \
              DEBUG_PRINT(( "\n" )); \
              } \
          else \
              { \
              DEBUG_PRINT(( "<Unknown field>\n" )); \
              } \
          }
  #define TRACE_DEBUG( message ) \
          DEBUG_PRINT( message ); \
          DEBUG_PRINT(( "\n" ));
#else
  #define TRACE_FIELDTYPE( attributeInfoPtr, stackPos )
  #define TRACE_DEBUG( message )
#endif /* NDEBUG */

#ifdef USE_CERTIFICATES

/****************************************************************************
*                                                                           *
*                               Utility Routines                            *
*                                                                           *
****************************************************************************/

/* Get the tag for a field from the attribute field definition */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3 ) ) \
static int getFieldTag( INOUT STREAM *stream, 
                        const ATTRIBUTE_INFO *attributeInfoPtr,
                        int *tag )
    {
    int status, value;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( tag, sizeof( int ) ) );

    /* Clear return value.  This is actually a bit difficult to do because 
       the output can have both positive values (tags) and negative values 
       (field codes), setting the output to -1000 is invalid for both 
       types */
    *tag = -1000;

    /* Check whether the field is tagged */
    status = value = getFieldEncodedTag( attributeInfoPtr );
    if( cryptStatusError( status ) )
        {
        /* If there's no tagging (i.e. the tag is the same as the field 
           type) we'll get an OK_SPECIAL return value, this isn't an 
           error */
        if( status != OK_SPECIAL )
            return( status );
        }
    else
        {
        /* It's a tagged field, return the encoded form */
        *tag = value;

        return( CRYPT_OK );
        }
    ENSURES( status == OK_SPECIAL );

    /* It's a non-tagged field, the tag is the same as the field type */
    value = attributeInfoPtr->fieldType;
    if( value == FIELDTYPE_TEXTSTRING )
        {
        static const int allowedStringTypes[] = {
            BER_STRING_BMP, BER_STRING_IA5, BER_STRING_ISO646, 
            BER_STRING_PRINTABLE, BER_STRING_T61, BER_STRING_UTF8,
            CRYPT_ERROR, CRYPT_ERROR 
            };
        int i;

        /* This is a variable-tag field that can have one of a number of 
           tags.  To handle this we peek ahead into the stream to see if an 
           acceptable tag is present and if not set the value to a non-
           matching tag value */
        status = value = peekTag( stream );
        if( cryptStatusError( status ) )
            return( status );
        for( i = 0; allowedStringTypes[ i ] != value && \
                    i < FAILSAFE_ARRAYSIZE( allowedStringTypes, int ); i++ )
            {
            /* If we've reached the end of the list of allowed types without
               finding a match, change the tag value from what we've found to
               make sure that it results in a non-match when the caller uses
               it */
            if( allowedStringTypes[ i ] == CRYPT_ERROR )
                {
                value++;
                break;
                }
            }
        ENSURES( i < FAILSAFE_ARRAYSIZE( allowedStringTypes, int ) );
        }
    if( value == FIELDTYPE_BLOB_BITSTRING || \
        value == FIELDTYPE_BLOB_SEQUENCE )
        {
        /* This is a typed blob that's read as a blob but still has a type
           for type-checking purposes */
        value = ( value == FIELDTYPE_BLOB_BITSTRING ) ? \
                BER_BITSTRING : BER_SEQUENCE;
        }

    ENSURES( ( ( value == FIELDTYPE_BLOB_ANY || value == FIELDTYPE_DN ) && \
               !( attributeInfoPtr->encodingFlags & FL_OPTIONAL ) ) || \
             ( value > 0 && value <= MAX_TAG ) );
             /* A FIELDTYPE_BLOB_ANY or FIELDTYPE_DN can't be optional 
                fields because with no type information for them available 
                there's no way to check whether we've encountered them or 
                not */
    *tag = value;

    return( CRYPT_OK );
    }

/* Read an explicit tag that wraps the actual item that we're after */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3 ) ) \
static int readExplicitTag( INOUT STREAM *stream, 
                            const ATTRIBUTE_INFO *attributeInfoPtr, 
                            OUT_TAG_Z int *tag )
    {
    int status;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( tag, sizeof( int ) ) );

    REQUIRES( attributeInfoPtr->encodingFlags & FL_EXPLICIT );
    REQUIRES( attributeInfoPtr->fieldEncodedType >= 0 && \
              attributeInfoPtr->fieldEncodedType < MAX_TAG );

    /* Clear return value */
    *tag = 0;

    /* Read the explicit wrapper */
    status = readConstructed( stream, NULL, 
                              attributeInfoPtr->fieldEncodedType );
    if( cryptStatusError( status ) )
        return( status );

    /* We've processed the explicit wrappper, we're now on the actual tag */
    *tag = attributeInfoPtr->fieldType;

    return( CRYPT_OK );
    }

/* Find the end of an item (either primitive or constructed) in the attribute
   table.  Sometimes we may have already entered a constructed object (for
   example when an attribute has a version number so we don't know until we've
   started processing it that we can't do anything with it), if this is the
   case then the depth parameter indicates how many nesting levels we have to 
   undo */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1 ) ) \
static int findItemEnd( const ATTRIBUTE_INFO **attributeInfoPtrPtr,
                        IN_RANGE( 0, 3 ) const int depth )
    {
    const ATTRIBUTE_INFO *attributeInfoPtr;
    int currentDepth = depth, iterationCount;

    assert( isReadPtr( attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO * ) ) );
    assert( isReadPtr( *attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO ) ) );

    REQUIRES( depth >= 0 && depth < 3 );

    /* Skip to the end of the (potentially) constructed item by recording the
       nesting level and continuing until either it reaches zero or we reach
       the end of the item */
    for( attributeInfoPtr = *attributeInfoPtrPtr, iterationCount = 0; 
         !( attributeInfoPtr->typeInfoFlags & FL_ATTR_ATTREND ) && \
            iterationCount < FAILSAFE_ITERATIONS_MED;
         attributeInfoPtr++, iterationCount++ )
        {
        /* If it's a sequence/set, increment the depth; if it's an end-of-
           constructed-item marker, decrement it by the appropriate amount */
        if( attributeInfoPtr->fieldType == BER_SEQUENCE || \
            attributeInfoPtr->fieldType == BER_SET )
            currentDepth++;
        currentDepth -= decodeNestingLevel( attributeInfoPtr->encodingFlags );
        if( currentDepth <= 0 )
            break;
        }
    ENSURES( iterationCount < FAILSAFE_ITERATIONS_MED );

    /* We return the next-to-last entry by stopping when we find the 
       FL_ATTR_ATTREND flag since we're going to move on to the next entry 
       once we return */
    *attributeInfoPtrPtr = attributeInfoPtr;
    return( CRYPT_OK );
    }

/****************************************************************************
*                                                                           *
*                           SET/SEQUENCE Management Routines                *
*                                                                           *
****************************************************************************/

/* When we're processing SETs/SEQUENCEs (generically referred to as a SET
   OF) we need to maintain a stack of state information to handle a nested 
   SET OF.  The following code implements the state stack */

#define SETOF_STATE_STACKSIZE   16

#define SETOF_FLAG_NONE         0x00    /* No flag value */
#define SETOF_FLAG_SUBTYPED     0x01    /* SET ends on a subtyped value */
#define SETOF_FLAG_RESTARTPOINT 0x02    /* SET OF rather than SET */
#define SETOF_FLAG_ISEMPTY      0x04    /* Cleared if SET OF contains at least one entry */

typedef struct {
    /* SET OF state information */
    const ATTRIBUTE_INFO *infoStart;    /* Start of SET OF attribute information */
    int startPos, endPos;   /* Start and end position of SET OF */
    int flags;              /* SET OF flags */

    /* Subtype information */
    CRYPT_ATTRIBUTE_TYPE subtypeParent; /* Parent type if this is subtyped */
    int inheritedFlags;     /* Flags inherited from parent if subtyped */
    } SETOF_STATE_INFO;

typedef struct {
    ARRAY( SETOF_STATE_STACKSIZE, stackPos ) \
    SETOF_STATE_INFO stateInfo[ SETOF_STATE_STACKSIZE + 8 ];
    int stackPos;           /* Current position in stack */
    } SETOF_STACK;

STDC_NONNULL_ARG( ( 1 ) ) \
static void setofStackInit( OUT SETOF_STACK *setofStack )
    {
    assert( isWritePtr( setofStack, sizeof( SETOF_STACK ) ) );

    memset( setofStack, 0, sizeof( SETOF_STACK ) );
    }

CHECK_RETVAL_BOOL STDC_NONNULL_ARG( ( 1 ) ) \
static BOOLEAN setofStackPush( INOUT SETOF_STACK *setofStack )
    {
    assert( isWritePtr( setofStack, sizeof( SETOF_STACK ) ) );

    /* Increment the stack pointer and make sure that we don't overflow */
    if( setofStack->stackPos < 0 || \
        setofStack->stackPos >= SETOF_STATE_STACKSIZE - 1 )
        return( FALSE );
    setofStack->stackPos++;
    ENSURES_B( setofStack->stackPos >= 1 && \
               setofStack->stackPos < SETOF_STATE_STACKSIZE );

    /* Initialise the new entry */
    memset( &setofStack->stateInfo[ setofStack->stackPos ], 0, \
            sizeof( SETOF_STATE_INFO ) );

    return( TRUE );
    }

CHECK_RETVAL_BOOL STDC_NONNULL_ARG( ( 1 ) ) \
static BOOLEAN setofStackPop( INOUT SETOF_STACK *setofStack )
    {
    assert( isWritePtr( setofStack, sizeof( SETOF_STACK ) ) );

    /* Decrement the stack pointer and make sure that we don't underflow */
    if( setofStack->stackPos <= 0 || \
        setofStack->stackPos >= SETOF_STATE_STACKSIZE )
        return( FALSE );
    setofStack->stackPos--;
    ENSURES_B( setofStack->stackPos >= 0 && \
               setofStack->stackPos < SETOF_STATE_STACKSIZE - 1 );

    return( TRUE );
    }

CHECK_RETVAL_PTR STDC_NONNULL_ARG( ( 1 ) ) \
static SETOF_STATE_INFO *setofTOS( const SETOF_STACK *setofStack )
    {
    assert( isReadPtr( setofStack, sizeof( SETOF_STACK ) ) );

    ENSURES_N( setofStack->stackPos >= 0 && \
               setofStack->stackPos < SETOF_STATE_STACKSIZE );

    return( ( SETOF_STATE_INFO * ) \
            &setofStack->stateInfo[ setofStack->stackPos ] );
    }

/* Process the start of a SET/SET OF/SEQUENCE/SEQUENCE OF */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3 ) ) \
static int beginSetof( INOUT STREAM *stream, 
                       INOUT SETOF_STACK *setofStack,
                       const ATTRIBUTE_INFO *attributeInfoPtr )
    {
    SETOF_STATE_INFO *setofInfoPtr;
    CRYPT_ATTRIBUTE_TYPE oldSubtypeParent;
    int oldInheritedFlags, setofLength, status;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( setofStack, sizeof( SETOF_STACK ) ) );
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );

    /* Determine the length and start position of the SET OF items.  If the
       tag is an explicit tag then we don't have to process it since it's
       already been handled by the caller */
    if( attributeInfoPtr->fieldEncodedType >= 0 && \
        !( attributeInfoPtr->encodingFlags & FL_EXPLICIT ) )
        {
        status = readConstructed( stream, &setofLength,
                                  attributeInfoPtr->fieldEncodedType );
        }
    else
        {
        if( attributeInfoPtr->fieldType == BER_SET )
            status = readSet( stream, &setofLength );
        else
            status = readSequence( stream, &setofLength );
        }
    if( cryptStatusError( status ) )
        return( status );

    /* When processing a SET/SEQUENCE with default values for the elements 
       the result may be a zero-length object, in which case we don't take 
       any action */
    if( setofLength <= 0 )
        return( CRYPT_OK );

    /* Remember assorted information such as where the SET/SEQUENCE ends.  
       In addition if this is a SET OF/SEQUENCE OF, remember this as a 
       restart point for when we're parsing the next item in the 
       SET/SEQUENCE OF */
    setofInfoPtr = setofTOS( setofStack );
    ENSURES( setofInfoPtr != NULL );
    oldSubtypeParent = setofInfoPtr->subtypeParent;
    oldInheritedFlags = setofInfoPtr->inheritedFlags;
    if( !setofStackPush( setofStack ) )
        {
        /* Stack overflow, there's a problem with the certificate */
        return( CRYPT_ERROR_OVERFLOW );
        }
    setofInfoPtr = setofTOS( setofStack );
    ENSURES( setofInfoPtr != NULL );
    setofInfoPtr->infoStart = attributeInfoPtr;
    if( attributeInfoPtr->encodingFlags & FL_SETOF )
        setofInfoPtr->flags |= SETOF_FLAG_RESTARTPOINT;
    if( !( attributeInfoPtr->encodingFlags & FL_EMPTYOK ) )
        setofInfoPtr->flags |= SETOF_FLAG_ISEMPTY;
    setofInfoPtr->subtypeParent = oldSubtypeParent;
    setofInfoPtr->inheritedFlags = oldInheritedFlags;
    setofInfoPtr->startPos = stell( stream );
    setofInfoPtr->endPos = setofInfoPtr->startPos + setofLength;

    return( CRYPT_OK );
    }

/* Check whether we've reached the end of a SET/SEQUENCE.  Returns OK_SPECIAL
   if the end has been reached */

CHECK_RETVAL_SPECIAL STDC_NONNULL_ARG( ( 1, 2, 3 ) ) \
static int checkSetofEnd( const STREAM *stream, 
                          INOUT SETOF_STACK *setofStack,
                          const ATTRIBUTE_INFO **attributeInfoPtrPtr )
    {
    const SETOF_STATE_INFO *setofInfoPtr = setofTOS( setofStack );
    const ATTRIBUTE_INFO *oldAttributeInfoPtr = *attributeInfoPtrPtr;
    const ATTRIBUTE_INFO *attributeInfoPtr = *attributeInfoPtrPtr;
    const int currentPos = stell( stream );
    int iterationCount;

    assert( isReadPtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( setofStack, sizeof( SETOF_STACK ) ) );
    assert( isReadPtr( attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO * ) ) );
    assert( isReadPtr( *attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO ) ) );

    REQUIRES( setofInfoPtr != NULL );
    REQUIRES( currentPos > 0 && currentPos < MAX_INTLENGTH );

    /* If we're still within the SET/SEQUENCE, we're done */
    if( setofStack->stackPos <= 0 || currentPos < setofInfoPtr->endPos )
        return( CRYPT_OK );

    /* We've reached the end of one or more layers of SET/SEQUENCE, keep 
       popping SET/SEQUENCE state information until we can continue */
    for( iterationCount = 0;
         setofStack->stackPos > 0 && \
            currentPos >= setofInfoPtr->endPos && \
            iterationCount < SETOF_STATE_STACKSIZE;
         iterationCount++ )
        {
        const int flags = setofInfoPtr->flags;

        /* Pop one level of parse state */
        if( !setofStackPop( setofStack ) )
            {
            /* Stack underflow, there's a problem with the certificate */
            return( CRYPT_ERROR_UNDERFLOW );
            }
        setofInfoPtr = setofTOS( setofStack );
        ENSURES( setofInfoPtr != NULL );
        attributeInfoPtr = setofInfoPtr->infoStart;
        ENSURES( setofInfoPtr->endPos > 0 && \
                 setofInfoPtr->endPos < MAX_INTLENGTH_SHORT );

        /* If it's a pure SET/SEQUENCE (not a SET OF/SEQUENCE OF) and there 
           are no more elements present, go to the end of the SET/SEQUENCE 
           information in the decoding table */
        if( !( flags & SETOF_FLAG_RESTARTPOINT ) && \
            currentPos >= setofInfoPtr->endPos )
            {
            int status;

            status = findItemEnd( &attributeInfoPtr, 0 );
            if( cryptStatusError( status ) )
                return( status );
            }
        }
    ENSURES( iterationCount < SETOF_STATE_STACKSIZE );

    *attributeInfoPtrPtr = attributeInfoPtr;
    return( ( attributeInfoPtr != oldAttributeInfoPtr ) ? \
            OK_SPECIAL : CRYPT_OK );
    }

/****************************************************************************
*                                                                           *
*                       Identified Item Management Routines                 *
*                                                                           *
****************************************************************************/

/* Given a pointer to a set of SEQUENCE { type, value } entries, return a
   pointer to the { value } entry appropriate for the data in the stream.  
   If the entry contains user data in the { value } portion then the 
   returned pointer points to this, if it contains a fixed value or isn't 
   present at all then the returned pointer points to the { type } portion */

CHECK_RETVAL_PTR STDC_NONNULL_ARG( ( 1, 2 ) ) \
static const ATTRIBUTE_INFO *findIdentifiedItem( INOUT STREAM *stream,
                                    const ATTRIBUTE_INFO *attributeInfoPtr )
    {
    BYTE oid[ MAX_OID_SIZE + 8 ];
    int oidLength, sequenceLength, iterationCount, status;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );

    REQUIRES_N( attributeInfoPtr->encodingFlags & FL_IDENTIFIER );

    /* Skip the header and read the OID.  We only check for a sane total
       length in the debug version since this isn't a fatal error */
    readSequence( stream, &sequenceLength );
    status = readEncodedOID( stream, oid, MAX_OID_SIZE, &oidLength, 
                             BER_OBJECT_IDENTIFIER );
    if( cryptStatusError( status ) )
        return( NULL );
    sequenceLength -= oidLength;
    ENSURES_N( sequenceLength >= 0 && sequenceLength < MAX_INTLENGTH );

    /* Walk down the list of entries trying to match it to an allowed value.
       Unfortunately we can't use the attributeInfoSize bounds check limit 
       here because we don't know how far through the attribute table we 
       already are, so we have to use a generic large value */
    for( iterationCount = 0;
         ( attributeInfoPtr->encodingFlags & FL_IDENTIFIER ) && \
            iterationCount < FAILSAFE_ITERATIONS_LARGE;
         iterationCount++ )
        {
        const BYTE *oidPtr;

        /* Skip the SEQUENCE and OID.  The fact that an OID follows the
           entry with the FL_IDENTIFIER field unless it's the finall catch-
           all blob entry has been verified during the startup check */
        attributeInfoPtr++;
        oidPtr = attributeInfoPtr->oid;
        if( !( attributeInfoPtr->encodingFlags & FL_NONENCODING ) )
            attributeInfoPtr++;
        else
            {
            /* If this is a blob field we've hit a don't-care value (usually 
               the last in a series of type-and-value pairs) which ensures 
               that { type }s added after the encoding table was defined 
               don't get processed as errors, skip the field and continue */
            if( attributeInfoPtr->fieldType == FIELDTYPE_BLOB_ANY )
                {
                /* If there's a { value } attached to the type, skip it */
                if( sequenceLength > 0 )
                    {
                    status = sSkip( stream, sequenceLength );
                    if( cryptStatusError( status ) )
                        return( NULL );
                    }
                return( attributeInfoPtr );
                }
            }
        ENSURES_N( oidPtr != NULL );

        /* If the OID matches, return a pointer to the value entry */
        if( oidLength == sizeofOID( oidPtr ) && \
            !memcmp( oidPtr, oid, sizeofOID( oidPtr ) ) )
            {
            /* If this is a fixed field and there's a value attached, skip
               it */
            if( ( attributeInfoPtr->encodingFlags & FL_NONENCODING ) && \
                sequenceLength > 0 )
                {
                status = sSkip( stream, sequenceLength );
                if( cryptStatusError( status ) )
                    return( NULL );
                }

            return( attributeInfoPtr );
            }

        /* The OID doesn't match, skip the { value } entry and continue.  We 
           set the current nesting depth parameter to 1 since we've already
           entered the SEQUENCE above */
        status = findItemEnd( &attributeInfoPtr, 1 );
        if( cryptStatusError( status ) )
            return( NULL );
        attributeInfoPtr++;     /* Move to start of next item */
        }
    ENSURES_N( iterationCount < FAILSAFE_ITERATIONS_LARGE );

    /* We reached the end of the set of entries without matching the OID */
    return( NULL );
    }

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 4, 5, 6, 7 ) ) \
static int processIdentifiedItem( INOUT STREAM *stream, 
                                  INOUT ATTRIBUTE_LIST **attributeListPtrPtr,
                                  IN_FLAGS( ATTR ) const int flags, 
                                  const SETOF_STACK *setofStack,
                                  const ATTRIBUTE_INFO **attributeInfoPtrPtr,
                                  OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                                    CRYPT_ATTRIBUTE_TYPE *errorLocus,
                                  OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                                    CRYPT_ERRTYPE_TYPE *errorType )
    {
    const SETOF_STATE_INFO *setofInfoPtr = setofTOS( setofStack );
    const ATTRIBUTE_INFO *attributeInfoPtr;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( attributeListPtrPtr, sizeof( ATTRIBUTE_LIST * ) ) );
    assert( isReadPtr( setofStack, sizeof( SETOF_STACK ) ) );
    assert( isReadPtr( attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO * ) ) );
    assert( isReadPtr( *attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );

assert( ( flags & ~( ATTR_FLAG_CRITICAL ) ) == 0 );
    REQUIRES( flags >= ATTR_FLAG_NONE && flags <= ATTR_FLAG_MAX );
    REQUIRES( setofInfoPtr != NULL );

    /* Search for the identified item from the start of the set of items.  
       The 0-th value is the SET OF/SEQUENCE OF so we start the search at 
       the next entry which is the first FL_IDENTIFIER */
    ENSURES( setofInfoPtr->infoStart->encodingFlags & FL_SETOF );
    attributeInfoPtr = findIdentifiedItem( stream, 
                                           setofInfoPtr->infoStart + 1 );
    if( attributeInfoPtr == NULL )
        return( CRYPT_ERROR_BADDATA );
    *attributeInfoPtrPtr = attributeInfoPtr;

    /* If it's a subtyped field, tell the caller to restart the decoding
       using the new attribute information.  This is typically used where
       the { type, value } combinations that we're processing are
       { OID, GeneralName }, so the process consists of locating the entry 
       that corresponds to the OID and then continuing the decoding with
       the the subtyped attribute information entry that points to the
       GeneralName decoding table */
    if( attributeInfoPtr->fieldType == FIELDTYPE_SUBTYPED )
        return( OK_SPECIAL );

    /* If it's not a special-case, non-encoding field, we're done */
    if( !( attributeInfoPtr->encodingFlags & FL_NONENCODING ) )
        return( CRYPT_OK );

    /* If the { type, value } pair has a fixed value then the information 
       being conveyed is its presence, not its contents, so we add an 
       attribute corresponding to its ID and continue.  The addition of the 
       attribute is a bit tricky, some of the fixed type-and-value pairs can 
       have multiple entries denoting things like { algorithm, weak key }, 
       { algorithm, average key }, { algorithm, strong key }, however all 
       that we're interested in is the strong key so we ignore the value and 
       only use the type (in his ordo est ordinem non servare).  Since the 
       same type can be present multiple times (with different { value }s) 
       we ignore data duplicate errors and continue.  If we're processing a 
       blob field type then we've ended up at a generic catch-any value and 
       can't do much with it */
    if( attributeInfoPtr->fieldType != FIELDTYPE_BLOB_ANY )
        {
        int status;

        /* Add the field type, discarding warnings about duplicates */
        TRACE_FIELDTYPE( attributeInfoPtr, 0 );
        status = addAttributeField( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                    attributeInfoPtr->fieldID, 
                                    CRYPT_ATTRIBUTE_NONE, CRYPT_UNUSED, 
                                    flags, errorLocus, errorType );
        if( cryptStatusError( status ) && status != CRYPT_ERROR_INITED )
            return( CRYPT_ERROR_BADDATA );
        }

    /* Reset the attribute information position in preparation for 
       processing the next value and tell the caller to continue using the 
       reset attribute information */
    *attributeInfoPtrPtr = setofInfoPtr->infoStart + 1;
    return( OK_SPECIAL );
    }

/* Read a sequence of identifier fields of the form { oid, value OPTIONAL }.
   This is used to read both SEQUENCE OF (via FIELDTYPE_IDENTIFIER) and 
   CHOICE (via FIELDTYPE_CHOICE), with SEQUENCE OF allowing multiple entries 
   and CHOICE allowing only a single entry */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3, 6, 7 ) ) \
static int readIdentifierFields( INOUT STREAM *stream, 
                                 INOUT ATTRIBUTE_LIST **attributeListPtrPtr,
                                 const ATTRIBUTE_INFO **attributeInfoPtrPtr, 
                                 IN_FLAGS( ATTR ) const int flags,
                                 IN_ATTRIBUTE const CRYPT_ATTRIBUTE_TYPE fieldID, 
                                 OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                                    CRYPT_ATTRIBUTE_TYPE *errorLocus,
                                 OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                                    CRYPT_ERRTYPE_TYPE *errorType )
    {
    const ATTRIBUTE_INFO *attributeInfoPtr;
    const BOOLEAN isChoice = ( fieldID != CRYPT_ATTRIBUTE_NONE );
    int count = 0, iterationCount;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( attributeListPtrPtr, sizeof( ATTRIBUTE_LIST * ) ) );
    assert( isReadPtr( attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO * ) ) );
    assert( isReadPtr( *attributeInfoPtrPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );

assert( ( flags == ATTR_FLAG_NONE ) || ( flags == ATTR_FLAG_CRITICAL ) );
    REQUIRES( flags >= ATTR_FLAG_NONE && flags <= ATTR_FLAG_MAX );

    for( iterationCount = 0;
         peekTag( stream ) == BER_OBJECT_IDENTIFIER && \
            iterationCount < FAILSAFE_ITERATIONS_MED;
         iterationCount++ )
        {
        BYTE oid[ MAX_OID_SIZE + 8 ];
        BOOLEAN addField = TRUE;
        int oidLength, innerIterationCount, status;

        /* The fact that the FIELDTYPE_IDENTIFIER field is present and 
           associated with an OID has been verified during the startup 
           check */
        attributeInfoPtr = *attributeInfoPtrPtr;
        ENSURES( attributeInfoPtr->fieldType == FIELDTYPE_IDENTIFIER && \
                 attributeInfoPtr->oid != NULL );

        /* Read the OID and walk down the list of possible OIDs up to the end
           of the group of alternatives trying to match it to an allowed
           value */
        status = readEncodedOID( stream, oid, MAX_OID_SIZE, &oidLength, 
                                 BER_OBJECT_IDENTIFIER );
        if( cryptStatusError( status ) )
            return( status );
        for( innerIterationCount = 0;
             ( oidLength != sizeofOID( attributeInfoPtr->oid ) || \
               memcmp( attributeInfoPtr->oid, oid, oidLength ) ) && \
               innerIterationCount < FAILSAFE_ITERATIONS_MED;
             innerIterationCount++ )
            {
            /* If we've reached the end of the list and the OID wasn't
               matched, exit */
            if( ( attributeInfoPtr->encodingFlags & FL_SEQEND_MASK ) || \
                ( attributeInfoPtr->typeInfoFlags & FL_ATTR_ATTREND ) )
                return( CRYPT_ERROR_BADDATA );

            attributeInfoPtr++;

            /* If this is a blob field we've hit a don't-care value which 
               ensures that { type }s added after the encoding table was 
               defined don't get processed as errors, skip the field and 
               continue */
            if( attributeInfoPtr->fieldType == FIELDTYPE_BLOB_ANY )
                {
                addField = FALSE;
                break;
                }

            /* The fact that the FIELDTYPE_IDENTIFIER field is present and 
               associated with an OID has been verified during the startup 
               check */
            ENSURES( attributeInfoPtr->fieldType == FIELDTYPE_IDENTIFIER && \
                     attributeInfoPtr->oid != NULL );
            }
        ENSURES( innerIterationCount < FAILSAFE_ITERATIONS_MED );
        TRACE_FIELDTYPE( attributeInfoPtr, 0 );
        if( addField )
            {
            /* The OID matches, add this field as an identifier field.  This
               will catch duplicate OIDs since we can't add the same 
               identifier field twice */
            if( isChoice )
                {
                /* If there's a field value present then this is a CHOICE of
                   attributes whose value is the field value so we add it with
                   this value */
                status = addAttributeField( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                            fieldID, CRYPT_ATTRIBUTE_NONE,
                                            attributeInfoPtr->fieldID,  
                                            flags, errorLocus, errorType );
                }
            else
                {
                /* It's a standard field */
                status = addAttributeField( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr,
                                            attributeInfoPtr->fieldID, 
                                            CRYPT_ATTRIBUTE_NONE, 
                                            CRYPT_UNUSED, flags, 
                                            errorLocus, errorType );
                }
            if( cryptStatusError( status ) )
                return( status );
            }
        count++;

        /* If there's more than one OID present in a CHOICE, it's an error */
        if( isChoice && count > 1 )
            {
            *errorLocus = attributeInfoPtr->fieldID,
            *errorType = CRYPT_ERRTYPE_ATTR_PRESENT;
            return( CRYPT_ERROR_BADDATA );
            }
        }
    ENSURES( iterationCount < FAILSAFE_ITERATIONS_MED );

    /* We've processed the non-data field(s), move on to the next field.
       We move to the last valid non-data field rather than the start of the
       field following it since the caller needs to be able to check whether
       there are more fields to follow using the current field's flags.
       Unfortunately we can't use the attributeInfoSize bounds check limit 
       here because we don't know how far through the attribute table we 
       already are, so we have to use a generic value */
    for( attributeInfoPtr = *attributeInfoPtrPtr, iterationCount = 0;
         !( attributeInfoPtr->encodingFlags & FL_SEQEND_MASK ) && \
            !( attributeInfoPtr->typeInfoFlags & FL_ATTR_ATTREND ) && \
            iterationCount < FAILSAFE_ITERATIONS_MED;
         attributeInfoPtr++, iterationCount++ );
    ENSURES( iterationCount < FAILSAFE_ITERATIONS_MED );
    *attributeInfoPtrPtr = attributeInfoPtr;

    return( CRYPT_OK );
    }

/****************************************************************************
*                                                                           *
*                       Attribute/Attribute Field Read Routines             *
*                                                                           *
****************************************************************************/

/* Generic error-handler that sets extended error codes */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2 ) ) \
static int fieldErrorReturn( OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                                CRYPT_ATTRIBUTE_TYPE *errorLocus,
                             OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                                CRYPT_ERRTYPE_TYPE *errorType, 
                             IN_ERROR const int status,
                             IN_ATTRIBUTE const CRYPT_ATTRIBUTE_TYPE fieldID )
    {
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );

    REQUIRES( cryptStatusError( status ) );
    REQUIRES( ( fieldID == CRYPT_ATTRIBUTE_NONE ) || \
              ( fieldID > CRYPT_CERTINFO_FIRST && \
                fieldID < CRYPT_CERTINFO_LAST ) );

    /* Since some fields are internal-use only (e.g. meaningless blob data,
       version numbers, and other paraphernalia) we only set the locus if
       it has a meaningful value */
    *errorLocus = ( fieldID > CRYPT_CERTINFO_FIRST && \
                    fieldID < CRYPT_CERTINFO_LAST ) ? \
                  fieldID : CRYPT_ATTRIBUTE_NONE;
    *errorType = CRYPT_ERRTYPE_ATTR_VALUE;

    return( status );
    }

/* Switch from the main encoding table to a subtype encoding table */

CHECK_RETVAL_PTR STDC_NONNULL_ARG( ( 1, 2 ) ) \
static ATTRIBUTE_INFO *switchToSubtype( const ATTRIBUTE_INFO *attributeInfoPtr,
                                        INOUT SETOF_STATE_INFO *setofInfoPtr )
    {
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isReadPtr( attributeInfoPtr->extraData, 
                       sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( setofInfoPtr, sizeof( SETOF_STATE_INFO ) ) );

    /* This has already been verified during the startup check */
    REQUIRES_N( attributeInfoPtr->extraData != NULL );

    /* Record the subtype parent information */
    setofInfoPtr->subtypeParent = attributeInfoPtr->fieldID;
    setofInfoPtr->inheritedFlags = \
                    ( attributeInfoPtr->encodingFlags & FL_MULTIVALUED ) ? \
                      ATTR_FLAG_MULTIVALUED : ATTR_FLAG_NONE;

    /* If the subtype is being used to process a list of { ... OPTIONAL, 
       ... OPTIONAL } and at least one entry must be present, remember 
       that we haven't seen any entries yet */
    if( !( attributeInfoPtr->encodingFlags & FL_EMPTYOK ) )
        setofInfoPtr->flags |= SETOF_FLAG_ISEMPTY;

    /* If the subtype ends once the current SET/SEQUENCE ends, remember this 
       so that we return to the main type when appropriate */
    if( ( attributeInfoPtr->encodingFlags & FL_SEQEND_MASK ) || \
        ( attributeInfoPtr->typeInfoFlags & FL_ATTR_ATTREND ) )
        setofInfoPtr->flags |= SETOF_FLAG_SUBTYPED;

    /* Switch to the subtype encoding table */
    return( ( ATTRIBUTE_INFO * ) attributeInfoPtr->extraData );
    }

/* Read the contents of an attribute field.  This uses the readXXXData() 
   variants of the read functions because the field that we're reading may 
   be tagged so we process the tag at a higher level and only read the 
   contents here */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3, 6, 7 ) ) \
static int readAttributeField( INOUT STREAM *stream, 
                               INOUT ATTRIBUTE_LIST **attributeListPtrPtr,
                               const ATTRIBUTE_INFO *attributeInfoPtr,
                               IN_ATTRIBUTE_OPT \
                                    const CRYPT_ATTRIBUTE_TYPE subtypeParent, 
                               IN_FLAGS( ATTR ) const int flags, 
                               OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                                    CRYPT_ATTRIBUTE_TYPE *errorLocus,
                               OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                                    CRYPT_ERRTYPE_TYPE *errorType )
    {
    CRYPT_ATTRIBUTE_TYPE fieldID, subFieldID;
    const int fieldType = attributeInfoPtr->fieldType;
    int length, status = CRYPT_OK;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( attributeListPtrPtr, sizeof( ATTRIBUTE_LIST * ) ) );
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );

    REQUIRES( ( subtypeParent == CRYPT_ATTRIBUTE_NONE ) || \
              ( subtypeParent > CRYPT_CERTINFO_FIRST && \
                subtypeParent < CRYPT_CERTINFO_LAST ) );
assert( ( flags & ~( ATTR_FLAG_NONE | ATTR_FLAG_CRITICAL | ATTR_FLAG_MULTIVALUED ) ) == 0 );
    REQUIRES( flags >= ATTR_FLAG_NONE && flags <= ATTR_FLAG_MAX );

    /* Set up the field identifiers depending on whether it's a normal field
       or a subfield of a parent field */
    if( subtypeParent == CRYPT_ATTRIBUTE_NONE )
        {
        fieldID = attributeInfoPtr->fieldID;
        subFieldID = CRYPT_ATTRIBUTE_NONE;
        }
    else
        {
        fieldID = subtypeParent;
        subFieldID = attributeInfoPtr->fieldID;
        }

    /* Read the field as appropriate */
    switch( fieldType )
        {
        case BER_INTEGER:
        case BER_ENUMERATED:
        case BER_BITSTRING:
        case BER_BOOLEAN:
        case BER_NULL:
            {
            int value;

            /* Read the data as appropriate */
            switch( fieldType )
                {
                case BER_BITSTRING:
                    status = readBitStringData( stream, &value );
                    break;

                case BER_BOOLEAN:
                    {
                    BOOLEAN boolean;

                    status = readBooleanData( stream, &boolean );
                    value = boolean;
                    break;
                    }

                case BER_ENUMERATED:
                    status = readEnumeratedData( stream, &value );
                    break;

                case BER_INTEGER:
                    {
                    long longValue;

                    status = readShortIntegerData( stream, &longValue );
                    value = ( int ) longValue;
                    break;
                    }

                case BER_NULL:
                    /* NULL values have no associated data so we explicitly 
                       set the value to CRYPT_UNUSED to ensure that this is 
                       returned on any attempt to read it */
                    value = CRYPT_UNUSED;
                    break;

                default:
                    retIntError();
                }
            if( cryptStatusError( status ) )
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );

            /* Add the data for this attribute field */
            return( addAttributeField( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                       fieldID, subFieldID, value, flags, 
                                       errorLocus, errorType ) );
            }

        case BER_TIME_GENERALIZED:
        case BER_TIME_UTC:
            {
            time_t timeVal;

            if( fieldType == BER_TIME_GENERALIZED )
                status = readGeneralizedTimeData( stream, &timeVal );
            else
                status = readUTCTimeData( stream, &timeVal );
            if( cryptStatusError( status ) )
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );

            /* Add the data for this attribute field */
            return( addAttributeFieldString( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                             fieldID, subFieldID, &timeVal, 
                                             sizeof( time_t ), flags, 
                                             errorLocus, errorType ) );
            }

        case BER_STRING_BMP:
        case BER_STRING_IA5:
        case BER_STRING_ISO646:
        case BER_STRING_NUMERIC:
        case BER_STRING_PRINTABLE:
        case BER_STRING_T61:
        case BER_STRING_UTF8:
        case BER_OCTETSTRING:
        case FIELDTYPE_BLOB_ANY:
        case FIELDTYPE_BLOB_BITSTRING:
        case FIELDTYPE_BLOB_SEQUENCE:
        case FIELDTYPE_TEXTSTRING:
            {
            /* If it's a string type or a blob read it in as a blob (the 
               only difference being that for a true blob we read the tag + 
               length as well) */
            BYTE buffer[ 256 + 8 ];

            /* Read in the string to a maximum length of 256 bytes */
            if( isBlobField( fieldType ) )
                {
                int tag;
                
                /* Reading in blob fields is somewhat difficult since these
                   are typically used to read SET/SEQUENCE OF kitchensink
                   values and so won't have a consistent tag that we can pass
                   to readRawObject().  To get around this we peek ahead into
                   the stream to get the tag and then pass that down to 
                   readRawObject().  Note that this requires that the blob 
                   have an internal structure (with its own { tag, length }
                   data) since the caller has already stripped off the tag
                   and length */
                status = tag = peekTag( stream );
                if( cryptStatusError( status ) )
                    {
                    return( fieldErrorReturn( errorLocus, errorType, status,
                                              attributeInfoPtr->fieldID ) );
                    }
                status = readRawObject( stream, buffer, 256, &length, tag );
                }
            else
                {
                status = readOctetStringData( stream, buffer, &length, \
                                              1, 256 );
                }
            if( cryptStatusError( status ) )
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );

            /* There are enough broken certificates out there with 
               enormously long disclaimers in the certificate policy 
               explicit text field that we have to specifically check for 
               them here and truncate the text at a valid length in order to 
               get it past the extension validity checking code */
            if( fieldID == CRYPT_CERTINFO_CERTPOLICY_EXPLICITTEXT && \
                length > 200 )
                length = 200;

            /* Add the data for this attribute field, setting the payload-
               blob flag to disable type-checking of the payload data so 
               that users can cram any old rubbish into the strings */
            return( addAttributeFieldString( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                             fieldID, subFieldID, buffer, length, 
                                             flags | ATTR_FLAG_BLOB_PAYLOAD,
                                             errorLocus, errorType ) );
            }

        case BER_OBJECT_IDENTIFIER:
            {
            BYTE oid[ MAX_OID_SIZE + 8 ];

            /* If it's an OID we need to reassemble the entire OID since 
               this is the form expected by addAttributeFieldString() */
            oid[ 0 ] = BER_OBJECT_IDENTIFIER;   /* Add skipped tag */
            status = readEncodedOID( stream, oid  + 1, MAX_OID_SIZE - 1, 
                                     &length, NO_TAG );
            if( cryptStatusError( status ) )
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );
            return( addAttributeFieldString( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                             fieldID, subFieldID, oid, length + 1, 
                                             flags, errorLocus, errorType ) );
            }

        case FIELDTYPE_DN:
            {
            DN_PTR *dnPtr;

            /* Read the DN */
            status = readDN( stream, &dnPtr );
            if( cryptStatusError( status ) )
                {
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );
                }

            /* Some buggy certificates can include zero-length DNs, which we 
               skip */
            if( dnPtr == NULL )
                return( CRYPT_OK );

            /* We're being asked to instantiate the field containing the DN,
               create the attribute field and fill in the DN value.  Since 
               the value that we're passing in is actually a DN_PTR rather 
               than a standard string value we set set the size field to the 
               pseudo-length of a DN_PTR_STORAGE value to keep the static/
               runtime code checks happy */
            status = addAttributeFieldString( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr, 
                                              fieldID, subFieldID, dnPtr, 
                                              sizeof( DN_PTR_STORAGE ), 
                                              flags, errorLocus, errorType );
            if( cryptStatusError( status ) )
                deleteDN( &dnPtr );
            return( status );
            }
        }

    retIntError();
    }

/* Read an attribute */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3, 6, 7 ) ) \
static int readAttribute( INOUT STREAM *stream, 
                          INOUT ATTRIBUTE_LIST **attributeListPtrPtr,
                          const ATTRIBUTE_INFO *attributeInfoPtr, 
                          IN_LENGTH_Z const int attributeLength, 
                          const BOOLEAN criticalFlag, 
                          OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                            CRYPT_ATTRIBUTE_TYPE *errorLocus,
                          OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                            CRYPT_ERRTYPE_TYPE *errorType )
    {
    SETOF_STACK setofStack;
    SETOF_STATE_INFO *setofInfoPtr;
    const int maxAttributeFields = \
        ( attributeInfoPtr->fieldID == CRYPT_CERTINFO_AUTONOMOUSSYSIDS || \
          attributeInfoPtr->fieldID == CRYPT_CERTINFO_IPADDRESSBLOCKS ) ? \
        5 + ( attributeLength / 3 ) : min( 5 + ( attributeLength / 3 ), 256 );
        /* The RPKI extensions can contain vast lists of 3-4 byte entries 
           that exceed the normal sanity-check limit */
    const int endPos = stell( stream ) + attributeLength;
    BOOLEAN attributeContinues = TRUE;
    int flags = criticalFlag ? ATTR_FLAG_CRITICAL : ATTR_FLAG_NONE;
    int attributeFieldsProcessed = 0, status = CRYPT_OK;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( attributeListPtrPtr, sizeof( ATTRIBUTE_LIST * ) ) );
    assert( isReadPtr( attributeInfoPtr, sizeof( ATTRIBUTE_INFO ) ) );
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );
    
    REQUIRES( attributeLength >= 0 && attributeLength < MAX_INTLENGTH );
    REQUIRES( endPos > 0 && endPos < MAX_INTLENGTH );

    /* Initialise the SET OF state stack */
    setofStackInit( &setofStack );
    setofInfoPtr = setofTOS( &setofStack );
    ENSURES( setofInfoPtr != NULL );

    /* Process each field in the attribute.  This is a simple FSM driven by
       the encoding table and the data that we encounter.  The various 
       states and associated actions are indicated by the comment tags */
    do
        {
        int tag;

        /* Inside a SET/SET OF/SEQUENCE/SEQUENCE OF: Check for the end of the
           item/collection of items.  This must be the first action taken
           since reaching the end of a SET/SEQUENCE pre-empts all other
           parsing actions */
        if( setofInfoPtr->endPos > 0 )
            {
            /* If we've reached the end of the collection of items, exit */
            status = checkSetofEnd( stream, &setofStack, &attributeInfoPtr );
            if( cryptStatusError( status ) && status != OK_SPECIAL )
                return( status );
            setofInfoPtr = setofTOS( &setofStack );
            ENSURES( setofInfoPtr != NULL );
            if( status == OK_SPECIAL )
                goto continueDecoding;

            /* If we're looking for a new item find the table entry that it
               corresponds to.  This takes a pointer to the start of a set of
               SEQUENCE { type, value } entries and returns a pointer to the
               appropriate value entry.

               The test for the start of a new item is a bit complex since we
               could be at the end of the previous item (i.e. on the next item
               flagged as an identifier) or at the end of the attribute (i.e.
               on the start of the next attribute) */
            if( ( attributeInfoPtr->typeInfoFlags & FL_ATTR_ATTRSTART ) || \
                ( attributeInfoPtr->encodingFlags & FL_IDENTIFIER ) )
                {
                status = processIdentifiedItem( stream, attributeListPtrPtr,
                                                flags, &setofStack, 
                                                &attributeInfoPtr, 
                                                errorLocus, errorType );
                if( cryptStatusError( status ) )
                    {
                    if( status == OK_SPECIAL )
                        {
                        /* We've switched to a new encoding table, continue
                           from there */
                        status = CRYPT_OK;
                        continue;
                        }
                    return( fieldErrorReturn( errorLocus, errorType, 
                                              CRYPT_ERROR_BADDATA,
                                              attributeInfoPtr->fieldID ) );
                    }
                }
            }

        /* Subtyped field: Switch to the new encoding table */
        if( attributeInfoPtr->fieldType == FIELDTYPE_SUBTYPED )
            {
            attributeInfoPtr = switchToSubtype( attributeInfoPtr, 
                                                setofInfoPtr );
            ENSURES( attributeInfoPtr != NULL );
            }

        /* CHOICE (of object identifiers): Read a single OID from a
           selection defined in a subtable.
           Identifier field: Read a sequence of one or more { oid, value }
           fields and continue */
        if( attributeInfoPtr->fieldType == FIELDTYPE_CHOICE || \
            attributeInfoPtr->fieldType == FIELDTYPE_IDENTIFIER )
            {
            if( attributeInfoPtr->fieldType == FIELDTYPE_CHOICE )
                {
                const ATTRIBUTE_INFO *extraDataPtr = \
                                            attributeInfoPtr->extraData;
                        /* Needed because ->extraData is read-only */

                /* This has already been verified during the startup check */
                ENSURES( extraDataPtr != NULL );
                
                status = readIdentifierFields( stream, attributeListPtrPtr,
                            &extraDataPtr, flags, attributeInfoPtr->fieldID,
                            errorLocus, errorType );
                }
            else
                {
                status = readIdentifierFields( stream, attributeListPtrPtr,
                            &attributeInfoPtr, flags, CRYPT_ATTRIBUTE_NONE, 
                            errorLocus, errorType );
                }
            if( cryptStatusError( status ) )
                {
                return( fieldErrorReturn( errorLocus, errorType,
                                          CRYPT_ERROR_BADDATA,
                                          attributeInfoPtr->fieldID ) );
                }
            if( setofInfoPtr->endPos > 0 )
                {
                /* Remember that we've seen an entry in the SET/SEQUENCE */
                setofInfoPtr->flags &= ~SETOF_FLAG_ISEMPTY;
                }
            goto continueDecoding;
            }

        /* Non-encoding field: Skip it and continue */
        if( attributeInfoPtr->encodingFlags & FL_NONENCODING )
            {
            /* Read the data and continue.  We don't check its value or set
               specific error information for the reasons given under the SET 
               OF handling code above (value check) and optional field code 
               below (error locus set) */
            TRACE_FIELDTYPE( attributeInfoPtr, setofStack.stackPos );
            status = readUniversal( stream );
            if( cryptStatusError( status ) )
                return( status );
            if( setofInfoPtr->endPos > 0 )
                {
                /* Remember that we've seen an entry in the SET/SEQUENCE */
                setofInfoPtr->flags &= ~SETOF_FLAG_ISEMPTY;
                }
            goto continueDecoding;
            }

        /* Get the tag for the field */
        status = getFieldTag( stream, attributeInfoPtr, &tag );
        if( cryptStatusError( status ) )
            return( status );

        /* Optional field: Check whether it's present and if it isn't, move
           on to the next field */
        if( ( attributeInfoPtr->encodingFlags & FL_OPTIONAL ) && \
            peekTag( stream ) != tag )
            {
            /* If it's a field with a default value, add that value.  This
               isn't needed for cryptlib's own use since it knows the default
               values for fields but can cause confusion for the caller if 
               all fields in an attribute have default values because the
               attribute will appear to disappear when it's read in as no
               fields are ever added */
            if( attributeInfoPtr->encodingFlags & FL_DEFAULT )
                {
                CRYPT_ATTRIBUTE_TYPE dummy1;
                CRYPT_ERRTYPE_TYPE dummy2;

                status = addAttributeField( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr,
                                            attributeInfoPtr->fieldID, 
                                            CRYPT_ATTRIBUTE_NONE,
                                            attributeInfoPtr->defaultValue, 
                                            flags, &dummy1, &dummy2 );
                if( cryptStatusError( status ) )
                    {
                    /* This is a field contributed from internal data so we
                       don't try and get an error locus or value for it
                       since this would only confuse the caller */
                    return( CRYPT_ERROR_BADDATA );
                    }
                }

            /* Skip to the end of the item and continue */
            status = findItemEnd( &attributeInfoPtr, 0 );
            if( cryptStatusError( status ) )
                return( status );
            
            /* Since this was a (non-present) optional attribute it 
               shouldn't be counted in the total when we continue decoding,
               so we adjust the fields-processed value to account for this */
            if( attributeFieldsProcessed > 0 )
                attributeFieldsProcessed--;

            goto continueDecoding;
            }

        /* Print a trace of what we're processing.  Everything before this
           point does its own special-case tracing if required so we don't
           trace before we get here to avoid displaying duplicate/
           misleading information */
        TRACE_FIELDTYPE( attributeInfoPtr, setofStack.stackPos );

        /* Explicitly tagged field: Read the explicit wrapper and make sure
           that it matches what we're expecting */
        if( attributeInfoPtr->encodingFlags & FL_EXPLICIT )
            {
            REQUIRES( tag == MAKE_CTAG( attributeInfoPtr->fieldEncodedType ) );
                      /* Always constructed */

            status = readExplicitTag( stream, attributeInfoPtr, &tag );
            if( cryptStatusError( status ) )
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );
            }

        /* Blob field or DN: We don't try and interpret blobs in any way, 
           and DNs are a composite structure read as a complete unit by the 
           lower-level code */
        if( isBlobField( attributeInfoPtr->fieldType ) || \
            attributeInfoPtr->fieldType == FIELDTYPE_DN )
            {
            status = readAttributeField( stream, attributeListPtrPtr,
                                         attributeInfoPtr,
                                         setofInfoPtr->subtypeParent,
                                         flags | setofInfoPtr->inheritedFlags,
                                         errorLocus, errorType );
            if( cryptStatusError( status ) )
                {
                /* Adding complex attributes such as DNs can return detailed
                   error codes that report the exact parameter that was wrong,
                   we don't need this much detail so we convert a parameter
                   error into a more general bad data status */
                return( fieldErrorReturn( errorLocus, errorType,
                                          cryptArgError( status ) ? \
                                            CRYPT_ERROR_BADDATA : status,
                                          attributeInfoPtr->fieldID ) );
                }
            if( setofInfoPtr->endPos > 0 )
                {
                /* Remember that we've seen an entry in the SET/SEQUENCE */
                setofInfoPtr->flags &= ~SETOF_FLAG_ISEMPTY;
                }
            goto continueDecoding;
            }

        /* Standard field: Read the tag for the field and make sure that it
           matches what we're expecting */
        if( peekTag( stream ) != tag )
            {
            return( fieldErrorReturn( errorLocus, errorType,
                                      CRYPT_ERROR_BADDATA,
                                      attributeInfoPtr->fieldID ) );
            }
        if( setofInfoPtr->endPos > 0 )
            {
            /* Remember that we've seen an entry in the SET/SEQUENCE */
            setofInfoPtr->flags &= ~SETOF_FLAG_ISEMPTY;
            }

        /* SET/SET OF/SEQUENCE/SEQUENCE OF start: Record its end position,
           stack the current processing state, and continue */
        if( attributeInfoPtr->fieldType == BER_SEQUENCE || \
            attributeInfoPtr->fieldType == BER_SET )
            {
            status = beginSetof( stream, &setofStack, attributeInfoPtr );
            if( cryptStatusError( status ) )
                return( fieldErrorReturn( errorLocus, errorType, status,
                                          attributeInfoPtr->fieldID ) );
            setofInfoPtr = setofTOS( &setofStack );
            ENSURES( setofInfoPtr != NULL );
            goto continueDecoding;
            }
        ENSURES( !( attributeInfoPtr->encodingFlags & FL_SETOF ) );

        /* We've checked the tag, skip it.  We do this at this level rather
           than in readAttributeField() because it doesn't know about 
           context-specific tagging requirements */
        status = readTag( stream );
        if( cryptStatusError( status ) )
            return( status );

        /* Standard field, read the field data */
        status = readAttributeField( stream, attributeListPtrPtr,
                                     attributeInfoPtr,
                                     setofInfoPtr->subtypeParent,
                                     flags | setofInfoPtr->inheritedFlags,
                                     errorLocus, errorType );
        if( cryptStatusError( status ) )
            {
            /* Adding invalid attribute data can return detailed error codes
               that report the exact parameter that was wrong, we don't
               need this much detail so we convert a parameter error into a
               more general bad data status */
            return( cryptArgError( status ) ? CRYPT_ERROR_BADDATA : status );
            }

        /* Move on to the next field */
continueDecoding:
        attributeContinues = \
                    ( attributeInfoPtr->typeInfoFlags & FL_ATTR_ATTREND ) ? \
                    FALSE : TRUE;
        attributeInfoPtr++;

        /* If this is the end of the attribute encoding information but we're
           inside a SET OF/SEQUENCE OF and there's more attribute data 
           present, go back to the restart point and try again */
        if( !attributeContinues && setofInfoPtr->endPos > 0 && \
            stell( stream ) < setofInfoPtr->endPos )
            {
            /* If we require at least one entry in the SET OF/SEQUENCE OF 
               but we haven't found one, this is an error */
            if( setofInfoPtr->flags & SETOF_FLAG_ISEMPTY )
                return( CRYPT_ERROR_BADDATA );

            /* If we haven't made any progress in processing the SET OF/
               SEQUENCE OF then further iterations through the loop won't
               make any difference, there's a bug in the decoder */
            ENSURES( stell( stream ) > setofInfoPtr->startPos );

            /* Retry from the restart point */
            attributeInfoPtr = setofInfoPtr->infoStart + 1;
            ENSURES( ( setofInfoPtr->flags & SETOF_FLAG_RESTARTPOINT ) || \
                     attributeInfoPtr->fieldType == FIELDTYPE_IDENTIFIER );
            attributeContinues = TRUE;
            }
        }
    while( ( attributeContinues || setofStack.stackPos > 1 ) && \
           stell( stream ) < endPos && \
           attributeFieldsProcessed++ < maxAttributeFields );

    /* If we got stuck in a loop trying to decode an attribute, complain and 
       exit.  At this point we could have encountered either a certificate-
       parsing error or a CRYPT_ERROR_INTERNAL internal error, since we 
       can't tell without human intervention we treat it as a certificate 
       error rather than throwing a retIntError() exception.  This is 
       particularly important for the CRYPT_CERTINFO_IPADDRESSBLOCKS and
       CRYPT_CERTINFO_AUTONOMOUSSYSIDS extensions, which are an end-run 
       around the nonexistence of attribute certificates but turn standard
       certificates into arbitrary-length capability lists that exceed 
       normal sanity-check limits */
    if( attributeFieldsProcessed >= maxAttributeFields )
        {
        DEBUG_DIAG(( "Processed more than %d fields in attribute, decoder "
                     "may be stuck", maxAttributeFields ));
        assert( DEBUG_WARN );
        return( CRYPT_ERROR_BADDATA );
        }

    /* Handle the special case of (a) the encoded data ending but fields with
       default values being present or (b) the encoded data continuing but no 
       more decoding information being present */
    if( attributeContinues )
        {
        /* If there are default fields to follow, add the default value, see
           the comment on the handling of default fields above for more 
           details.  For now we only add the first field since the only 
           attributes where this case can occur have a single default value 
           as the next possible entry, burrowing down further causes 
           complications due to default values present in optional sequences.  
           As usual we don't set any specific error information for the 
           default fields */
        if( attributeInfoPtr->encodingFlags & FL_DEFAULT )
            {
            CRYPT_ATTRIBUTE_TYPE dummy1;
            CRYPT_ERRTYPE_TYPE dummy2;

            status = addAttributeField( ( ATTRIBUTE_PTR ** ) attributeListPtrPtr,
                                        attributeInfoPtr->fieldID, 
                                        CRYPT_ATTRIBUTE_NONE,
                                        attributeInfoPtr->defaultValue, 
                                        flags, &dummy1, &dummy2 );
            if( cryptStatusError( status ) )
                return( status );
            }
        }
    else
        {
        int iterationCount;
        
        /* Some attributes have a SEQUENCE OF fields of no great use (e.g.
           Microsoft's extensive crlDistributionPoints lists providing
           redundant pointers to the same inaccessible site-internal
           servers, although these are already handled above), if there's
           any extraneous data left we just skip it */
        for( iterationCount = 0;
             stell( stream ) < endPos && cryptStatusOK( status ) && \
                iterationCount < FAILSAFE_ITERATIONS_LARGE;
             iterationCount++ )
            {
            DEBUG_DIAG(( "Skipping extraneous data at end of extension" ));
            assert( DEBUG_WARN );
            status = readUniversal( stream );
            }
        ENSURES( iterationCount < FAILSAFE_ITERATIONS_LARGE );
        }

    return( status );
    }

/****************************************************************************
*                                                                           *
*                       Attribute Collection Read Routines                  *
*                                                                           *
****************************************************************************/

/* Read the certificate object-specific wrapper for a set of attributes */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2 ) ) \
static int readAttributeWrapper( INOUT STREAM *stream, 
                                 OUT_LENGTH_Z int *lengthPtr, 
                                 IN_ENUM_OPT( CRYPT_CERTTYPE ) \
                                    const CRYPT_CERTTYPE_TYPE type,
                                 IN_LENGTH const int attributeLength )
    {
    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( lengthPtr, sizeof( int ) ) );

    REQUIRES( type >= CRYPT_CERTTYPE_NONE && type < CRYPT_CERTTYPE_LAST );
              /* Single CRL entries have the special-case type 
                 CRYPT_CERTTYPE_NONE */
    REQUIRES( attributeLength >= 0 && attributeLength < MAX_INTLENGTH );

    /* Clear return value */
    *lengthPtr = 0;

    /* Read the appropriate extensions tag for the certificate object and
       determine how far we can read.  CRLs and OCSP requests/responses have
       two extension types that have different tagging, per-entry extensions
       and entire-CRL/request extensions.  To differentiate between the two
       we read per-entry extensions with a type of CRYPT_CERTTYPE_NONE */
    switch( type )
        {
        case CRYPT_CERTTYPE_CERTIFICATE:
            readConstructed( stream, NULL, CTAG_CE_EXTENSIONS );
            return( readSequence( stream, lengthPtr ) );

        case CRYPT_CERTTYPE_CRL:
            readConstructed( stream, NULL, CTAG_CL_EXTENSIONS );
            return( readSequence( stream, lengthPtr ) );

        case CRYPT_CERTTYPE_ATTRIBUTE_CERT:
        case CRYPT_CERTTYPE_PKIUSER:
        case CRYPT_CERTTYPE_NONE:
            /* Any outer wrapper for per-entry CRL/OCSP extensions has
               already been read by the caller so there's only the inner
               SEQUENCE left to read */
            return( readSequence( stream, lengthPtr ) );

        case CRYPT_CERTTYPE_CMS_ATTRIBUTES:
            return( readConstructed( stream, lengthPtr,
                                     CTAG_SI_AUTHENTICATEDATTRIBUTES ) );

        case CRYPT_CERTTYPE_REQUEST_CERT:
        case CRYPT_CERTTYPE_REQUEST_REVOCATION:
            /* CRMF/CMP attributes don't contain any wrapper so there's
               nothing to read */
            *lengthPtr = attributeLength;
            return( CRYPT_OK );

        case CRYPT_CERTTYPE_RTCS_REQUEST:
            return( readSet( stream, lengthPtr ) );

        case CRYPT_CERTTYPE_RTCS_RESPONSE:
            return( readConstructed( stream, lengthPtr, CTAG_RP_EXTENSIONS ) );

        case CRYPT_CERTTYPE_OCSP_REQUEST:
            readConstructed( stream, NULL, CTAG_OR_EXTENSIONS );
            return( readSequence( stream, lengthPtr ) );

        case CRYPT_CERTTYPE_OCSP_RESPONSE:
            readConstructed( stream, NULL, CTAG_OP_EXTENSIONS );
            return( readSequence( stream, lengthPtr ) );
        }

    retIntError();
    }

#ifdef USE_CERTREQ 

/* Read a PKCS #10 certificate request wrapper for a set of attributes.  
   This isn't as simple as it should be because there are two approaches to 
   adding extensions to a request, the PKCS #10 approach which puts them all 
   inside a PKCS #9 extensionRequest attribute and the SET approach which
   lists them all individually (CRMF is a separate case handled by 
   readAttributeWrapper() above).  In addition Microsoft invented their own 
   incompatible version of the PKCS #9 extensionRequest which is exactly
   the same as the PKCS #9 one but with a MS OID, however they also invented 
   other values to add containing God knows what sort of data (long Unicode 
   strings describing the Windows module that created it (as if you'd need 
   that to know where it came from), the scripts from "Gilligan's Island", 
   every "Brady Bunch" episode ever made, dust from under somebody's bed 
   from the 1930s, etc).  Because of this, the following code skips over 
   unknown garbage until it finds a valid extension.  Since all attributes 
   may be either skipped or processed at this stage we include provisions 
   for bailing out if we exhaust the available attributes */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 3, 5, 6 ) ) \
static int readCertReqWrapper( INOUT STREAM *stream, 
                               INOUT ATTRIBUTE_LIST **attributeListPtrPtr,
                               OUT_LENGTH_Z int *lengthPtr, 
                               IN_LENGTH const int attributeLength,
                               OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                                    CRYPT_ATTRIBUTE_TYPE *errorLocus,
                               OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                                    CRYPT_ERRTYPE_TYPE *errorType )
    {
    const int endPos = stell( stream ) + attributeLength;
    int attributesProcessed, status;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( attributeListPtrPtr, sizeof( ATTRIBUTE_LIST * ) ) );
    assert( isWritePtr( lengthPtr, sizeof( int ) ) );
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );

    REQUIRES( attributeLength >= 0 && attributeLength < MAX_INTLENGTH );
    REQUIRES( endPos > 0 && endPos < MAX_INTLENGTH );

    /* Clear return value */
    *lengthPtr = 0;
            
    for( attributesProcessed = 0; attributesProcessed < 16; \
         attributesProcessed++ )
        {
#ifdef USE_CERT_OBSOLETE
        const ATTRIBUTE_INFO *attributeInfoPtr;
#endif /* USE_CERT_OBSOLETE */
        BYTE oid[ MAX_OID_SIZE + 8 ];
        int oidLength;

        /* If we've run out of attributes without finding anything useful, 
           exit */
        if( stell( stream ) > endPos - MIN_ATTRIBUTE_SIZE )
            return( CRYPT_OK );

        /* Read the wrapper SEQUENCE and OID */
        readSequence( stream, NULL );
        status = readEncodedOID( stream, oid, MAX_OID_SIZE, &oidLength, 
                                 BER_OBJECT_IDENTIFIER );
        if( cryptStatusError( status ) )
            return( status );

#ifdef USE_CERT_OBSOLETE
        /* Check for a known attribute, which can happen with SET 
           certificate requests.  If it's a known attribute, process it */
        attributeInfoPtr = oidToAttribute( ATTRIBUTE_CERTIFICATE, 
                                           oid, oidLength );
        if( attributeInfoPtr != NULL )
            {
            status = readSet( stream, lengthPtr );
            if( cryptStatusOK( status ) )
                {
                status = readAttribute( stream, attributeListPtrPtr,
                                        attributeInfoPtr, *lengthPtr, FALSE, 
                                        errorLocus, errorType );
                }
            if( cryptStatusError( status ) )
                return( status );
            }
        else
#endif /* USE_CERT_OBSOLETE */
            {
            /* It's not a known attribute, check whether it's a CRMF or MS 
               wrapper attribute */
            if( !memcmp( oid, OID_PKCS9_EXTREQ, oidLength ) || \
                !memcmp( oid, OID_MS_EXTREQ, oidLength ) )
                {
                /* Skip the wrapper to reveal the encapsulated attributes */
                readSet( stream, NULL );
                return( readSequence( stream, lengthPtr ) );
                }

            /* It's unknown MS garbage, skip it */
            status = readUniversal( stream );
            if( cryptStatusError( status ) )
                return( status );
            }
        }

    /* As with the check in readAttribute() above this could be either a 
       certificate-parsing error or a CRYPT_ERROR_INTERNAL internal error, 
       since we can't tell without human intervention we treat it as a 
       certificate error rather than throwing a retIntError() exception */
    DEBUG_DIAG(( "Unknown certificate request wrapper type encountered" ));
    assert( DEBUG_WARN );
    return( CRYPT_ERROR_BADDATA );
    }
#endif /* USE_CERTREQ */

/* Read a set of attributes */

CHECK_RETVAL STDC_NONNULL_ARG( ( 1, 2, 5, 6 ) ) \
int readAttributes( INOUT STREAM *stream, 
                    INOUT ATTRIBUTE_PTR **attributeListPtrPtr,
                    IN_ENUM_OPT( CRYPT_CERTTYPE ) const CRYPT_CERTTYPE_TYPE type, 
                    IN_LENGTH_Z const int attributeLength,
                    OUT_ENUM_OPT( CRYPT_ATTRIBUTE ) \
                        CRYPT_ATTRIBUTE_TYPE *errorLocus,
                    OUT_ENUM_OPT( CRYPT_ERRTYPE ) \
                        CRYPT_ERRTYPE_TYPE *errorType )
    {
    const ATTRIBUTE_TYPE attributeType = ( type == CRYPT_CERTTYPE_CMS_ATTRIBUTES || \
                                           type == CRYPT_CERTTYPE_RTCS_REQUEST || \
                                           type == CRYPT_CERTTYPE_RTCS_RESPONSE ) ? \
                                         ATTRIBUTE_CMS : ATTRIBUTE_CERTIFICATE;
    const BOOLEAN wrapperTagSet = ( attributeType == ATTRIBUTE_CMS ) ? \
                                  TRUE : FALSE;
    int length, endPos, complianceLevel, iterationCount, status;

    assert( isWritePtr( stream, sizeof( STREAM ) ) );
    assert( isWritePtr( attributeListPtrPtr, sizeof( ATTRIBUTE_LIST * ) ) );
    assert( isWritePtr( errorLocus, sizeof( CRYPT_ATTRIBUTE_TYPE ) ) );
    assert( isWritePtr( errorType, sizeof( CRYPT_ERRTYPE_TYPE ) ) );

    REQUIRES( type >= CRYPT_CERTTYPE_NONE && type < CRYPT_CERTTYPE_LAST );
              /* Single CRL entries have the special-case type 
                 CRYPT_CERTTYPE_NONE */
    REQUIRES( ( type == CRYPT_CERTTYPE_CMS_ATTRIBUTES && \
                attributeLength == 0 ) || \
              ( type != CRYPT_CERTTYPE_CMS_ATTRIBUTES && \
                attributeLength >= 0 && attributeLength < MAX_INTLENGTH ) );
              /* CMS attributes are pure attribute data with no 
                 encapsulation to indicate the length so the length is 
                 implicitly "everything that's present" */

    status = krnlSendMessage( DEFAULTUSER_OBJECT_HANDLE,
                              IMESSAGE_GETATTRIBUTE, &complianceLevel,
                              CRYPT_OPTION_CERT_COMPLIANCELEVEL );
    if( cryptStatusError( status ) )
        return( status );

    /* Read the wrapper for the certificate object's attributes and 
       determine how far we can read */
#ifdef USE_CERTREQ 
    if( type == CRYPT_CERTTYPE_CERTREQUEST )
        {
        status = readCertReqWrapper( stream, 
                                     ( ATTRIBUTE_LIST ** ) attributeListPtrPtr, 
                                     &length, attributeLength, errorLocus, 
                                     errorType );
        }
    else
#endif /* USE_CERTREQ */
        status = readAttributeWrapper( stream, &length, type, attributeLength );
    if( cryptStatusError( status ) )
        return( status );
    endPos = stell( stream ) + length;

    /* Read the collection of attributes.  We allow for a bit of slop for
       software that gets the length encoding wrong by a few bytes */
    TRACE_DEBUG(( "\nReading attributes for certificate object starting at "
                  "offset %d.", stell( stream ) ));
    for( iterationCount = 0;
         stell( stream ) <= endPos - MIN_ATTRIBUTE_SIZE && \
            iterationCount < FAILSAFE_ITERATIONS_LARGE;
         iterationCount++ )
        {
        const ATTRIBUTE_INFO *attributeInfoPtr;
        BYTE oid[ MAX_OID_SIZE + 8 ];
        BOOLEAN criticalFlag = FALSE, ignoreAttribute = FALSE;
        void *attributeDataPtr;
        int oidLength, attributeDataLength;

        /* Read the outer wrapper and determine the attribute type based on
           the OID */
        readSequence( stream, NULL );
        status = readEncodedOID( stream, oid, MAX_OID_SIZE, &oidLength, 
                                 BER_OBJECT_IDENTIFIER );
        if( cryptStatusError( status ) )
            {
            TRACE_DEBUG(( "Couldn't read attribute OID, status %d.", 
                          status ));
            return( status );
            }
        attributeInfoPtr = oidToAttribute( attributeType, oid, oidLength );
        if( attributeInfoPtr != NULL && \
            complianceLevel < decodeComplianceLevel( attributeInfoPtr->typeInfoFlags ) )
            {
            /* If we're running at a lower compliance level than that
               required for the attribute, ignore it by treating it as a
               blob-type attribute */
            TRACE_DEBUG(( "Reading %s (%d) as a blob.", 
                          attributeInfoPtr->description, 
                          attributeInfoPtr->fieldID ));
            attributeInfoPtr = NULL;
            ignoreAttribute = TRUE;
            }

        /* Read the optional critical flag if it's a certificate.  If the
           extension is marked critical and we don't recognise it we don't 
           reject it at this point because that'd make it impossible to 
           examine the contents of the certificate or display it to the 
           user.  Instead we reject the certificate when we try and check 
           it */
        if( attributeType != ATTRIBUTE_CMS && \
            peekTag( stream ) == BER_BOOLEAN )
            {
            status = readBoolean( stream, &criticalFlag );
            if( cryptStatusError( status ) )
                {
                *errorLocus = ( attributeInfoPtr != NULL ) ? \
                              attributeInfoPtr->fieldID : CRYPT_ATTRIBUTE_NONE;
                *errorType = CRYPT_ERRTYPE_ATTR_VALUE;
                return( status );
                }
            }

        /* Read the wrapper around the attribute payload */
        if( wrapperTagSet )
            status = readSet( stream, &attributeDataLength );
        else
            status = readOctetStringHole( stream, &attributeDataLength, 2,
                                          DEFAULT_TAG );
        if( cryptStatusError( status ) )
            {
            *errorLocus = ( attributeInfoPtr != NULL ) ? \
                          attributeInfoPtr->fieldID : CRYPT_ATTRIBUTE_NONE;
            *errorType = CRYPT_ERRTYPE_ATTR_VALUE;
            TRACE_DEBUG(( "Couldn't read attribute wrapper for %s, status %d.", 
                          ( attributeInfoPtr->description != NULL ) ? \
                            attributeInfoPtr->description : \
                            "(unrecognised blob attribute)", status ));
            return( status );
            }

        /* If it's a known attribute, parse the payload */
        if( attributeInfoPtr != NULL )
            {
            status = readAttribute( stream, 
                                    ( ATTRIBUTE_LIST ** ) attributeListPtrPtr,
                                    attributeInfoPtr, attributeDataLength,
                                    criticalFlag, errorLocus, errorType );
            if( cryptStatusError( status ) )
                {
                TRACE_DEBUG(( "Error %d reading %s attribute.", status,
                              ( attributeInfoPtr->description != NULL ) ? \
                                attributeInfoPtr->description : \
                                "(unrecognised blob attribute)" ));
                return( status );
                }
            continue;
            }

        /* If it's a zero-length unrecognised attribute, don't add anything.
           A zero length indicates that the attribute contains all default
           values, however since we don't recognise the attribute we can't
           fill these in so the attribute is in effect not present */
        if( attributeDataLength <= 0 )
            continue;

        /* It's an unrecognised or ignored attribute type, add the raw data
           to the list of attributes */
        status = sMemGetDataBlock( stream, &attributeDataPtr, 
                                   attributeDataLength );
        if( cryptStatusOK( status ) )
            {
            ANALYSER_HINT( attributeDataPtr != NULL );
            status = addAttribute( attributeType, attributeListPtrPtr, 
                                   oid, oidLength, criticalFlag, 
                                   attributeDataPtr, attributeDataLength, 
                                   ignoreAttribute ? \
                                        ATTR_FLAG_BLOB | ATTR_FLAG_IGNORED : \
                                        ATTR_FLAG_BLOB );
            }
        if( cryptStatusError( status ) )
            {
            if( status == CRYPT_ERROR_INITED )
                {
                /* If there's a duplicate attribute present, set error
                   information for it and flag it as a bad data error.  We
                   can't set an error locus since it's an unknown blob */
                *errorLocus = CRYPT_ATTRIBUTE_NONE;
                *errorType = CRYPT_ERRTYPE_ATTR_PRESENT;
                status = CRYPT_ERROR_BADDATA;
                }
            TRACE_DEBUG(( "Error %d adding unrecognised blob attribute data.", 
                          status ));
            return( status );
            }
        sSkip( stream, attributeDataLength );   /* Skip the attribute data */
        }
    ENSURES( iterationCount < FAILSAFE_ITERATIONS_LARGE );
    TRACE_DEBUG(( "Finished reading attributes for certificate object ending at "
                  "offset %d.\n", stell( stream ) ));

    return( CRYPT_OK );
    }
#endif /* USE_CERTIFICATES */