/*++ Copyright (c) 1996 Microsoft Corporation Module Name: ViewSup.c Abstract: This module implements the Index management routines for NtOfs Author: Tom Miller [TomM] 5-Jan-1996 Revision History: --*/ #include "NtfsProc.h" #include "Index.h" // // The local debug trace level // #define Dbg (DEBUG_TRACE_VIEWSUP) // // Define a tag for general pool allocations from this module // #undef MODULE_POOL_TAG #define MODULE_POOL_TAG ('vFtN') // // Temporary definitions for test // BOOLEAN NtOfsDoIndexTest = TRUE; BOOLEAN NtOfsLeaveTestIndex = FALSE; extern ATTRIBUTE_DEFINITION_COLUMNS NtfsAttributeDefinitions[]; // // Define a context for NtOfsReadRecords, which is primarily an IndexContext // and a copy of the last Key returned. // typedef struct _READ_CONTEXT { // // IndexContext (cursor) for the enumeration. // INDEX_CONTEXT IndexContext; // // The last key returned is allocated from paged pool. We have to // separately record how much is allocated, and how long the current // key is using, the latter being in the KeyLength field of IndexKey. // SmallKeyBuffer will store a small key in this structure without going // to pool. // INDEX_KEY LastReturnedKey; ULONG AllocatedKeyLength; ULONG SmallKeyBuffer[3]; } READ_CONTEXT, *PREAD_CONTEXT; #ifdef ALLOC_PRAGMA #pragma alloc_text(PAGE, NtfsQueryViewIndex) #pragma alloc_text(PAGE, NtOfsCreateIndex) #pragma alloc_text(PAGE, NtOfsCloseIndex) #pragma alloc_text(PAGE, NtOfsDeleteIndex) #pragma alloc_text(PAGE, NtOfsFindRecord) #pragma alloc_text(PAGE, NtOfsAddRecords) #pragma alloc_text(PAGE, NtOfsDeleteRecords) #pragma alloc_text(PAGE, NtOfsUpdateRecord) #pragma alloc_text(PAGE, NtOfsReadRecords) #pragma alloc_text(PAGE, NtOfsFreeReadContext) #pragma alloc_text(PAGE, NtOfsFindLastRecord) #pragma alloc_text(PAGE, NtOfsCollateUlong) #pragma alloc_text(PAGE, NtOfsCollateUlongs) #pragma alloc_text(PAGE, NtOfsCollateUnicode) #pragma alloc_text(PAGE, NtOfsMatchAll) #pragma alloc_text(PAGE, NtOfsMatchUlongExact) #pragma alloc_text(PAGE, NtOfsMatchUlongsExact) #pragma alloc_text(PAGE, NtOfsMatchUnicodeExpression) #pragma alloc_text(PAGE, NtOfsMatchUnicodeString) #pragma alloc_text(PAGE, NtOfsCollateSid) #endif NTFSAPI NTSTATUS NtOfsCreateIndex ( IN PIRP_CONTEXT IrpContext, IN PFCB Fcb, IN UNICODE_STRING Name, IN CREATE_OPTIONS CreateOptions, IN ULONG DeleteCollationData, IN ULONG CollationRule, IN PCOLLATION_FUNCTION CollationFunction, IN PVOID CollationData OPTIONAL, OUT PSCB *Scb ) /*++ Routine Description: This routine may be called to create / open a view index within a given file for a given CollationRule. Arguments: Fcb - File in which the index is to be created. Name - Name of the index for all related Scbs and attributes on disk. CreateOptions - Standard create flags. DeleteCollationData - Specifies 1 if the NtfsFreePool should be called for CollationData when no longer required, or 0 if NtfsFreePool should never be called. CollationRule - A binary code to store in the index root to convey the collation function to ChkDsk. These rules are defined in ntfs.h, and must have a one-to-one correspondence with the CollationFunction below. CollationFunction - Function to be called to collate the index. CollationData - Data pointer to be passed to CollationFunction. Scb - Returns an Scb as handle for the index. Return Value: STATUS_OBJECT_NAME_COLLISION -- if CreateNew and index already exists STATUS_OBJECT_NAME_NOT_FOUND -- if OpenExisting and index does not exist --*/ { ATTRIBUTE_ENUMERATION_CONTEXT LocalContext; BOOLEAN FoundAttribute; NTSTATUS Status = STATUS_SUCCESS; PBCB FileRecordBcb = NULL; PFILE_RECORD_SEGMENT_HEADER FileRecord; LONGLONG FileRecordOffset; struct { INDEX_ROOT IndexRoot; INDEX_ENTRY EndEntry; } R; ASSERT_IRP_CONTEXT( IrpContext ); PAGED_CODE(); // // First we will initialize the Index Root structure which is the value // of the attribute we need to create. // RtlZeroMemory( &R, sizeof(R) ); R.IndexRoot.CollationRule = CollationRule; R.IndexRoot.BytesPerIndexBuffer = NTOFS_VIEW_INDEX_BUFFER_SIZE; R.IndexRoot.BlocksPerIndexBuffer = (UCHAR)ClustersFromBytes( Fcb->Vcb, NTOFS_VIEW_INDEX_BUFFER_SIZE ); if (NTOFS_VIEW_INDEX_BUFFER_SIZE < Fcb->Vcb->BytesPerCluster) { R.IndexRoot.BlocksPerIndexBuffer = NTOFS_VIEW_INDEX_BUFFER_SIZE / DEFAULT_INDEX_BLOCK_SIZE; } R.IndexRoot.IndexHeader.FirstIndexEntry = QuadAlign(sizeof(INDEX_HEADER)); R.IndexRoot.IndexHeader.FirstFreeByte = R.IndexRoot.IndexHeader.BytesAvailable = QuadAlign(sizeof(INDEX_HEADER)) + QuadAlign(sizeof(INDEX_ENTRY)); // // Now we need to put in the special End entry. // R.EndEntry.Length = sizeof(INDEX_ENTRY); SetFlag( R.EndEntry.Flags, INDEX_ENTRY_END ); // // Now, just create the Index Root Attribute. // NtfsInitializeAttributeContext( &LocalContext ); NtfsAcquireExclusiveFcb( IrpContext, Fcb, NULL, 0 ); try { // // First see if the index already exists, by searching for the root // attribute. // FoundAttribute = NtfsLookupAttributeByName( IrpContext, Fcb, &Fcb->FileReference, $INDEX_ROOT, &Name, NULL, TRUE, &LocalContext ); // // If it is not there, and the CreateOptions allow, then let's create // the index root now. (First cleaning up the attribute context from // the lookup). // if (!FoundAttribute && (CreateOptions <= CREATE_OR_OPEN)) { NtfsCleanupAttributeContext( IrpContext, &LocalContext ); NtfsCreateAttributeWithValue( IrpContext, Fcb, $INDEX_ROOT, &Name, &R, sizeof(R), 0, NULL, TRUE, &LocalContext ); // // If the index is already there, and we were asked to create it, then // return an error. // } else if (FoundAttribute && (CreateOptions == CREATE_NEW)) { try_return( Status = STATUS_OBJECT_NAME_COLLISION ); // // If the index is not there, and we were supposed to open existing, then // return an error. // } else if (!FoundAttribute && (CreateOptions == OPEN_EXISTING)) { try_return( Status = STATUS_OBJECT_NAME_NOT_FOUND ); } // // Otherwise create/find the Scb and reference it. // *Scb = NtfsCreateScb( IrpContext, Fcb, $INDEX_ALLOCATION, &Name, FALSE, NULL ); SetFlag( (*Scb)->ScbState, SCB_STATE_VIEW_INDEX ); (*Scb)->ScbType.Index.CollationFunction = CollationFunction; // // Handle the case where CollationData is to be deleted. // if (DeleteCollationData) { SetFlag((*Scb)->ScbState, SCB_STATE_DELETE_COLLATION_DATA); if ((*Scb)->ScbType.Index.CollationData != NULL) { NtfsFreePool(CollationData); } else { (*Scb)->ScbType.Index.CollationData = CollationData; } // // Otherwise just jam the pointer the caller passed. // } else { (*Scb)->ScbType.Index.CollationData = CollationData; } NtfsIncrementCloseCounts( *Scb, TRUE, FALSE ); // // We have to set the view index present bit, so read it, save the // old data and set the flag here. // NtfsPinMftRecord( IrpContext, Fcb->Vcb, &Fcb->FileReference, FALSE, &FileRecordBcb, &FileRecord, &FileRecordOffset ); // // If necessary, set the flag to indicate that this file will have // no unnamed data stream and any attempt to open this file without // specifying a named stream will fail, but without marking the // volume corrupt. // if (!FlagOn( FileRecord->Flags, FILE_VIEW_INDEX_PRESENT )) { // // We have to be very careful when using the InitialzeFileRecordSegment // log record. This action is applied unconditionally. DoAction doesn't // check the previous LSN in the page. It may be garbage on a newly initialized // file record. We log the entire file record to avoid the case where we // might overwrite a later Lsn with this earlier Lsn during restart. // // // Log the existing file record as the undo action. // FileRecord->Lsn = NtfsWriteLog( IrpContext, Fcb->Vcb->MftScb, FileRecordBcb, Noop, NULL, 0, InitializeFileRecordSegment, FileRecord, FileRecord->FirstFreeByte, FileRecordOffset, 0, 0, Fcb->Vcb->BytesPerFileRecordSegment ); SetFlag( FileRecord->Flags, FILE_VIEW_INDEX_PRESENT ); // // Log the new file record. // FileRecord->Lsn = NtfsWriteLog( IrpContext, Fcb->Vcb->MftScb, FileRecordBcb, InitializeFileRecordSegment, FileRecord, FileRecord->FirstFreeByte, Noop, NULL, 0, FileRecordOffset, 0, 0, Fcb->Vcb->BytesPerFileRecordSegment ); } try_exit: NOTHING; } finally { NtfsCleanupAttributeContext( IrpContext, &LocalContext ); NtfsUnpinBcb( IrpContext, &FileRecordBcb ); NtfsReleaseFcb( IrpContext, Fcb ); } return Status; } NTFSAPI VOID NtOfsCloseIndex ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb ) /*++ Routine Description: This routine may be called to close a previously returned handle on a view index. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. Return Value: None. --*/ { ASSERT( NtfsIsExclusiveFcb( Scb->Fcb )); NtfsDecrementCloseCounts( IrpContext, Scb, NULL, TRUE, FALSE, FALSE, NULL ); } NTFSAPI VOID NtOfsDeleteIndex ( IN PIRP_CONTEXT IrpContext, IN PFCB Fcb, IN PSCB Scb ) /*++ Routine Description: This routine may be called to delete an index. Arguments: Fcb - Supplies an Fcb as the previously returned object handle for the file Scb - Supplies an Scb as the previously returned handle for this index. Return Value: None (Deleting a nonexistant index is benign). --*/ { ATTRIBUTE_ENUMERATION_CONTEXT LocalContext; ATTRIBUTE_TYPE_CODE AttributeTypeCode; BOOLEAN FoundAttribute; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT(($BITMAP - $INDEX_ALLOCATION) == ($INDEX_ALLOCATION - $INDEX_ROOT)); PAGED_CODE(); NtfsAcquireExclusiveScb( IrpContext, Scb ); try { // // First see if there is some index allocation, and if so truncate it // away allowing this operation to be broken up. // NtfsInitializeAttributeContext( &LocalContext ); if (NtfsLookupAttributeByName( IrpContext, Fcb, &Fcb->FileReference, $INDEX_ALLOCATION, &Scb->AttributeName, NULL, FALSE, &LocalContext )) { NtfsCreateInternalAttributeStream( IrpContext, Scb, TRUE, NULL ); NtfsDeleteAllocation( IrpContext, NULL, Scb, 0, MAXLONGLONG, TRUE, TRUE ); } NtfsCleanupAttributeContext( IrpContext, &LocalContext ); for (AttributeTypeCode = $INDEX_ROOT; AttributeTypeCode <= $BITMAP; AttributeTypeCode += ($INDEX_ALLOCATION - $INDEX_ROOT)) { // // Initialize the attribute context on each trip through the loop. // NtfsInitializeAttributeContext( &LocalContext ); // // First see if the index already exists, by searching for the root // attribute. // FoundAttribute = NtfsLookupAttributeByName( IrpContext, Fcb, &Fcb->FileReference, AttributeTypeCode, &Scb->AttributeName, NULL, TRUE, &LocalContext ); // // Loop while we see the right records. // while (FoundAttribute) { NtfsDeleteAttributeRecord( IrpContext, Fcb, DELETE_LOG_OPERATION | DELETE_RELEASE_FILE_RECORD | DELETE_RELEASE_ALLOCATION, &LocalContext ); FoundAttribute = NtfsLookupNextAttributeByName( IrpContext, Fcb, AttributeTypeCode, &Scb->AttributeName, TRUE, &LocalContext ); } NtfsCleanupAttributeContext( IrpContext, &LocalContext ); } SetFlag( Scb->ScbState, SCB_STATE_ATTRIBUTE_DELETED ); } finally { NtfsCleanupAttributeContext( IrpContext, &LocalContext ); NtfsReleaseScb( IrpContext, Scb ); } } NTFSAPI NTSTATUS NtOfsFindRecord ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb, IN PINDEX_KEY IndexKey, OUT PINDEX_ROW IndexRow, OUT PMAP_HANDLE MapHandle, IN OUT PQUICK_INDEX_HINT QuickIndexHint OPTIONAL ) /*++ Routine Description: This routine may be called to find the first occurrence of a key in an index, and return cached information which may can accelerate the update on the data for that key if the index buffer is not changed. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. IndexKey - Supplies the key to find. IndexRow - Returns a description of the Key and Data *in place*, for read-only access, valid only until the Bcb is unpinned. (Neither key nor data may be modified in place!) MapHandle - Returns a map handle for accessing the key and data directly. QuickIndexHint - Supplies a previously returned hint, or all zeros on first use. Returns location information which may be held an arbitrary amount of time, which can accelerate a subsequent call to NtOfsUpdateRecord for the data in this key, iff changes to the index do not prohibit use of this hint. Return Value: STATUS_SUCCESS -- if operation was successful. STATUS_NO_MATCH -- if the specified key does not exist. --*/ { INDEX_CONTEXT IndexContext; PINDEX_LOOKUP_STACK Sp; PINDEX_ENTRY IndexEntry; NTSTATUS Status; PQUICK_INDEX QuickIndex = (PQUICK_INDEX)QuickIndexHint; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_SCB( Scb ); PAGED_CODE(); Status = STATUS_SUCCESS; NtfsInitializeIndexContext( &IndexContext ); ASSERT_SHARED_SCB( Scb ); try { // // Use the second location in the index context to perform the // read. // Sp = IndexContext.Current = IndexContext.Base + 1; // // If the index entry for this filename hasn't moved we can go // directly to the location in the buffer. For this to be the case the // following must be true. // // - The entry must already be in an index buffer (BufferOffset test) // - The index stream may not have been truncated (ChangeCount test) // - The Lsn in the page can't have changed // if (ARGUMENT_PRESENT( QuickIndexHint ) && (QuickIndex->BufferOffset != 0) && (QuickIndex->ChangeCount == Scb->ScbType.Index.ChangeCount)) { ReadIndexBuffer( IrpContext, Scb, QuickIndex->IndexBlock, FALSE, Sp ); // // If the Lsn matches then we can use this buffer directly. // if (QuickIndex->CapturedLsn.QuadPart == Sp->CapturedLsn.QuadPart) { Sp->IndexEntry = (PINDEX_ENTRY) Add2Ptr( Sp->StartOfBuffer, QuickIndex->BufferOffset ); // // Otherwise we need to reinitialize the index context and take // the long path below. // } else { NtfsReinitializeIndexContext( IrpContext, &IndexContext ); } } // // If we did not get the index entry via the hint, get it now. // if (Sp->Bcb == NULL) { // // Position to first possible match. // FindFirstIndexEntry( IrpContext, Scb, IndexKey, &IndexContext ); // // See if there is an actual match. // if (!FindNextIndexEntry( IrpContext, Scb, IndexKey, FALSE, FALSE, &IndexContext, FALSE, NULL )) { try_return( Status = STATUS_NO_MATCH ); } } // // Basic consistency check // IndexEntry = IndexContext.Current->IndexEntry; if ((IndexEntry->DataOffset + IndexEntry->DataLength > IndexEntry->Length) || (IndexEntry->AttributeLength + sizeof( INDEX_ENTRY ) > IndexEntry->Length)) { NtfsRaiseStatus( IrpContext, STATUS_FILE_CORRUPT_ERROR, NULL, Scb->Fcb ); } // // If we found the key in the base, then return the Bcb from the // attribute context and return no hint (BufferOffset = 0). // if (IndexContext.Current == IndexContext.Base) { MapHandle->Buffer = NULL; MapHandle->Bcb = NtfsFoundBcb(&IndexContext.AttributeContext); NtfsFoundBcb(&IndexContext.AttributeContext) = NULL; if (ARGUMENT_PRESENT( QuickIndexHint )) { QuickIndex->BufferOffset = 0; } // // If we found the key in an index buffer, then return the Bcb from // the lookup stack, and record the hint for the caller. // } else { Sp = IndexContext.Current; MapHandle->Buffer = Sp->StartOfBuffer; MapHandle->Bcb = Sp->Bcb; Sp->Bcb = NULL; if (ARGUMENT_PRESENT( QuickIndexHint )) { QuickIndex->ChangeCount = Scb->ScbType.Index.ChangeCount; QuickIndex->BufferOffset = PtrOffset( Sp->StartOfBuffer, Sp->IndexEntry ); QuickIndex->CapturedLsn = ((PINDEX_ALLOCATION_BUFFER) Sp->StartOfBuffer)->Lsn; QuickIndex->IndexBlock = ((PINDEX_ALLOCATION_BUFFER) Sp->StartOfBuffer)->ThisBlock; } } // // Return the IndexRow described directly in the buffer. // IndexRow->KeyPart.Key = (IndexEntry + 1); IndexRow->KeyPart.KeyLength = IndexEntry->AttributeLength; IndexRow->DataPart.Data = Add2Ptr( IndexEntry, IndexEntry->DataOffset ); IndexRow->DataPart.DataLength = IndexEntry->DataLength; try_exit: NOTHING; } finally { NtfsCleanupIndexContext( IrpContext, &IndexContext ); } return Status; } NTFSAPI NTSTATUS NtOfsFindLastRecord ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb, IN PINDEX_KEY MaxIndexKey, OUT PINDEX_ROW IndexRow, OUT PMAP_HANDLE MapHandle ) /*++ Routine Description: This routine may be called to find the highest key in an index. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. MaxIndexKey - Supplies the maximum possible key value (such as MAXULONG, etc.), and this key must not actually be in use! IndexRow - Returns a description of the Key and Data *in place*, for read-only access, valid only until the Bcb is unpinned. (Neither key nor data may be modified in place!) MapHandle - Returns a map handle for accessing the key and data directly. Return Value: STATUS_SUCCESS -- if operation was successful. STATUS_NO_MATCH -- if the specified key does not exist (index is empty). --*/ { INDEX_CONTEXT IndexContext; PINDEX_LOOKUP_STACK Sp; PINDEX_ENTRY IndexEntry, NextIndexEntry; NTSTATUS Status; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_SCB( Scb ); PAGED_CODE(); Status = STATUS_SUCCESS; NtfsInitializeIndexContext( &IndexContext ); NtfsAcquireSharedScb( IrpContext, Scb ); try { // // Slide down the "right" side of the tree. // FindFirstIndexEntry( IrpContext, Scb, MaxIndexKey, &IndexContext ); // // If this happens, the index must be empty. // Sp = IndexContext.Current; IndexEntry = NtfsFirstIndexEntry(Sp->IndexHeader); if (FlagOn(IndexEntry->Flags, INDEX_ENTRY_END)) { try_return( Status = STATUS_NO_MATCH ); } // // If we found the key in the base, then return the Bcb from the // attribute context and return no hint (BufferOffset = 0). // if (IndexContext.Current == IndexContext.Base) { MapHandle->Bcb = NtfsFoundBcb(&IndexContext.AttributeContext); NtfsFoundBcb(&IndexContext.AttributeContext) = NULL; // // If we found the key in an index buffer, then return the Bcb from // the lookup stack, and record the hint for the caller. // } else { MapHandle->Bcb = Sp->Bcb; Sp->Bcb = NULL; } // // Complete the MapHandle to disallow pinning. // MapHandle->Buffer = NULL; // // Now rescan the last buffer to return the second to last index entry, // if there is one. // NextIndexEntry = IndexEntry; do { IndexEntry = NextIndexEntry; if (IndexEntry->Length == 0) { NtfsRaiseStatus( IrpContext, STATUS_FILE_CORRUPT_ERROR, NULL, Scb->Fcb ); } NextIndexEntry = NtfsNextIndexEntry(IndexEntry); NtfsCheckIndexBound( NextIndexEntry, Sp->IndexHeader ); } while (!FlagOn(NextIndexEntry->Flags, INDEX_ENTRY_END)); // // Return the IndexRow described directly in the buffer. // IndexRow->KeyPart.Key = (IndexEntry + 1); IndexRow->KeyPart.KeyLength = IndexEntry->AttributeLength; IndexRow->DataPart.Data = Add2Ptr( IndexEntry, IndexEntry->DataOffset ); IndexRow->DataPart.DataLength = IndexEntry->DataLength; try_exit: NOTHING; } finally { NtfsCleanupIndexContext( IrpContext, &IndexContext ); NtfsReleaseScb( IrpContext, Scb ); } return Status; } NTFSAPI VOID NtOfsAddRecords ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb, IN ULONG Count, IN PINDEX_ROW IndexRow, IN ULONG SequentialInsertMode ) /*++ Routine Description: This routine may be called to add one or more records to an index. If SequentialInsertMode is nonzero, this is a hint to the index package to keep all BTree buffers as full as possible, by splitting as close to the end of the buffer as possible. If specified as zero, random inserts are assumed, and buffers are always split in the middle for better balance. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. Count - Supplies the number of records being added. IndexRow - Supplies an array of Count entries, containing the Keys and Data to add. SequentialInsertMode - If specified as nozero, the implementation may choose to split all index buffers at the end for maximum fill. Return Value: None. Raises: STATUS_DUPLICATE_NAME -- if the specified key already exists. --*/ { INDEX_CONTEXT IndexContext; struct { INDEX_ENTRY IndexEntry; PVOID Key; PVOID Data; } IE; ULONG i; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_SCB( Scb ); UNREFERENCED_PARAMETER(SequentialInsertMode); PAGED_CODE(); NtfsInitializeIndexContext( &IndexContext ); NtfsAcquireExclusiveScb( IrpContext, Scb ); try { // // Loop to add all entries // for (i = 0; i < Count; i++) { // // Position to first possible match. // FindFirstIndexEntry( IrpContext, Scb, &IndexRow->KeyPart, &IndexContext ); // // See if there is an actual match. // if (FindNextIndexEntry( IrpContext, Scb, &IndexRow->KeyPart, FALSE, FALSE, &IndexContext, FALSE, NULL )) { NtfsRaiseStatus( IrpContext, STATUS_DUPLICATE_NAME, NULL, NULL ); } // // Initialize the Index Entry in pointer form. // // Note that the final index entry ends up looking like this: // // (IndexEntry)(Key)(Data) // // where all fields are long-aligned and: // // Key is at IndexEntry + sizeof(INDEX_ENTRY), and of length AttributeLength // Data is at IndexEntry + DataOffset and of length DataLength // IE.IndexEntry.AttributeLength = (USHORT)IndexRow->KeyPart.KeyLength; IE.IndexEntry.DataOffset = (USHORT)(sizeof(INDEX_ENTRY) + LongAlign( IndexRow->KeyPart.KeyLength )); IE.IndexEntry.DataLength = (USHORT)IndexRow->DataPart.DataLength; IE.IndexEntry.ReservedForZero = 0; IE.IndexEntry.Length = (USHORT)(QuadAlign(IE.IndexEntry.DataOffset + IndexRow->DataPart.DataLength)); IE.IndexEntry.Flags = INDEX_ENTRY_POINTER_FORM; IE.IndexEntry.Reserved = 0; IE.Key = IndexRow->KeyPart.Key; IE.Data = IndexRow->DataPart.Data; // // Now add it to the index. We can only add to a leaf, so force our // position back to the correct spot in a leaf first. // IndexContext.Current = IndexContext.Top; AddToIndex( IrpContext, Scb, (PINDEX_ENTRY)&IE, &IndexContext, NULL, FALSE ); NtfsReinitializeIndexContext( IrpContext, &IndexContext ); IndexRow += 1; } } finally { NtfsCleanupIndexContext( IrpContext, &IndexContext ); NtfsReleaseScb( IrpContext, Scb ); } } NTFSAPI VOID NtOfsDeleteRecords ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb, IN ULONG Count, IN PINDEX_KEY IndexKey ) /*++ Routine Description: This routine may be called to delete one or more records from an index. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. Count - Supplies the number of records being deleted. IndexKey - Supplies an array of Count entries, containing the Keys to be deleted. Return Value: None. (This call is benign if any records do not exist.) --*/ { INDEX_CONTEXT IndexContext; ULONG i; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_SCB( Scb ); PAGED_CODE(); NtfsInitializeIndexContext( &IndexContext ); NtfsAcquireExclusiveScb( IrpContext, Scb ); try { // // Loop to add all entries // for (i = 0; i < Count; i++) { // // Position to first possible match. // FindFirstIndexEntry( IrpContext, Scb, IndexKey, &IndexContext ); // // See if there is an actual match. // if (FindNextIndexEntry( IrpContext, Scb, IndexKey, FALSE, FALSE, &IndexContext, FALSE, NULL )) { // // Delete it. // DeleteFromIndex( IrpContext, Scb, &IndexContext ); } NtfsReinitializeIndexContext( IrpContext, &IndexContext ); IndexKey += 1; } } finally { NtfsCleanupIndexContext( IrpContext, &IndexContext ); NtfsReleaseScb( IrpContext, Scb ); } } NTFSAPI VOID NtOfsUpdateRecord ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb, IN ULONG Count, IN PINDEX_ROW IndexRow, IN OUT PQUICK_INDEX_HINT QuickIndexHint OPTIONAL, IN OUT PMAP_HANDLE MapHandle OPTIONAL ) /*++ Routine Description: This routine may be called to update the data portion of a record in an index. If QuickIndexHint is specified, then the update may occur by directly accessing the buffer containing the specified key, iff other changes to the index do not prevent that. If changes prevent the quick update, then the record is looked up by key in order to perform the data update. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. Count - Supplies the count of updates described in IndexRow. For counts greater than 1, QuickIndexHint and MapHandle must not be supplied. IndexRow - Supplies the key to be updated and the new data for that key. QuickIndexHint - Supplies a optional quick index for this row returned from a previous call to NtOfsFindRecord, updated on return. MapHandle - Supplies an optional MapHandle to accompany the QuickIndex. If MapHandle is supplied, then the QuickIndexHint must be guaranteed valid. MapHandle is updated (pinned) on return. MapHandle is ignored if QuickIndexHint is not specified. Return Value: None. Raises: STATUS_INFO_LENGTH_MISMATCH -- if the specified data is a different length from the data in the key. STATUS_NO_MATCH -- if the specified key does not exist. --*/ { INDEX_CONTEXT IndexContext; PQUICK_INDEX QuickIndex = (PQUICK_INDEX)QuickIndexHint; PVOID DataInIndex; PINDEX_ENTRY IndexEntry; PVCB Vcb = Scb->Vcb; PINDEX_LOOKUP_STACK Sp; PINDEX_ALLOCATION_BUFFER IndexBuffer; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_SCB( Scb ); ASSERT_SHARED_SCB( Scb ); ASSERT(Count != 0); PAGED_CODE(); NtfsInitializeIndexContext( &IndexContext ); try { // // If the index entry for this filename hasn't moved we can go // directly to the location in the buffer. For this to be the case the // following must be true. // // - The entry must already be in an index buffer (BufferOffset test) // - The index stream may not have been truncated (ChangeCount test) // - The Lsn in the page can't have changed // if (ARGUMENT_PRESENT( QuickIndexHint ) && (QuickIndex->BufferOffset != 0) && (QuickIndex->ChangeCount == Scb->ScbType.Index.ChangeCount)) { ASSERT(Count == 1); // // Use the top location in the index context to perform the // read. // Sp = IndexContext.Base; // // If we have a MapHandle already, we do not need to read the // IndexBuffer. // if (ARGUMENT_PRESENT(MapHandle)) { IndexBuffer = MapHandle->Buffer; Sp->Bcb = MapHandle->Bcb; MapHandle->Bcb = NULL; Sp->CapturedLsn.QuadPart = QuickIndex->CapturedLsn.QuadPart; } else { ReadIndexBuffer( IrpContext, Scb, QuickIndex->IndexBlock, FALSE, Sp ); IndexBuffer = Sp->StartOfBuffer; } // // If the Lsn matches then we can use this buffer directly. // if (QuickIndex->CapturedLsn.QuadPart == Sp->CapturedLsn.QuadPart) { IndexEntry = (PINDEX_ENTRY) Add2Ptr( IndexBuffer, QuickIndex->BufferOffset ); if (IndexEntry->DataLength < IndexRow->DataPart.DataLength) { NtfsRaiseStatus( IrpContext, STATUS_INFO_LENGTH_MISMATCH, NULL, NULL ); } DataInIndex = Add2Ptr( IndexEntry, IndexEntry->DataOffset ); // // Pin the index buffer // NtfsPinMappedData( IrpContext, Scb, LlBytesFromIndexBlocks( IndexBuffer->ThisBlock, Scb->ScbType.Index.IndexBlockByteShift ), Scb->ScbType.Index.BytesPerIndexBuffer, &Sp->Bcb ); // // Write a log record to change our ParentIndexEntry. // // // Write the log record, but do not update the IndexBuffer Lsn, // since nothing moved and we don't want to force index contexts // to have to rescan. // // Indexbuffer->Lsn = // // ASSERT(Scb->ScbType.Index.ClustersPerIndexBuffer != 0); NtfsWriteLog( IrpContext, Scb, Sp->Bcb, UpdateRecordDataAllocation, IndexRow->DataPart.Data, IndexRow->DataPart.DataLength, UpdateRecordDataAllocation, DataInIndex, IndexRow->DataPart.DataLength, LlBytesFromIndexBlocks( IndexBuffer->ThisBlock, Scb->ScbType.Index.IndexBlockByteShift ), 0, QuickIndex->BufferOffset, Scb->ScbType.Index.BytesPerIndexBuffer ); // // Now call the Restart routine to do it. // NtOfsRestartUpdateDataInIndex( IndexEntry, IndexRow->DataPart.Data, IndexRow->DataPart.DataLength ); // // If there is a MapHandle, we must update the Bcb pointer. // if (ARGUMENT_PRESENT(MapHandle)) { MapHandle->Bcb = Sp->Bcb; Sp->Bcb = NULL; } leave; // // Otherwise we need to unpin the Bcb and take // the long path below. // } else { ASSERT(!ARGUMENT_PRESENT(MapHandle)); NtfsUnpinBcb( IrpContext, &Sp->Bcb ); } } // // Loop to apply all updates. // do { // // Position to first possible match. // FindFirstIndexEntry( IrpContext, Scb, &IndexRow->KeyPart, &IndexContext ); // // See if there is an actual match. // if (FindNextIndexEntry( IrpContext, Scb, &IndexRow->KeyPart, FALSE, FALSE, &IndexContext, FALSE, NULL )) { // // Point to the index entry and the data within it. // IndexEntry = IndexContext.Current->IndexEntry; if (IndexEntry->DataLength < IndexRow->DataPart.DataLength) { NtfsRaiseStatus( IrpContext, STATUS_INFO_LENGTH_MISMATCH, NULL, NULL ); } DataInIndex = Add2Ptr( IndexEntry, IndexEntry->DataOffset ); // // Now pin the entry. // if (IndexContext.Current == IndexContext.Base) { PFILE_RECORD_SEGMENT_HEADER FileRecord; PATTRIBUTE_RECORD_HEADER Attribute; PATTRIBUTE_ENUMERATION_CONTEXT Context = &IndexContext.AttributeContext; // // Pin the root // NtfsPinMappedAttribute( IrpContext, Vcb, Context ); // // Write a log record to change our ParentIndexEntry. // FileRecord = NtfsContainingFileRecord(Context); Attribute = NtfsFoundAttribute(Context); // // Write the log record, but do not update the FileRecord Lsn, // since nothing moved and we don't want to force index contexts // to have to rescan. // // FileRecord->Lsn = // NtfsWriteLog( IrpContext, Vcb->MftScb, NtfsFoundBcb(Context), UpdateRecordDataRoot, IndexRow->DataPart.Data, IndexRow->DataPart.DataLength, UpdateRecordDataRoot, DataInIndex, IndexRow->DataPart.DataLength, NtfsMftOffset( Context ), (ULONG)((PCHAR)Attribute - (PCHAR)FileRecord), (ULONG)((PCHAR)IndexEntry - (PCHAR)Attribute), Vcb->BytesPerFileRecordSegment ); if (ARGUMENT_PRESENT( QuickIndexHint )) { ASSERT( Count == 1 ); QuickIndex->BufferOffset = 0; } } else { Sp = IndexContext.Current; IndexBuffer = (PINDEX_ALLOCATION_BUFFER)Sp->StartOfBuffer; // // Pin the index buffer // NtfsPinMappedData( IrpContext, Scb, LlBytesFromIndexBlocks( IndexBuffer->ThisBlock, Scb->ScbType.Index.IndexBlockByteShift ), Scb->ScbType.Index.BytesPerIndexBuffer, &Sp->Bcb ); // // Write a log record to change our ParentIndexEntry. // // // Write the log record, but do not update the IndexBuffer Lsn, // since nothing moved and we don't want to force index contexts // to have to rescan. // // Indexbuffer->Lsn = // NtfsWriteLog( IrpContext, Scb, Sp->Bcb, UpdateRecordDataAllocation, IndexRow->DataPart.Data, IndexRow->DataPart.DataLength, UpdateRecordDataAllocation, DataInIndex, IndexRow->DataPart.DataLength, LlBytesFromIndexBlocks( IndexBuffer->ThisBlock, Scb->ScbType.Index.IndexBlockByteShift ), 0, (ULONG)((PCHAR)Sp->IndexEntry - (PCHAR)IndexBuffer), Scb->ScbType.Index.BytesPerIndexBuffer ); if (ARGUMENT_PRESENT( QuickIndexHint )) { ASSERT( Count == 1 ); QuickIndex->ChangeCount = Scb->ScbType.Index.ChangeCount; QuickIndex->BufferOffset = PtrOffset( Sp->StartOfBuffer, Sp->IndexEntry ); QuickIndex->CapturedLsn = ((PINDEX_ALLOCATION_BUFFER) Sp->StartOfBuffer)->Lsn; QuickIndex->IndexBlock = ((PINDEX_ALLOCATION_BUFFER) Sp->StartOfBuffer)->ThisBlock; } } // // Now call the Restart routine to do it. // NtOfsRestartUpdateDataInIndex( IndexEntry, IndexRow->DataPart.Data, IndexRow->DataPart.DataLength ); // // If the file name is not in the index, this is a bad file. // } else { NtfsRaiseStatus( IrpContext, STATUS_NO_MATCH, NULL, NULL ); } // // Get ready for the next pass through. // NtfsReinitializeIndexContext( IrpContext, &IndexContext ); IndexRow += 1; } while (--Count); } finally { NtfsCleanupIndexContext( IrpContext, &IndexContext ); } return; } NTFSAPI NTSTATUS NtOfsReadRecords ( IN PIRP_CONTEXT IrpContext, IN PSCB Scb, IN OUT PREAD_CONTEXT *ReadContext, IN PINDEX_KEY IndexKey OPTIONAL, IN PMATCH_FUNCTION MatchFunction, IN PVOID MatchData, IN OUT ULONG *Count, OUT PINDEX_ROW Rows, IN ULONG BufferLength, OUT PVOID Buffer ) /*++ Routine Description: This routine may be called to enumerate rows in an index, in collated order. It only returns records accepted by the match function. IndexKey may be specified at any time to start a new search from IndexKey, and IndexKey must be specified on the first call for a given IrpContext (and *ReadContext must be NULL). The read terminates when either *Count records have been returned, or BufferLength has been exhausted, or there are no more matching records. NtOfsReadRecords will seek to the appropriate point in the BTree (as defined by the IndexKey or saved position and the CollateFunction) and begin calling MatchFunction for each record. It continues doing this while MatchFunction returns STATUS_SUCCESS. If MatchFunction returns STATUS_NO_MORE_MATCHES, NtOfsReadRecords will cache this result and not call MatchFunction again until called with a non-NULL IndexKey. Note that this call is self-synchronized, such that successive calls to the routine are guaranteed to make progress through the index and to return items in Collation order, in spite of Add and Delete record calls being interspersed with Read records calls. Arguments: Scb - Supplies an Scb as the previously returned handle for this index. ReadContext - On the first call this must supply a pointer to NULL. On return a pointer to a private context structure is returned, which must then be supplied on all subsequent calls. This structure must be eventually be freed via NtOfsFreeReadContext. IndexKey - If specified, supplies the key from which the enumeration is to start/resume. It must be specified on the first call when *ReadContext is NULL. MatchFunction - Supplies the MatchFunction to be called to determine which rows to return. MatchData - Supplies the MatchData to be specified on each call to the MatchFunction. Count - Supplies the count of how many rows may be received, and returns the number of rows actually being returned. Rows - Returns the Count row descriptions. BufferLength - Supplies the length of the buffer in bytes, into which the row keys and data are copied upon return. Buffer - Supplies the buffer into which the rows may be copied. Return Value: STATUS_SUCCESS -- if operation was successful. STATUS_NO_MATCH -- if there is no match for the specified IndexKey. STATUS_NO_MORE_MATCHES -- if a match is returned or previously returned, but there are no more matches. --*/ { PINDEX_CONTEXT IndexContext; PINDEX_ENTRY IndexEntry; ULONG LengthToCopy; BOOLEAN MustRestart; ULONG BytesRemaining = BufferLength; ULONG ReturnCount = 0; NTSTATUS Status; BOOLEAN NextFlag; ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_SCB( Scb ); PAGED_CODE(); // // On the first lookup, their must be a key. // ASSERT((IndexKey != NULL) || (*ReadContext != NULL)); // // Everything must be Ulong aligned and sized. // ASSERT(IsLongAligned(Buffer)); ASSERT(IsLongAligned(BufferLength)); Status = STATUS_SUCCESS; NextFlag = FALSE; // // Pick up the IndexContext, allocating one if we need to. // if (*ReadContext == NULL) { *ReadContext = NtfsAllocatePool(PagedPool, sizeof(READ_CONTEXT) ); NtfsInitializeIndexContext( &(*ReadContext)->IndexContext ); (*ReadContext)->LastReturnedKey.Key = &(*ReadContext)->SmallKeyBuffer[0]; (*ReadContext)->LastReturnedKey.KeyLength = 0; (*ReadContext)->AllocatedKeyLength = sizeof(READ_CONTEXT) - FIELD_OFFSET(READ_CONTEXT, SmallKeyBuffer[0]); } IndexContext = &(*ReadContext)->IndexContext; // // Store the MatchFunction and Data in the IndexContext, for the enumerations. // IndexContext->MatchFunction = MatchFunction; IndexContext->MatchData = MatchData; NtfsAcquireSharedScb( IrpContext, Scb ); try { // // If a Key was passed, position to the first possible match. // if (ARGUMENT_PRESENT(IndexKey)) { FindFirstIndexEntry( IrpContext, Scb, IndexKey, IndexContext ); // // Otherwise return here if we hit the end of the matches last time. // } else if ((*ReadContext)->LastReturnedKey.KeyLength == 0) { try_return( Status = STATUS_NO_MORE_MATCHES ); } // // Loop while we still have space to store rows. // while (ReturnCount <= *Count) { // // If we're already at the end, don't call FindNextIndexEntry again. // ASSERT(ARGUMENT_PRESENT(IndexKey) || ((*ReadContext)->LastReturnedKey.KeyLength != 0)); // // See if there is an actual match. // if (!FindNextIndexEntry( IrpContext, Scb, NULL, // Not needed because of Match Function TRUE, FALSE, IndexContext, NextFlag, &MustRestart )) { // // First handle the restart case by resuming from the last // key returned, and skip that one. // if (MustRestart) { ASSERT(!ARGUMENT_PRESENT(IndexKey)); NtfsReinitializeIndexContext( IrpContext, IndexContext ); FindFirstIndexEntry( IrpContext, Scb, &(*ReadContext)->LastReturnedKey, IndexContext ); // // Set NextFlag to TRUE, so we can go back and skip // the key we resumed on. // NextFlag = TRUE; continue; } // // No (more) entries - remember that the enumeration is done. // (*ReadContext)->LastReturnedKey.KeyLength = 0; // // Return the appropriate code based on whether we have returned // any matches yet or not. // if ((ReturnCount == 0) && ARGUMENT_PRESENT(IndexKey)) { Status = STATUS_NO_MATCH; } else { Status = STATUS_NO_MORE_MATCHES; } try_return(Status); } // // We always need to go one beyond the one we can return to keep // all resume cases the same, so now is the time to get out if the // count is finished. // if (ReturnCount == *Count) { break; } // // Now we must always move to the next. // NextFlag = TRUE; // // First try to copy the key. // IndexEntry = IndexContext->Current->IndexEntry; LengthToCopy = IndexEntry->AttributeLength; if (LengthToCopy > BytesRemaining) { break; } RtlCopyMemory( Buffer, IndexEntry + 1, LengthToCopy ); Rows->KeyPart.Key = Buffer; Rows->KeyPart.KeyLength = LengthToCopy; LengthToCopy = LongAlign(LengthToCopy); Buffer = Add2Ptr( Buffer, LengthToCopy ); BytesRemaining -= LengthToCopy; // // Now try to copy the data. // LengthToCopy = IndexEntry->DataLength; if (LengthToCopy > BytesRemaining) { break; } RtlCopyMemory( Buffer, Add2Ptr(IndexEntry, IndexEntry->DataOffset), LengthToCopy ); Rows->DataPart.Data = Buffer; Rows->DataPart.DataLength = LengthToCopy; LengthToCopy = LongAlign(LengthToCopy); Buffer = Add2Ptr( Buffer, LengthToCopy ); BytesRemaining -= LengthToCopy; // // Capture this key before looping back. // // First see if there is enough space. // if (Rows->KeyPart.KeyLength > (*ReadContext)->AllocatedKeyLength) { PVOID NewBuffer; // // Allocate a new buffer. // LengthToCopy = LongAlign(Rows->KeyPart.KeyLength + 16); NewBuffer = NtfsAllocatePool(PagedPool, LengthToCopy ); // // Delete old key buffer? // if ((*ReadContext)->LastReturnedKey.Key != &(*ReadContext)->SmallKeyBuffer[0]) { NtfsFreePool( (*ReadContext)->LastReturnedKey.Key ); } (*ReadContext)->LastReturnedKey.Key = NewBuffer; (*ReadContext)->AllocatedKeyLength = LengthToCopy; } RtlCopyMemory( (*ReadContext)->LastReturnedKey.Key, Rows->KeyPart.Key, Rows->KeyPart.KeyLength ); (*ReadContext)->LastReturnedKey.KeyLength = Rows->KeyPart.KeyLength; Rows += 1; ReturnCount += 1; } try_exit: NOTHING; } finally { #ifdef BENL_DBG ASSERT( (*ReadContext)->AllocatedKeyLength >= (*ReadContext)->LastReturnedKey.KeyLength ); #endif NtfsReinitializeIndexContext( IrpContext, IndexContext ); NtfsReleaseScb( IrpContext, Scb ); } *Count = ReturnCount; // // If we are already returning something, but we got an error, change it // to success to return what we have. Then we may or may not get this error // again anyway when we are called back. This loop is currently not designed // to resume correctly in all cases if there are already items returned. // if (ReturnCount != 0) { Status = STATUS_SUCCESS; } return Status; } NTFSAPI VOID NtOfsFreeReadContext ( IN PREAD_CONTEXT ReadContext ) /*++ Routine Description: This routine is called to free an ReadContext created by NtOfsReadRecords. Arguments: ReadContext - Supplies the context to free. Return Value: STATUS_SUCCESS -- if operation was successful. --*/ { PAGED_CODE(); if (ReadContext->LastReturnedKey.Key != NULL && ReadContext->LastReturnedKey.Key != &ReadContext->SmallKeyBuffer[0]) { NtfsFreePool( ReadContext->LastReturnedKey.Key ); } if (ReadContext->IndexContext.Base != ReadContext->IndexContext.LookupStack) { NtfsFreePool( ReadContext->IndexContext.Base ); } NtfsFreePool( ReadContext ); } NTSTATUS NtfsQueryViewIndex ( IN PIRP_CONTEXT IrpContext, IN PIRP Irp, IN PVCB Vcb, IN PSCB Scb, IN PCCB Ccb ) /*++ Routine Description: This routine performs the query view index operation. It is responsible for either completing or enqueuing the input Irp. Arguments: Irp - Supplies the Irp to process Vcb - Supplies its Vcb Scb - Supplies its Scb Ccb - Supplies its Ccb Return Value: NTSTATUS - The return status for the operation --*/ { NTSTATUS Status = STATUS_SUCCESS; PIO_STACK_LOCATION IrpSp; PUCHAR Buffer; CLONG UserBufferLength; ULONG BaseLength; ULONG SidLength; FILE_INFORMATION_CLASS FileInformationClass; BOOLEAN RestartScan; BOOLEAN ReturnSingleEntry; BOOLEAN GotEntry; BOOLEAN LastPass; BOOLEAN FirstPass = TRUE; ULONG NextEntry; ULONG LastEntry; ULONG VariableLength; PVOID CurrentEntryBuffer = NULL; PINDEX_KEY IndexKey; ULONG IndexKeyLength = 0; PREAD_CONTEXT ReadContext = NULL; PFILE_OBJECTID_INFORMATION ObjIdInfoPtr; PFILE_QUOTA_INFORMATION QuotaInfoPtr = NULL; PQUOTA_USER_DATA QuotaUserData; PFILE_REPARSE_POINT_INFORMATION ReparsePointInfoPtr; BOOLEAN ScbAcquired = FALSE; BOOLEAN CcbAcquired = FALSE; BOOLEAN FirstQueryForThisCcb = FALSE; BOOLEAN IndexKeyAllocated = FALSE; BOOLEAN IndexKeyKeyAllocated = FALSE; BOOLEAN AccessingUserBuffer = FALSE; ULONG ReadRecordBuffer[20]; ULONG VariableBytesToCopy = 0; BOOLEAN AnotherEntryWillFit = TRUE; BOOLEAN AtEndOfIndex = FALSE; PUNICODE_STRING RestartKey = NULL; ULONG BytesRemainingInBuffer; NTSTATUS ReadRecordStatus; ULONG Count; INDEX_ROW IndexRow; // // We need to be certain that the scratch buffer is big enough. // ASSERT( sizeof(ReadRecordBuffer) >= sizeof(FILE_OBJECTID_INFORMATION) ); ASSERT( sizeof(ReadRecordBuffer) >= sizeof(FILE_QUOTA_INFORMATION) ); ASSERT( sizeof(ReadRecordBuffer) >= sizeof(FILE_REPARSE_POINT_INFORMATION) ); ASSERT_IRP_CONTEXT( IrpContext ); ASSERT_IRP( Irp ); ASSERT_VCB( Vcb ); ASSERT_CCB( Ccb ); ASSERT_SCB( Scb ); PAGED_CODE(); // // Get the current Stack location // IrpSp = IoGetCurrentIrpStackLocation( Irp ); DebugTrace( +1, Dbg, ("NtfsQueryViewIndex...\n") ); DebugTrace( 0, Dbg, ("IrpContext = %08lx\n", IrpContext) ); DebugTrace( 0, Dbg, ("Irp = %08lx\n", Irp) ); DebugTrace( 0, Dbg, (" ->Length = %08lx\n", IrpSp->Parameters.QueryDirectory.Length) ); DebugTrace( 0, Dbg, (" ->FileInformationClass = %08lx\n", IrpSp->Parameters.QueryDirectory.FileInformationClass) ); DebugTrace( 0, Dbg, (" ->SystemBuffer = %08lx\n", Irp->AssociatedIrp.SystemBuffer) ); DebugTrace( 0, Dbg, (" ->RestartScan = %08lx\n", FlagOn(IrpSp->Flags, SL_RESTART_SCAN)) ); DebugTrace( 0, Dbg, (" ->ReturnSingleEntry = %08lx\n", FlagOn(IrpSp->Flags, SL_RETURN_SINGLE_ENTRY)) ); DebugTrace( 0, Dbg, ("Vcb = %08lx\n", Vcb) ); DebugTrace( 0, Dbg, ("Scb = %08lx\n", Scb) ); DebugTrace( 0, Dbg, ("Ccb = %08lx\n", Ccb) ); // // Because we probably need to do the I/O anyway we'll reject any request // right now that cannot wait for I/O. We do not want to abort after // processing a few index entries. // if (!FlagOn(IrpContext->State, IRP_CONTEXT_STATE_WAIT)) { DebugTrace( 0, Dbg, ("Automatically enqueue Irp to Fsp\n") ); Status = NtfsPostRequest( IrpContext, Irp ); DebugTrace( -1, Dbg, ("NtfsQueryViewIndex -> %08lx\n", Status) ); return Status; } // // Reference our input parameters to make things easier // UserBufferLength = IrpSp->Parameters.QueryDirectory.Length; FileInformationClass = IrpSp->Parameters.QueryDirectory.FileInformationClass; RestartKey = IrpSp->Parameters.QueryDirectory.FileName; // // Look in the Ccb to see the type of search. // RestartScan = BooleanFlagOn(IrpSp->Flags, SL_RESTART_SCAN); ReturnSingleEntry = BooleanFlagOn(IrpSp->Flags, SL_RETURN_SINGLE_ENTRY); // // Determine the size of the constant part of the structure and make sure the Scb // and info class are in agreement. There may be some security implications in // letting a user treat, say, the reparse index as the object id index. // switch (FileInformationClass) { case FileObjectIdInformation: BaseLength = sizeof( FILE_OBJECTID_INFORMATION ); IndexKeyLength = OBJECT_ID_KEY_LENGTH; if (Scb != Vcb->ObjectIdTableScb) { Status = STATUS_INVALID_INFO_CLASS; } break; case FileQuotaInformation: BaseLength = sizeof( FILE_QUOTA_INFORMATION ); IndexKeyLength = sizeof( ULONG ); if (Scb != Vcb->QuotaTableScb) { Status = STATUS_INVALID_INFO_CLASS; } break; case FileReparsePointInformation: BaseLength = sizeof( FILE_REPARSE_POINT_INFORMATION ); IndexKeyLength = sizeof( REPARSE_INDEX_KEY ); if (Scb != Vcb->ReparsePointTableScb) { Status = STATUS_INVALID_INFO_CLASS; } break; default: Status = STATUS_INVALID_INFO_CLASS; break; } if (Status != STATUS_SUCCESS) { NtfsCompleteRequest( IrpContext, Irp, Status ); DebugTrace( -1, Dbg, ("NtfsQueryViewIndex -> %08lx\n", Status) ); return Status; } try { // // We only allow one active request in this handle at a time. If this is // not a synchronous request then wait on the handle. // if (!FlagOn( IrpSp->FileObject->Flags, FO_SYNCHRONOUS_IO )) { EOF_WAIT_BLOCK WaitBlock; NtfsAcquireIndexCcb( Scb, Ccb, &WaitBlock ); CcbAcquired = TRUE; } // // Initialize the value we use to start the index enumeration. Read carefully, // we're working with both IndexKey and IndexKey->Key here. // IndexKey = NtfsAllocatePool( PagedPool, sizeof(INDEX_KEY) ); IndexKeyAllocated = TRUE; IndexKey->KeyLength = IndexKeyLength; IndexKey->Key = NtfsAllocatePool( PagedPool, IndexKeyLength ); IndexKeyKeyAllocated = TRUE; // // When we first come into this function, there are a few interesting // cases we need to consider to get everything initialized correctly. // // 1 We were called with some value from which to (re)start the enumeration, // i.e. the caller wants to start somewhere in the middle of the index. // // 2 This is the first time we've been called to enumerate this index with // this Ccb, in which case we want to start from the beginning. This // is substantially similar to the case where the caller has enumerated // this index before, but wishes to restart the scan from the beginning of // the index. // // 3 This is _not_ the first time we've been called to enumerate this index // with this Ccb, and no restart key was specified, and the caller does // not wish to restart the scan. In this case, we need to pick up where // the last call left off. // if (RestartKey != NULL) { DebugTrace( 0, Dbg, ("Restart key NOT null (case 1)\n") ); // // If we have a leftover query buffer from a previous call, free // it, since we're no longer interested in where it left us. // if (Ccb->QueryBuffer != NULL) { ASSERT(FlagOn( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED )); NtOfsFreeReadContext( Ccb->QueryBuffer ); Ccb->QueryBuffer = NULL; ClearFlag( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED | CCB_FLAG_LAST_INDEX_ROW_RETURNED ); } // // Store the initial value from which to start the enumeration, // being careful not to write beyond the size of our allocated buffer. // if (RestartKey->Length > IndexKeyLength) { NtfsFreePool( IndexKey->Key ); IndexKeyKeyAllocated = FALSE; IndexKey->Key = NtfsAllocatePool( PagedPool, RestartKey->Length ); IndexKeyKeyAllocated = TRUE; } // // Copy the key, and store the length. // RtlCopyMemory( IndexKey->Key, RestartKey->Buffer, RestartKey->Length ); IndexKey->KeyLength = IndexKeyLength = RestartKey->Length; } else if (RestartScan || (Ccb->QueryBuffer == NULL)) { DebugTrace( 0, Dbg, ("RestartScan || Qb null (case 2)") ); // // The restart scan case is similar to the case where we're called with a // RestartKey in that we want to deallocate any leftover info in the Ccb. // The only difference is that we don't have a key from which to restart // so we just set the key back to the appropriate starting value. If // the Ccb has no query buffer, then this is our first enumeration call // since the handle was opened, and we need to start from scratch. // if (Ccb->QueryBuffer != NULL) { DebugTrace( 0, Dbg, ("Qb NOT null\n") ); ASSERT(FlagOn( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED )); NtOfsFreeReadContext( Ccb->QueryBuffer ); Ccb->QueryBuffer = NULL; ClearFlag( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED | CCB_FLAG_LAST_INDEX_ROW_RETURNED ); } else { DebugTrace( 0, Dbg, ("Qb null\n") ); FirstQueryForThisCcb = TRUE; } if (FileInformationClass == FileQuotaInformation) { // // In the quota case, we have some special requirements for the fist key, // so we want to init it to handle the case where we haven't been called // with a restart key. // *((PULONG) IndexKey->Key) = QUOTA_FISRT_USER_ID; } else { RtlZeroMemory( IndexKey->Key, IndexKeyLength ); } } else { DebugTrace( 0, Dbg, ("Ccb->QueryBuffer NOT null (case 3)\n") ); ASSERT(Ccb->QueryBuffer != NULL); ASSERT(FlagOn( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED )); // // We have a leftover query buffer in the Ccb, and we were _not_ // called with a restart key, and we're not restarting the enumeration. // Let's just pick up the enumeration where the Ccb's buffer tells us // the last call left off. // ReadContext = Ccb->QueryBuffer; // // If the previous pass set the keylength to 0, it must have hit the // end of the index. Check the CCB_FLAG_LAST_INDEX_ROW_RETURNED bit // to see if we already made our special last pass through here to // return the last row. // if (ReadContext->LastReturnedKey.KeyLength == 0) { DebugTrace( 0, Dbg, ("LastReturnedKey had 0 length\n") ); if (FlagOn(Ccb->Flags, CCB_FLAG_LAST_INDEX_ROW_RETURNED)) { // // We're at the end of the index, and the last entry has already // been returned to our caller. We're all done now. // try_return( Status = STATUS_NO_MORE_FILES ); } else { // // We're at the end of the index, but we have not yet returned the // last entry to our caller. We can't break out yet. // AtEndOfIndex = TRUE; // // Remember that we are returning the last entry now. // SetFlag( Ccb->Flags, CCB_FLAG_LAST_INDEX_ROW_RETURNED ); // // We need to set this to a nonzero value so NtOfsReadRecords // will honor our request to read the last record. // if (IndexKeyLength > ReadContext->AllocatedKeyLength) { ReadContext->LastReturnedKey.KeyLength = ReadContext->AllocatedKeyLength; } else { ReadContext->LastReturnedKey.KeyLength = IndexKeyLength; } } } else if (ReadContext->LastReturnedKey.KeyLength > IndexKeyLength) { // // There's not enough room to store the initial value from which to // start the enumeration. Free the buffer and get a bigger one. // NtfsFreePool( IndexKey->Key ); IndexKeyKeyAllocated = FALSE; IndexKey->Key = NtfsAllocatePool( PagedPool, ReadContext->LastReturnedKey.KeyLength ); IndexKeyKeyAllocated = TRUE; } // // Make sure we're either using the small key buffer, or we've allocated // a buffer that's big enough. // ASSERT( (ReadContext->LastReturnedKey.Key == &ReadContext->SmallKeyBuffer[0]) || (ReadContext->LastReturnedKey.KeyLength <= ReadContext->AllocatedKeyLength) ); // // Store the initial value from which to start the enumeration. // RtlCopyMemory( IndexKey->Key, ReadContext->LastReturnedKey.Key, ReadContext->LastReturnedKey.KeyLength ); IndexKeyLength = ReadContext->LastReturnedKey.KeyLength; } Irp->IoStatus.Information = 0; // // Acquire shared access to the Scb. // NtfsAcquireSharedScb( IrpContext, Scb ); ScbAcquired = TRUE; // // If the volume is no longer mounted, we should fail this // request. Since we have the Scb shared now, we know that // a dismount request can't sneak in. // if (FlagOn( Scb->ScbState, SCB_STATE_VOLUME_DISMOUNTED )) { try_return( Status = STATUS_VOLUME_DISMOUNTED ); } try { // // If we are in the Fsp now because we had to wait earlier, // we must map the user buffer, otherwise we can use the // user's buffer directly. // AccessingUserBuffer = TRUE; Buffer = NtfsMapUserBuffer( Irp, NormalPagePriority ); AccessingUserBuffer = FALSE; // // At this point we are about to enter our query loop. We have // already decided if we need to call restart or continue when we // go after an index entry. The variables LastEntry and NextEntry are // used to index into the user buffer. LastEntry is the last entry // we added to the user buffer, and NextEntry is the current // one we're working on. // LastEntry = 0; NextEntry = 0; while (TRUE) { DebugTrace( 0, Dbg, ("Top of Loop\n") ); DebugTrace( 0, Dbg, ("LastEntry = %08lx\n", LastEntry) ); DebugTrace( 0, Dbg, ("NextEntry = %08lx\n", NextEntry) ); // // Check to see if we should quit the loop because we are only // returning a single entry. We actually want to spin around // the loop top twice so that our enumeration has us left off // at the last entry we didn't return. // LastPass = (ReturnSingleEntry && !FirstPass); // // Be sure to pessimistically reinitialize these locals each // time through the loop. // GotEntry = FALSE; IndexRow.KeyPart.KeyLength = 0; IndexRow.DataPart.DataLength = 0; Count = 1; // // On the first pass for this IrpContext, we MUST take this code path // with a null readcontext and a non-null IndexKey. See the comment // where NtOfsReadRecords is implemented. // if (FirstPass) { DebugTrace( 0, Dbg, ("First pass\n") ); ReadContext = NULL; ReadRecordStatus = NtOfsReadRecords( IrpContext, Scb, &ReadContext, IndexKey, NtOfsMatchAll, IndexKey, &Count, &IndexRow, sizeof( ReadRecordBuffer ), ReadRecordBuffer ); // // We want to store the new ReadContext in the Ccb. Free // anything leftover in the Ccb first. // if (Ccb->QueryBuffer != NULL) { ASSERT(FlagOn( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED )); NtOfsFreeReadContext( Ccb->QueryBuffer ); } Ccb->QueryBuffer = ReadContext; SetFlag( Ccb->Flags, CCB_FLAG_READ_CONTEXT_ALLOCATED ); GotEntry = (NT_SUCCESS( ReadRecordStatus ) && (IndexRow.KeyPart.KeyLength != 0)); } else if ((!AtEndOfIndex)) { DebugTrace( 0, Dbg, ("Nth pass\n") ); // // We don't want to do this if the previous trip through the loop // took us to the end of the index. // ReadContext = Ccb->QueryBuffer; ASSERT(ReadContext != NULL); // // Lookup the next index entry and set ourselves // up for subsequent iterations through the loop. // ReadRecordStatus = NtOfsReadRecords( IrpContext, Scb, &ReadContext, NULL, NtOfsMatchAll, NULL, &Count, &IndexRow, sizeof( ReadRecordBuffer ), ReadRecordBuffer ); GotEntry = (NT_SUCCESS( ReadRecordStatus ) && (IndexRow.KeyPart.KeyLength != 0)); } // // If we only want to do the top part of the loop this time, get out here. // For more info, see the comment above where we set LastPass. Basically, // if we're just advancing our index pointer in the return single case, // if we don't have room to return another entry in the return multiple // case, or if we came in pointing to the end of the index and we've // already made one pass to return that entry to the caller, we should // get out now. // if (LastPass || !AnotherEntryWillFit || (AtEndOfIndex && !FirstPass)) { DebugTrace( 0, Dbg, ("LastPass = %08lx\n", LastPass) ); DebugTrace( 0, Dbg, ("AnotherEntryWillFit = %08lx\n", AnotherEntryWillFit) ); DebugTrace( 0, Dbg, ("...breaking out\n") ); if ((FileInformationClass == FileQuotaInformation) && (QuotaInfoPtr != NULL)) { // // In the quota enumeration case, we need to zero this field // to indicate the end of the list. // QuotaInfoPtr->NextEntryOffset = 0; } break; } // // Now check to see if we actually got another index entry. If // we didn't then we also need to check if we never got any // or if we just ran out. If we just ran out then we break out // of the main loop and finish the Irp after the loop. // if (!GotEntry) { DebugTrace( 0, Dbg, ("GotEntry is FALSE\n") ); if (!FirstPass) { if (FirstQueryForThisCcb) { try_return( Status = STATUS_NO_SUCH_FILE ); } try_return( Status = STATUS_NO_MORE_FILES ); } break; } // // If the previous pass set the keylength to 0, it must have hit the // end of the index. // if (ReadContext->LastReturnedKey.KeyLength == 0) { DebugTrace( 0, Dbg, ("LastReturnedKey had 0 length (mid-loop)\n") ); // // We're at the end of the index, but we have not yet returned the // last entry to our caller. We can't break out yet. // AtEndOfIndex = TRUE; // // Remember that we are returning the last entry now. // SetFlag( Ccb->Flags, CCB_FLAG_LAST_INDEX_ROW_RETURNED ); } // // Here are the rules concerning filling up the buffer: // // 1. The Io system guarantees that there will always be // enough room for at least one base record. // // 2. If the full first record (including variable length data) // cannot fit, as much of the data as possible is copied // and STATUS_BUFFER_OVERFLOW is returned. // // 3. If a subsequent record cannot completely fit into the // buffer, none of it (as in 0 bytes) is copied, and // STATUS_SUCCESS is returned. A subsequent query will // pick up with this record. // BytesRemainingInBuffer = UserBufferLength - NextEntry; if ( (NextEntry != 0) && ( (BaseLength + VariableLength > BytesRemainingInBuffer) || (UserBufferLength < NextEntry) ) ) { DebugTrace( 0, Dbg, ("Next entry won't fit\n") ); try_return( Status = STATUS_SUCCESS ); } ASSERT( BytesRemainingInBuffer >= BaseLength ); // // Zero the base part of the structure. // AccessingUserBuffer = TRUE; RtlZeroMemory( &Buffer[NextEntry], BaseLength ); // // Now we have an entry to return to our caller. We'll // case on the type of information requested and fill up the // user buffer if everything fits. // switch (FileInformationClass) { case FileObjectIdInformation: ObjIdInfoPtr = (PFILE_OBJECTID_INFORMATION) (&Buffer[NextEntry]); if (IndexRow.DataPart.DataLength == sizeof(NTFS_OBJECTID_INFORMATION)) { RtlCopyMemory( &ObjIdInfoPtr->FileReference, &(((NTFS_OBJECTID_INFORMATION *) IndexRow.DataPart.Data)->FileSystemReference), sizeof( LONGLONG ) ); RtlCopyMemory( &ObjIdInfoPtr->ExtendedInfo, ((NTFS_OBJECTID_INFORMATION *) IndexRow.DataPart.Data)->ExtendedInfo, OBJECT_ID_EXT_INFO_LENGTH ); } else { ASSERTMSG( "Bad objectid index datalength", FALSE ); SetFlag( Vcb->ObjectIdState, VCB_OBJECT_ID_CORRUPT ); try_return( STATUS_NO_MORE_FILES ); } if (IndexRow.KeyPart.KeyLength == IndexKeyLength) { RtlCopyMemory( &ObjIdInfoPtr->ObjectId, IndexRow.KeyPart.Key, IndexRow.KeyPart.KeyLength ); } else { ASSERTMSG( "Bad objectid index keylength", FALSE ); SetFlag( Vcb->ObjectIdState, VCB_OBJECT_ID_CORRUPT ); try_return( STATUS_NO_MORE_FILES ); } // // Object Ids have no variable length data, so we can skip // over some of the tricky code below. // VariableLength = 0; break; case FileQuotaInformation: QuotaInfoPtr = (PFILE_QUOTA_INFORMATION) (&Buffer[NextEntry]); QuotaUserData = (PQUOTA_USER_DATA) IndexRow.DataPart.Data; // // Skip this entry if it has been deleted. // if (FlagOn( QuotaUserData->QuotaFlags, QUOTA_FLAG_ID_DELETED )) { continue; } SidLength = IndexRow.DataPart.DataLength - SIZEOF_QUOTA_USER_DATA; QuotaInfoPtr->ChangeTime.QuadPart = QuotaUserData->QuotaChangeTime; QuotaInfoPtr->QuotaUsed.QuadPart = QuotaUserData->QuotaUsed; QuotaInfoPtr->QuotaThreshold.QuadPart = QuotaUserData->QuotaThreshold; QuotaInfoPtr->QuotaLimit.QuadPart = QuotaUserData->QuotaLimit; QuotaInfoPtr->SidLength = SidLength; RtlCopyMemory( &QuotaInfoPtr->Sid, &QuotaUserData->QuotaSid, SidLength ); QuotaInfoPtr->NextEntryOffset = QuadAlign( SidLength + SIZEOF_QUOTA_USER_DATA ); VariableLength = QuotaInfoPtr->SidLength; break; case FileReparsePointInformation: ReparsePointInfoPtr = (PFILE_REPARSE_POINT_INFORMATION) (&Buffer[NextEntry]); if (IndexRow.KeyPart.KeyLength == sizeof(REPARSE_INDEX_KEY)) { ReparsePointInfoPtr->Tag = ((PREPARSE_INDEX_KEY) IndexRow.KeyPart.Key)->FileReparseTag; ReparsePointInfoPtr->FileReference = ((PREPARSE_INDEX_KEY) IndexRow.KeyPart.Key)->FileId.QuadPart; } else { ASSERTMSG( "Bad reparse point index key length", FALSE ); } // // Reparse points have no variable length data, so we can skip // over some of the tricky code below. // VariableLength = 0; break; default: try_return( Status = STATUS_INVALID_INFO_CLASS ); } if (VariableLength != 0) { // // Compute how many bytes we can copy. This should only be less // than the variable length if we are only returning a single // entry. // if (BytesRemainingInBuffer >= BaseLength + VariableLength) { VariableBytesToCopy = VariableLength; } else { VariableBytesToCopy = BytesRemainingInBuffer - BaseLength; if (FileInformationClass == FileQuotaInformation) { // // In the quota enumeration case, we need to zero this field // to indicate the end of the list. // QuotaInfoPtr->NextEntryOffset = 0; } Status = STATUS_BUFFER_OVERFLOW; } } else { VariableBytesToCopy = 0; } // // Set up the previous next entry offset. // if (FileInformationClass == FileQuotaInformation) { *((PULONG)(&Buffer[LastEntry])) = NextEntry - LastEntry; } AccessingUserBuffer = FALSE; // // And indicate how much of the user buffer we have currently // used up. We must compute this value before we long align // ourselves for the next entry. This is the point where we // quad-align the length of the previous entry. // Irp->IoStatus.Information = QuadAlign( Irp->IoStatus.Information) + BaseLength + VariableBytesToCopy; // // If we weren't able to copy the whole index entry, then we bail here. // if ( !NT_SUCCESS( Status ) ) { DebugTrace( 0, Dbg, ("Couldn't copy the whole index entry, exiting\n") ); try_return( Status ); } // // Set ourselves up for the next iteration. // LastEntry = NextEntry; NextEntry += (ULONG)QuadAlign( BaseLength + VariableBytesToCopy ); FirstPass = FALSE; // // Determine whether we should be able to fit another entry // in the user's buffer after this one. // AnotherEntryWillFit = ((NextEntry + BaseLength) <= UserBufferLength); } } except(EXCEPTION_EXECUTE_HANDLER) { Status = GetExceptionCode(); DebugTrace( -1, Dbg, ("NtfsQueryViewIndex raising %08lx\n", Status) ); if (FsRtlIsNtstatusExpected( Status )) { NtfsRaiseStatus( IrpContext, Status, NULL, NULL ); } else { ExRaiseStatus( AccessingUserBuffer ? STATUS_INVALID_USER_BUFFER : Status ); } } // // At this point we've successfully filled up some of the buffer so // now is the time to set our status to success. // Status = STATUS_SUCCESS; try_exit: // // Abort transaction on error by raising. // NtfsCleanupTransaction( IrpContext, Status, FALSE ); // // Set the last access flag in the Fcb if the caller // didn't set it explicitly. // if (!FlagOn( Ccb->Flags, CCB_FLAG_USER_SET_LAST_ACCESS_TIME ) && !FlagOn( NtfsData.Flags, NTFS_FLAGS_DISABLE_LAST_ACCESS )) { NtfsGetCurrentTime( IrpContext, Scb->Fcb->CurrentLastAccess ); SetFlag( Scb->Fcb->InfoFlags, FCB_INFO_UPDATE_LAST_ACCESS ); } } finally { DebugUnwind( NtfsQueryViewIndex ); if (ScbAcquired) { NtfsReleaseScb( IrpContext, Scb ); } NtfsCleanupAfterEnumeration( IrpContext, Ccb ); if (CcbAcquired) { NtfsReleaseIndexCcb( Scb, Ccb ); } if (IndexKeyAllocated) { if (IndexKeyKeyAllocated) { NtfsFreePool( IndexKey->Key ); } NtfsFreePool( IndexKey ); } if (!AbnormalTermination()) { NtfsCompleteRequest( IrpContext, Irp, Status ); } } // // And return to our caller // DebugTrace( -1, Dbg, ("NtfsQueryViewIndex -> %08lx\n", Status) ); return Status; } /*++ Routine Descriptions: Standard collation routines for creating simple indices. Arguments: Key1 - First key to compare. Key2 - Second key to compare. CollationData - Optional data to support the collation. Return Value: LessThan, EqualTo, or Greater than, for how Key1 compares with Key2. --*/ FSRTL_COMPARISON_RESULT NtOfsCollateUlong ( IN PINDEX_KEY Key1, IN PINDEX_KEY Key2, IN PVOID CollationData ) { ULONG u1, u2; UNREFERENCED_PARAMETER(CollationData); ASSERT( Key1->KeyLength == 4 ); ASSERT( Key2->KeyLength == 4 ); u1 = *(PULONG)Key1->Key; u2 = *(PULONG)Key2->Key; if (u1 > u2) { return GreaterThan; } else if (u1 < u2) { return LessThan; } return EqualTo; } FSRTL_COMPARISON_RESULT NtOfsCollateUlongs ( IN PINDEX_KEY Key1, IN PINDEX_KEY Key2, IN PVOID CollationData ) { PULONG pu1, pu2; ULONG count; FSRTL_COMPARISON_RESULT result = EqualTo; UNREFERENCED_PARAMETER(CollationData); ASSERT( (Key1->KeyLength & 3) == 0 ); ASSERT( (Key2->KeyLength & 3) == 0 ); count = Key1->KeyLength; if (count != Key2->KeyLength) { result = LessThan; if (count > Key2->KeyLength) { count = Key2->KeyLength; result = GreaterThan; } } pu1 = (PULONG)Key1->Key; pu2 = (PULONG)Key2->Key; while (count > 0) { if (*pu1 > *pu2) { return GreaterThan; } else if (*(pu1++) < *(pu2++)) { return LessThan; } count -= 4; } return result; } FSRTL_COMPARISON_RESULT NtOfsCollateSid ( IN PINDEX_KEY Key1, IN PINDEX_KEY Key2, IN PVOID CollationData ) { LONG Compare; PAGED_CODE( ); UNREFERENCED_PARAMETER(CollationData); // // The length of a valid SID is imbedded in the data // so the function will mismatch be for the data runs out. // Compare = memcmp( Key1->Key, Key2->Key, Key1->KeyLength ); if (Compare > 0) { return GreaterThan; } else if (Compare < 0) { return LessThan; } return EqualTo; } FSRTL_COMPARISON_RESULT NtOfsCollateUnicode ( IN PINDEX_KEY Key1, IN PINDEX_KEY Key2, IN PVOID CollationData ) { UNICODE_STRING String1, String2; PAGED_CODE(); // // Build the unicode strings and call namesup. // String1.Length = String1.MaximumLength = (USHORT)Key1->KeyLength; String1.Buffer = Key1->Key; String2.Length = String2.MaximumLength = (USHORT)Key2->KeyLength; String2.Buffer = Key2->Key; return NtfsCollateNames( ((PUPCASE_TABLE_AND_KEY)CollationData)->UpcaseTable, ((PUPCASE_TABLE_AND_KEY)CollationData)->UpcaseTableSize, &String1, &String2, LessThan, TRUE ); } /*++ Routine Descriptions: Standard match routines for find / enumerate in simple indices. Arguments: IndexRow - Row to check for a match. MatchData - Optional data for determining a match. Return Value: STATUS_SUCCESS if the IndexRow matches STATUS_NO_MATCH if the IndexRow does not match, but the enumeration should continue STATUS_NO_MORE_MATCHES if the IndexRow does not match, and the enumeration should terminate --*/ NTSTATUS NtOfsMatchAll ( IN PINDEX_ROW IndexRow, IN OUT PVOID MatchData ) { UNREFERENCED_PARAMETER(IndexRow); UNREFERENCED_PARAMETER(MatchData); return STATUS_SUCCESS; } NTSTATUS NtOfsMatchUlongExact ( IN PINDEX_ROW IndexRow, IN OUT PVOID MatchData ) { ULONG u1, u2; ASSERT( IndexRow->KeyPart.KeyLength == 4 ); u1 = *(PULONG)IndexRow->KeyPart.Key; u2 = *(PULONG)((PINDEX_KEY)MatchData)->Key; if (u1 == u2) { return STATUS_SUCCESS; } else if (u1 < u2) { return STATUS_NO_MATCH; } return STATUS_NO_MORE_MATCHES; } NTSTATUS NtOfsMatchUlongsExact ( IN PINDEX_ROW IndexRow, IN OUT PVOID MatchData ) { PULONG pu1, pu2; ULONG count; NTSTATUS status = STATUS_SUCCESS; ASSERT( (((PINDEX_KEY)MatchData)->KeyLength & 3) == 0 ); ASSERT( (IndexRow->KeyPart.KeyLength & 3) == 0 ); count = ((PINDEX_KEY)MatchData)->KeyLength; if (count != IndexRow->KeyPart.KeyLength) { status = STATUS_NO_MORE_MATCHES; if (count > IndexRow->KeyPart.KeyLength) { count = IndexRow->KeyPart.KeyLength; status = STATUS_NO_MATCH; } } pu1 = (PULONG)((PINDEX_KEY)MatchData)->Key; pu2 = (PULONG)IndexRow->KeyPart.Key; while (count > 0) { if (*pu1 > *pu2) { return STATUS_NO_MATCH; } else if (*(pu1++) < *(pu2++)) { return STATUS_NO_MORE_MATCHES; } count -= 4; } return status; } NTSTATUS NtOfsMatchUnicodeExpression ( IN PINDEX_ROW IndexRow, IN OUT PVOID MatchData ) { UNICODE_STRING MatchString, IndexString; FSRTL_COMPARISON_RESULT BlindResult; PUPCASE_TABLE_AND_KEY UpcaseTableAndKey = (PUPCASE_TABLE_AND_KEY)MatchData; PAGED_CODE(); // // Build the unicode strings and call namesup. // MatchString.Length = MatchString.MaximumLength = (USHORT)UpcaseTableAndKey->Key.KeyLength; MatchString.Buffer = UpcaseTableAndKey->Key.Key; IndexString.Length = IndexString.MaximumLength = (USHORT)IndexRow->KeyPart.KeyLength; IndexString.Buffer = IndexRow->KeyPart.Key; if (NtfsIsNameInExpression( UpcaseTableAndKey->UpcaseTable, &MatchString, &IndexString, TRUE )) { return STATUS_SUCCESS; } else if ((BlindResult = NtfsCollateNames(UpcaseTableAndKey->UpcaseTable, UpcaseTableAndKey->UpcaseTableSize, &MatchString, &IndexString, GreaterThan, TRUE)) != LessThan) { return STATUS_NO_MATCH; } else { return STATUS_NO_MORE_MATCHES; } } NTSTATUS NtOfsMatchUnicodeString ( IN PINDEX_ROW IndexRow, IN OUT PVOID MatchData ) { UNICODE_STRING MatchString, IndexString; FSRTL_COMPARISON_RESULT BlindResult; PUPCASE_TABLE_AND_KEY UpcaseTableAndKey = (PUPCASE_TABLE_AND_KEY)MatchData; PAGED_CODE(); // // Build the unicode strings and call namesup. // MatchString.Length = MatchString.MaximumLength = (USHORT)UpcaseTableAndKey->Key.KeyLength; MatchString.Buffer = UpcaseTableAndKey->Key.Key; IndexString.Length = IndexString.MaximumLength = (USHORT)IndexRow->KeyPart.KeyLength; IndexString.Buffer = IndexRow->KeyPart.Key; if (NtfsAreNamesEqual( UpcaseTableAndKey->UpcaseTable, &MatchString, &IndexString, TRUE )) { return STATUS_SUCCESS; } else if ((BlindResult = NtfsCollateNames(UpcaseTableAndKey->UpcaseTable, UpcaseTableAndKey->UpcaseTableSize, &MatchString, &IndexString, GreaterThan, TRUE)) != LessThan) { return STATUS_NO_MATCH; } else { return STATUS_NO_MORE_MATCHES; } } #ifdef TOMM VOID NtOfsIndexTest ( PIRP_CONTEXT IrpContext, PFCB TestFcb ) { PSCB AdScb; NTSTATUS Status; ULONG i; MAP_HANDLE MapHandle; ULONG Count; UPCASE_TABLE_AND_KEY UpcaseTableAndKey; QUICK_INDEX_HINT QuickHint; INDEX_KEY IndexKey; INDEX_ROW IndexRow[6]; UCHAR Buffer[6*160]; PREAD_CONTEXT ReadContext = NULL; UNICODE_STRING IndexName = CONSTANT_UNICODE_STRING( L"$Test" ); USHORT MaxKey = MAXUSHORT; USHORT MinKey = 0; DbgPrint("NtOfs Make NtOfsDoIndexTest FALSE to suppress test\n"); DbgPrint("NtOfs Make NtOfsLeaveTestIndex TRUE to leave test index\n"); DbgBreakPoint(); if (!NtOfsDoIndexTest) { return; } NtOfsDoIndexTest = FALSE; UpcaseTableAndKey.UpcaseTable = TestFcb->Vcb->UpcaseTable; UpcaseTableAndKey.UpcaseTableSize = TestFcb->Vcb->UpcaseTableSize; UpcaseTableAndKey.Key.Key = NULL; UpcaseTableAndKey.Key.KeyLength = 0; // // Create Test Index // DbgPrint("NtOfs creating test index\n"); NtOfsCreateIndex( IrpContext, TestFcb, IndexName, CREATE_NEW, 0, COLLATION_NTOFS_ULONG, &NtOfsCollateUnicode, &UpcaseTableAndKey, &AdScb ); DbgPrint("NtOfs created Test Index Scb %08lx\n", AdScb); // // Access empty index // DbgPrint("NtOfs lookup last in empty index\n"); IndexKey.Key = &MaxKey; IndexKey.KeyLength = sizeof(MaxKey); Status = NtOfsFindLastRecord( IrpContext, AdScb, &IndexKey, &IndexRow[0], &MapHandle ); ASSERT(!NT_SUCCESS(Status)); // // Add some keys! // DbgPrint("NtOfs adding keys to index\n"); for (i = 0; i < $EA/0x10; i++) { IndexRow[0].KeyPart.Key = &NtfsAttributeDefinitions[i].AttributeName; IndexRow[0].KeyPart.KeyLength = 0x80; IndexRow[0].DataPart.Data = (PCHAR)IndexRow[0].KeyPart.Key + 0x80; IndexRow[0].DataPart.DataLength = sizeof(ATTRIBUTE_DEFINITION_COLUMNS) - 0x84; NtOfsAddRecords( IrpContext, AdScb, 1, &IndexRow[0], 0 ); } // // Now find the last key // DbgPrint("NtOfs checkin last key in index\n"); IndexKey.Key = &MaxKey; IndexKey.KeyLength = sizeof(MaxKey); Status = NtOfsFindLastRecord( IrpContext, AdScb, &IndexKey, &IndexRow[0], &MapHandle ); ASSERT(NT_SUCCESS(Status)); ASSERT(RtlCompareMemory(IndexRow[0].KeyPart.Key, L"$VOLUME_NAME", sizeof(L"$VOLUME_NAME") - sizeof( WCHAR )) == (sizeof(L"$VOLUME_NAME") - sizeof( WCHAR ))); NtOfsReleaseMap( IrpContext, &MapHandle ); // // See if they are all there. // DbgPrint("NtOfs looking up all keys in index\n"); for (i = 0; i < $EA/0x10; i++) { IndexKey.Key = &NtfsAttributeDefinitions[i].AttributeName; IndexKey.KeyLength = 0x80; Status = NtOfsFindRecord( IrpContext, AdScb, &IndexKey, &IndexRow[0], &MapHandle, NULL ); if (!NT_SUCCESS(Status)) { DbgPrint("NtOfsIterationFailure with i = %08lx, Status = %08lx\n", i, Status); } NtOfsReleaseMap( IrpContext, &MapHandle ); } // // Now enumerate the entire index // IndexKey.Key = &MinKey; IndexKey.KeyLength = sizeof(MinKey); Count = 6; DbgPrint("NtOfs enumerating index:\n\n"); while (NT_SUCCESS(Status = NtOfsReadRecords( IrpContext, AdScb, &ReadContext, (ReadContext == NULL) ? &IndexKey : NULL, &NtOfsMatchAll, NULL, &Count, IndexRow, sizeof(Buffer), Buffer ))) { for (i = 0; i < Count; i++) { DbgPrint( "IndexKey = %ws, AttributeTypeCode = %lx\n", IndexRow[i].KeyPart.Key, *(PULONG)IndexRow[i].DataPart.Data ); } DbgPrint( "\n" ); } NtOfsFreeReadContext( ReadContext ); ReadContext = NULL; // // Loop to update all records. // DbgPrint("NtOfs updating up all keys in index\n"); for (i = 0; i < $EA/0x10; i++) { IndexKey.Key = &NtfsAttributeDefinitions[i].AttributeName; IndexKey.KeyLength = 0x80; RtlZeroMemory( &QuickHint, sizeof(QUICK_INDEX_HINT) ); NtOfsFindRecord( IrpContext, AdScb, &IndexKey, &IndexRow[0], &MapHandle, &QuickHint ); // // Copy and update the data. // RtlCopyMemory( Buffer, IndexRow[0].DataPart.Data, IndexRow[0].DataPart.DataLength ); *(PULONG)Buffer += 0x100; IndexRow[0].DataPart.Data = Buffer; // // Perform update with all valid combinations of hint and map handle. // NtOfsUpdateRecord( IrpContext, AdScb, 1, &IndexRow[0], (i <= $FILE_NAME/0x10) ? NULL : &QuickHint, (i < $INDEX_ROOT/0x10) ? NULL : &MapHandle ); NtOfsReleaseMap( IrpContext, &MapHandle ); } // // Now enumerate the entire index again to see the updates. // IndexKey.Key = &MinKey; IndexKey.KeyLength = sizeof(MinKey); Count = 6; DbgPrint("NtOfs enumerating index after updates:\n\n"); while (NT_SUCCESS(Status = NtOfsReadRecords( IrpContext, AdScb, &ReadContext, (ReadContext == NULL) ? &IndexKey : NULL, &NtOfsMatchAll, NULL, &Count, IndexRow, sizeof(Buffer), Buffer ))) { for (i = 0; i < Count; i++) { DbgPrint( "IndexKey = %ws, AttributeTypeCode = %lx\n", IndexRow[i].KeyPart.Key, *(PULONG)IndexRow[i].DataPart.Data ); } DbgPrint( "\n" ); } NtOfsFreeReadContext( ReadContext ); ReadContext = NULL; // // Now delete the keys // if (!NtOfsLeaveTestIndex) { DbgPrint("NtOfs deleting all keys in index:\n\n"); for (i = 0; i < $EA/0x10; i++) { IndexKey.Key = &NtfsAttributeDefinitions[i].AttributeName; IndexKey.KeyLength = 0x80; NtOfsDeleteRecords( IrpContext, AdScb, 1, &IndexKey ); } // // Access empty index // DbgPrint("NtOfs lookup last key in empty index:\n\n"); IndexKey.Key = &MaxKey; IndexKey.KeyLength = sizeof(MaxKey); Status = NtOfsFindLastRecord( IrpContext, AdScb, &IndexKey, &IndexRow[0], &MapHandle ); ASSERT(!NT_SUCCESS(Status)); DbgPrint("NtOfs deleting index:\n"); NtOfsDeleteIndex( IrpContext, TestFcb, AdScb ); } DbgPrint("NtOfs closing index:\n"); NtOfsCloseIndex( IrpContext, AdScb ); DbgPrint("NtOfs test complete!\n\n"); return; // // Make sure these at least compile until we have some real callers. // { MAP_HANDLE M; PVOID B; LONGLONG O; ULONG L; LSN Lsn; NtOfsInitializeMapHandle( &M ); NtOfsMapAttribute( IrpContext, AdScb, O, L, &B, &M ); NtOfsPreparePinWrite( IrpContext, AdScb, O, L, &B, &M ); NtOfsPinRead( IrpContext, AdScb, O, L, &M ); NtOfsDirty( IrpContext, &M, &Lsn ); NtOfsReleaseMap( IrpContext, &M ); NtOfsPutData( IrpContext, AdScb, O, L, &B ); } } #endif TOMM