Intravision/reactos
1
0
Fork 0
mirror of https://github.com/reactos/reactos.git synced 2024-11-01 12:26:32 +00:00
Code Issues Packages Projects Releases Wiki Activity
reactos/drivers/filesystems/fastfat/close.c

1290 lines
33 KiB
C
Raw Blame History

/*++
Copyright (c) 1989-2000 Microsoft Corporation
Module Name:
Close.c
Abstract:
This module implements the File Close routine for Fat called by the
dispatch driver.
--*/
#include "fatprocs.h"
//
// The Bug check file id for this module
//
#define BugCheckFileId (FAT_BUG_CHECK_CLOSE)
//
// The local debug trace level
//
#define Dbg (DEBUG_TRACE_CLOSE)
ULONG FatMaxDelayedCloseCount;
#define FatAcquireCloseMutex() { \
NT_ASSERT(KeAreApcsDisabled()); \
ExAcquireFastMutexUnsafe( &FatCloseQueueMutex ); \
}
#define FatReleaseCloseMutex() { \
NT_ASSERT(KeAreApcsDisabled()); \
ExReleaseFastMutexUnsafe( &FatCloseQueueMutex ); \
}
//
// Local procedure prototypes
//
_Requires_lock_held_(_Global_critical_region_)
VOID
FatQueueClose (
IN PCLOSE_CONTEXT CloseContext,
IN BOOLEAN DelayClose
);
_Requires_lock_held_(_Global_critical_region_)
PCLOSE_CONTEXT
FatRemoveClose (
PVCB Vcb OPTIONAL,
PVCB LastVcbHint OPTIONAL
);
IO_WORKITEM_ROUTINE FatCloseWorker;
VOID
NTAPI
FatCloseWorker (
_In_ PDEVICE_OBJECT DeviceObject,
_In_opt_ PVOID Context
);
#ifdef ALLOC_PRAGMA
#pragma alloc_text(PAGE, FatFsdClose)
#pragma alloc_text(PAGE, FatFspClose)
#pragma alloc_text(PAGE, FatRemoveClose)
#pragma alloc_text(PAGE, FatCommonClose)
#pragma alloc_text(PAGE, FatCloseWorker)
#endif
_Function_class_(IRP_MJ_CLOSE)
_Function_class_(DRIVER_DISPATCH)
NTSTATUS
NTAPI
FatFsdClose (
_In_ PVOLUME_DEVICE_OBJECT VolumeDeviceObject,
_Inout_ PIRP Irp
)
/*++
Routine Description:
This routine implements the FSD part of Close.
Arguments:
VolumeDeviceObject - Supplies the volume device object where the
file exists
Irp - Supplies the Irp being processed
Return Value:
NTSTATUS - The FSD status for the IRP
--*/
{
NTSTATUS Status = STATUS_SUCCESS;
PIO_STACK_LOCATION IrpSp;
PFILE_OBJECT FileObject;
PVCB Vcb;
PFCB Fcb;
PCCB Ccb;
TYPE_OF_OPEN TypeOfOpen;
BOOLEAN TopLevel;
BOOLEAN VcbDeleted = FALSE;
PAGED_CODE();
//
// If we were called with our file system device object instead of a
// volume device object, just complete this request with STATUS_SUCCESS
//
if (FatDeviceIsFatFsdo( VolumeDeviceObject)) {
Irp->IoStatus.Status = STATUS_SUCCESS;
Irp->IoStatus.Information = FILE_OPENED;
IoCompleteRequest( Irp, IO_DISK_INCREMENT );
return STATUS_SUCCESS;
}
DebugTrace(+1, Dbg, "FatFsdClose\n", 0);
//
// Call the common Close routine
//
FsRtlEnterFileSystem();
TopLevel = FatIsIrpTopLevel( Irp );
//
// Get a pointer to the current stack location and the file object
//
IrpSp = IoGetCurrentIrpStackLocation( Irp );
FileObject = IrpSp->FileObject;
//
// Decode the file object and set the read-only bit in the Ccb.
//
TypeOfOpen = FatDecodeFileObject( FileObject, &Vcb, &Fcb, &Ccb );
if (Ccb && IsFileObjectReadOnly(FileObject)) {
SetFlag( Ccb->Flags, CCB_FLAG_READ_ONLY );
}
_SEH2_TRY {
PCLOSE_CONTEXT CloseContext = NULL;
//
// If we are top level, WAIT can be TRUE, otherwise make it FALSE
// to avoid deadlocks, unless this is a top
// level request not originating from the system process.
//
BOOLEAN Wait = TopLevel && (PsGetCurrentProcess() != FatData.OurProcess);
#if (NTDDI_VERSION >= NTDDI_WIN8)
//
// To catch the odd case where a close comes in without a preceding cleanup,
// call the oplock package to get rid of any oplock state. This can only
// be safely done in the FSD path.
//
if ((Fcb != NULL) &&
!FlagOn( FileObject->Flags, FO_CLEANUP_COMPLETE ) &&
FatIsFileOplockable( Fcb )) {
//
// This is equivalent to handling cleanup, and it always cleans up any
// oplock immediately. Also, we don't need any locking of the FCB here;
// the oplock's own lock will be sufficient for this purpose.
//
FsRtlCheckOplockEx( FatGetFcbOplock(Fcb),
Irp,
0,
NULL,
NULL,
NULL );
}
#endif
//
// Metadata streams have had close contexts preallocated. Pull one out now, while we're
// guaranteed the VCB exists.
//
if ( (TypeOfOpen == VirtualVolumeFile) || (TypeOfOpen == DirectoryFile) || (TypeOfOpen == EaFile)
) {
CloseContext = FatAllocateCloseContext( Vcb );
NT_ASSERT( CloseContext != NULL );
CloseContext->Free = TRUE;
}
//
// Call the common Close routine if we are not delaying this close.
//
if ((((TypeOfOpen == UserFileOpen) ||
(TypeOfOpen == UserDirectoryOpen)) &&
FlagOn(Fcb->FcbState, FCB_STATE_DELAY_CLOSE) &&
!FatData.ShutdownStarted) ||
(FatCommonClose( Vcb, Fcb, Ccb, TypeOfOpen, Wait, TopLevel, &VcbDeleted ) == STATUS_PENDING)) {
//
// Ship it off to the delayed close queue if we tried to close, and got STATUS_PENDING, or
// if the user open told us to delay the close.
//
//
// Metadata streams have had close contexts preallocated. If we have a user open,
// pull the close context out of the Ccb.
//
if( CloseContext == NULL ) {
//
// Free up any query template strings before using the close context fields,
// which overlap (union)
//
FatDeallocateCcbStrings( Ccb );
CloseContext = &Ccb->CloseContext;
CloseContext->Free = FALSE;
SetFlag( Ccb->Flags, CCB_FLAG_CLOSE_CONTEXT );
}
//
// If the status is pending, then let's get the information we
// need into the close context we already have bagged, complete
// the request, and post it. It is important we allocate nothing
// in the close path.
//
CloseContext->Vcb = Vcb;
CloseContext->Fcb = Fcb;
CloseContext->TypeOfOpen = TypeOfOpen;
//
// Send it off, either to an ExWorkerThread or to the async
// close list.
//
FatQueueClose( CloseContext,
(BOOLEAN)(Fcb && FlagOn(Fcb->FcbState, FCB_STATE_DELAY_CLOSE)));
} else {
//
// The close proceeded synchronously, so for the metadata objects we
// can now drop the close context we preallocated.
//
if ((TypeOfOpen == VirtualVolumeFile) ||
(TypeOfOpen == DirectoryFile) ||
(TypeOfOpen == EaFile)
) {
if (CloseContext != NULL) {
ExFreePool( CloseContext );
}
}
}
FatCompleteRequest( FatNull, Irp, Status );
}
_SEH2_EXCEPT(FatExceptionFilter( NULL, _SEH2_GetExceptionInformation() )) {
//
// We had some trouble trying to perform the requested
// operation, so we'll abort the I/O request with the
// error status that we get back from the exception code.
//
Status = FatProcessException( NULL, Irp, _SEH2_GetExceptionCode() );
} _SEH2_END;
if (TopLevel) { IoSetTopLevelIrp( NULL ); }
FsRtlExitFileSystem();
//
// And return to our caller
//
DebugTrace(-1, Dbg, "FatFsdClose -> %08lx\n", Status);
UNREFERENCED_PARAMETER( VolumeDeviceObject );
return Status;
}
VOID
NTAPI
FatCloseWorker (
_In_ PDEVICE_OBJECT DeviceObject,
_In_opt_ PVOID Context
)
/*++
Routine Description:
This routine is a shim between the IO worker package and FatFspClose.
Arguments:
DeviceObject - Registration device object, unused
Context - Context value, unused
Return Value:
None.
--*/
{
PAGED_CODE();
UNREFERENCED_PARAMETER( DeviceObject );
FsRtlEnterFileSystem();
FatFspClose (Context);
FsRtlExitFileSystem();
}
_Requires_lock_held_(_Global_critical_region_)
VOID
FatFspClose (
IN PVCB Vcb OPTIONAL
)
/*++
Routine Description:
This routine implements the FSP part of Close.
Arguments:
Vcb - If present, tells us to only close file objects opened on the
specified volume.
Return Value:
None.
--*/
{
PCLOSE_CONTEXT CloseContext;
PVCB CurrentVcb = NULL;
PVCB LastVcb = NULL;
BOOLEAN FreeContext = FALSE;
BOOLEAN TopLevel = FALSE;
ULONG LoopsWithVcbHeld = 0;
PAGED_CODE();
DebugTrace(+1, Dbg, "FatFspClose\n", 0);
//
// Set the top level IRP for the true FSP operation.
//
if (!ARGUMENT_PRESENT( Vcb )) {
IoSetTopLevelIrp( (PIRP)FSRTL_FSP_TOP_LEVEL_IRP );
TopLevel = TRUE;
}
while ((CloseContext = FatRemoveClose(Vcb, LastVcb)) != NULL) {
//
// If we are in the FSP (i.e. Vcb == NULL), then try to keep ahead of
// creates by doing several closes with one acquisition of the Vcb.
//
// Note that we cannot be holding the Vcb on entry to FatCommonClose
// if this is last close as we will try to acquire FatData, and
// worse the volume (and therefore the Vcb) may go away.
//
if (!ARGUMENT_PRESENT(Vcb)) {
if (!FatData.ShutdownStarted) {
if (CloseContext->Vcb != CurrentVcb) {
LoopsWithVcbHeld = 0;
//
// Release a previously held Vcb, if any.
//
if (CurrentVcb != NULL) {
ExReleaseResourceLite( &CurrentVcb->Resource);
}
//
// Get the new Vcb.
//
CurrentVcb = CloseContext->Vcb;
(VOID)ExAcquireResourceExclusiveLite( &CurrentVcb->Resource, TRUE );
} else {
//
// Share the resource occasionally if we seem to be finding a lot
// of closes for a single volume.
//
if (++LoopsWithVcbHeld >= 20) {
if (ExGetSharedWaiterCount( &CurrentVcb->Resource ) +
ExGetExclusiveWaiterCount( &CurrentVcb->Resource )) {
ExReleaseResourceLite( &CurrentVcb->Resource);
(VOID)ExAcquireResourceExclusiveLite( &CurrentVcb->Resource, TRUE );
}
LoopsWithVcbHeld = 0;
}
}
//
// Now check the Open count. We may be about to delete this volume!
//
// The test below must be <= 1 because there could still be outstanding
// stream references on this VCB that are not counted in the OpenFileCount.
// For example if there are no open files OpenFileCount could be zero and we would
// not release the resource here. The call to FatCommonClose() below may cause
// the VCB to be torn down and we will try to release memory we don't
// own later.
//
if (CurrentVcb->OpenFileCount <= 1) {
ExReleaseResourceLite( &CurrentVcb->Resource);
CurrentVcb = NULL;
}
//
// If shutdown has started while processing our list, drop the
// current Vcb resource.
//
} else if (CurrentVcb != NULL) {
ExReleaseResourceLite( &CurrentVcb->Resource);
CurrentVcb = NULL;
}
}
LastVcb = CurrentVcb;
//
// Call the common Close routine. Protected in a try {} except {}
//
_SEH2_TRY {
//
// The close context either is in the CCB, automatically freed,
// or was from pool for a metadata fileobject, CCB is NULL, and
// we'll need to free it.
//
FreeContext = CloseContext->Free;
(VOID)FatCommonClose( CloseContext->Vcb,
CloseContext->Fcb,
(FreeContext ? NULL :
CONTAINING_RECORD( CloseContext, CCB, CloseContext)),
CloseContext->TypeOfOpen,
TRUE,
TopLevel,
NULL );
} _SEH2_EXCEPT(FatExceptionFilter( NULL, _SEH2_GetExceptionInformation() )) {
//
// Ignore anything we expect.
//
NOTHING;
} _SEH2_END;
//
// Drop the context if it came from pool.
//
if (FreeContext) {
ExFreePool( CloseContext );
}
}
//
// Release a previously held Vcb, if any.
//
if (CurrentVcb != NULL) {
ExReleaseResourceLite( &CurrentVcb->Resource);
}
//
// Clean up the top level IRP hint if we owned it.
//
if (!ARGUMENT_PRESENT( Vcb )) {
IoSetTopLevelIrp( NULL );
}
//
// And return to our caller
//
DebugTrace(-1, Dbg, "FatFspClose -> NULL\n", 0);
}
_Requires_lock_held_(_Global_critical_region_)
VOID
FatQueueClose (
IN PCLOSE_CONTEXT CloseContext,
IN BOOLEAN DelayClose
)
/*++
Routine Description:
Enqueue a deferred close to one of the two delayed close queues.
Arguments:
CloseContext - a close context to enqueue for the delayed close thread.
DelayClose - whether this should go on the delayed close queue (unreferenced
objects).
Return Value:
None.
--*/
{
BOOLEAN StartWorker = FALSE;
FatAcquireCloseMutex();
if (DelayClose) {
InsertTailList( &FatData.DelayedCloseList,
&CloseContext->GlobalLinks );
InsertTailList( &CloseContext->Vcb->DelayedCloseList,
&CloseContext->VcbLinks );
FatData.DelayedCloseCount += 1;
if ((FatData.DelayedCloseCount > FatMaxDelayedCloseCount) &&
!FatData.AsyncCloseActive) {
FatData.AsyncCloseActive = TRUE;
StartWorker = TRUE;
}
} else {
InsertTailList( &FatData.AsyncCloseList,
&CloseContext->GlobalLinks );
InsertTailList( &CloseContext->Vcb->AsyncCloseList,
&CloseContext->VcbLinks );
FatData.AsyncCloseCount += 1;
if (!FatData.AsyncCloseActive) {
FatData.AsyncCloseActive = TRUE;
StartWorker = TRUE;
}
}
FatReleaseCloseMutex();
if (StartWorker) {
IoQueueWorkItem( FatData.FatCloseItem, FatCloseWorker, CriticalWorkQueue, NULL );
}
}
_Requires_lock_held_(_Global_critical_region_)
PCLOSE_CONTEXT
FatRemoveClose (
PVCB Vcb OPTIONAL,
PVCB LastVcbHint OPTIONAL
)
/*++
Routine Description:
Dequeue a deferred close from one of the two delayed close queues.
Arguments:
Vcb - if specified, only returns close for this volume.
LastVcbHint - if specified and other starvation avoidance is required by
the system condition, will attempt to return closes for this volume.
Return Value:
A close to perform.
--*/
{
PLIST_ENTRY Entry;
PCLOSE_CONTEXT CloseContext;
BOOLEAN WorkerThread;
PAGED_CODE();
FatAcquireCloseMutex();
//
// Remember if this is the worker thread, so we can pull down the active
// flag should we run everything out.
//
WorkerThread = (Vcb == NULL);
//
// If the queues are above the limits by a significant amount, we have
// to try hard to pull them down. To do this, we will aggressively try
// to find closes for the last volume the caller looked at. This will
// make sure we fully utilize the acquisition of the volume, which can
// be a hugely expensive resource to get (create/close/cleanup use it
// exclusively).
//
// Only do this in the delayed close thread. We will know this is the
// case by seeing a NULL mandatory Vcb.
//
if (Vcb == NULL && LastVcbHint != NULL) {
//
// Flip over to aggressive at twice the legal limit, and flip it
// off at the legal limit.
//
if (!FatData.HighAsync && FatData.AsyncCloseCount > FatMaxDelayedCloseCount*2) {
FatData.HighAsync = TRUE;
} else if (FatData.HighAsync && FatData.AsyncCloseCount < FatMaxDelayedCloseCount) {
FatData.HighAsync = FALSE;
}
if (!FatData.HighDelayed && FatData.DelayedCloseCount > FatMaxDelayedCloseCount*2) {
FatData.HighDelayed = TRUE;
} else if (FatData.HighDelayed && FatData.DelayedCloseCount < FatMaxDelayedCloseCount) {
FatData.HighDelayed = FALSE;
}
if (FatData.HighAsync || FatData.HighDelayed) {
Vcb = LastVcbHint;
}
}
//
// Do the case when we don't care about which Vcb the close is on.
// This is the case when we are in an ExWorkerThread and aren't
// under pressure.
//
if (Vcb == NULL) {
AnyClose:
//
// First check the list of async closes.
//
if (!IsListEmpty( &FatData.AsyncCloseList )) {
Entry = RemoveHeadList( &FatData.AsyncCloseList );
FatData.AsyncCloseCount -= 1;
CloseContext = CONTAINING_RECORD( Entry,
CLOSE_CONTEXT,
GlobalLinks );
RemoveEntryList( &CloseContext->VcbLinks );
//
// Do any delayed closes over half the limit, unless shutdown has
// started (then kill them all).
//
} else if (!IsListEmpty( &FatData.DelayedCloseList ) &&
(FatData.DelayedCloseCount > FatMaxDelayedCloseCount/2 ||
FatData.ShutdownStarted)) {
Entry = RemoveHeadList( &FatData.DelayedCloseList );
FatData.DelayedCloseCount -= 1;
CloseContext = CONTAINING_RECORD( Entry,
CLOSE_CONTEXT,
GlobalLinks );
RemoveEntryList( &CloseContext->VcbLinks );
//
// There are no more closes to perform; show that we are done.
//
} else {
CloseContext = NULL;
if (WorkerThread) {
FatData.AsyncCloseActive = FALSE;
}
}
//
// We're running down a specific volume.
//
} else {
//
// First check the list of async closes.
//
if (!IsListEmpty( &Vcb->AsyncCloseList )) {
Entry = RemoveHeadList( &Vcb->AsyncCloseList );
FatData.AsyncCloseCount -= 1;
CloseContext = CONTAINING_RECORD( Entry,
CLOSE_CONTEXT,
VcbLinks );
RemoveEntryList( &CloseContext->GlobalLinks );
//
// Do any delayed closes.
//
} else if (!IsListEmpty( &Vcb->DelayedCloseList )) {
Entry = RemoveHeadList( &Vcb->DelayedCloseList );
FatData.DelayedCloseCount -= 1;
CloseContext = CONTAINING_RECORD( Entry,
CLOSE_CONTEXT,
VcbLinks );
RemoveEntryList( &CloseContext->GlobalLinks );
//
// If we were trying to run down the queues but didn't find anything for this
// volume, flip over to accept anything and try again.
//
} else if (LastVcbHint) {
goto AnyClose;
//
// There are no more closes to perform; show that we are done.
//
} else {
CloseContext = NULL;
}
}
FatReleaseCloseMutex();
return CloseContext;
}
_Requires_lock_held_(_Global_critical_region_)
NTSTATUS
FatCommonClose (
IN PVCB Vcb,
IN PFCB Fcb,
IN PCCB Ccb,
IN TYPE_OF_OPEN TypeOfOpen,
IN BOOLEAN Wait,
IN BOOLEAN TopLevel,
OUT PBOOLEAN VcbDeleted OPTIONAL
)
/*++
Routine Description:
This is the common routine for closing a file/directory called by both
the fsd and fsp threads.
Close is invoked whenever the last reference to a file object is deleted.
Cleanup is invoked when the last handle to a file object is closed, and
is called before close.
The function of close is to completely tear down and remove the fcb/dcb/ccb
structures associated with the file object.
Arguments:
Fcb - Supplies the file to process.
Wait - If this is TRUE we are allowed to block for the Vcb, if FALSE
then we must try to acquire the Vcb anyway.
TopLevel - If this is TRUE this is a top level request.
VcbDeleted - Returns whether the VCB was deleted by this call.
Return Value:
NTSTATUS - The return status for the operation
--*/
{
NTSTATUS Status = STATUS_SUCCESS;
PDCB ParentDcb;
BOOLEAN RecursiveClose;
BOOLEAN LocalVcbDeleted;
IRP_CONTEXT IrpContext;
PAGED_CODE();
DebugTrace(+1, Dbg, "FatCommonClose...\n", 0);
//
// Initialize the callers variable, if needed.
//
LocalVcbDeleted = FALSE;
if (ARGUMENT_PRESENT( VcbDeleted )) {
*VcbDeleted = LocalVcbDeleted;
}
//
// Special case the unopened file object
//
if (TypeOfOpen == UnopenedFileObject) {
DebugTrace(0, Dbg, "Close unopened file object\n", 0);
Status = STATUS_SUCCESS;
DebugTrace(-1, Dbg, "FatCommonClose -> %08lx\n", Status);
return Status;
}
//
// Set up our stack IrpContext.
//
RtlZeroMemory( &IrpContext, sizeof(IRP_CONTEXT) );
IrpContext.NodeTypeCode = FAT_NTC_IRP_CONTEXT;
IrpContext.NodeByteSize = sizeof( IrpContext );
IrpContext.MajorFunction = IRP_MJ_CLOSE;
IrpContext.Vcb = Vcb;
if (Wait) {
SetFlag( IrpContext.Flags, IRP_CONTEXT_FLAG_WAIT );
}
//
// Acquire exclusive access to the Vcb and enqueue the irp if we didn't
// get access.
//
#ifdef _MSC_VER
#pragma prefast( suppress: 28137, "prefast wants Wait to be a constant, but that's not possible for fastfat" )
#endif
if (!ExAcquireResourceExclusiveLite( &Vcb->Resource, Wait )) {
return STATUS_PENDING;
}
//
// The following test makes sure that we don't blow away an Fcb if we
// are trying to do a Supersede/Overwrite open above us. This test
// does not apply for the EA file.
//
if (FlagOn(Vcb->VcbState, VCB_STATE_FLAG_CREATE_IN_PROGRESS) &&
Vcb->EaFcb != Fcb) {
ExReleaseResourceLite( &Vcb->Resource );
return STATUS_PENDING;
}
//
// Setting the following flag prevents recursive closes of directory file
// objects, which are handled in a special case loop.
//
if ( FlagOn(Vcb->VcbState, VCB_STATE_FLAG_CLOSE_IN_PROGRESS) ) {
RecursiveClose = TRUE;
} else {
SetFlag(Vcb->VcbState, VCB_STATE_FLAG_CLOSE_IN_PROGRESS);
RecursiveClose = FALSE;
//
// Since we are at the top of the close chain, we need to add
// a reference to the VCB. This will keep it from going away
// on us until we are ready to check for a dismount below.
//
Vcb->OpenFileCount += 1;
}
_SEH2_TRY {
//
// Case on the type of open that we are trying to close.
//
switch (TypeOfOpen) {
case VirtualVolumeFile:
DebugTrace(0, Dbg, "Close VirtualVolumeFile\n", 0);
//
// Remove this internal, residual open from the count.
//
InterlockedDecrement( (LONG*)&(Vcb->InternalOpenCount) );
InterlockedDecrement( (LONG*)&(Vcb->ResidualOpenCount) );
try_return( Status = STATUS_SUCCESS );
break;
case UserVolumeOpen:
DebugTrace(0, Dbg, "Close UserVolumeOpen\n", 0);
Vcb->DirectAccessOpenCount -= 1;
Vcb->OpenFileCount -= 1;
if (FlagOn(Ccb->Flags, CCB_FLAG_READ_ONLY)) { Vcb->ReadOnlyCount -= 1; }
FatDeleteCcb( &IrpContext, &Ccb );
try_return( Status = STATUS_SUCCESS );
break;
case EaFile:
DebugTrace(0, Dbg, "Close EaFile\n", 0);
//
// Remove this internal, residual open from the count.
//
InterlockedDecrement( (LONG*)&(Vcb->InternalOpenCount) );
InterlockedDecrement( (LONG*)&(Vcb->ResidualOpenCount) );
try_return( Status = STATUS_SUCCESS );
break;
case DirectoryFile:
DebugTrace(0, Dbg, "Close DirectoryFile\n", 0);
InterlockedDecrement( (LONG*)&Fcb->Specific.Dcb.DirectoryFileOpenCount );
//
// Remove this internal open from the count.
//
InterlockedDecrement( (LONG*)&(Vcb->InternalOpenCount) );
//
// If this is the root directory, it is a residual open
// as well.
//
if (NodeType( Fcb ) == FAT_NTC_ROOT_DCB) {
InterlockedDecrement( (LONG*)&(Vcb->ResidualOpenCount) );
}
//
// If this is a recursive close, just return here.
//
if ( RecursiveClose ) {
try_return( Status = STATUS_SUCCESS );
} else {
break;
}
case UserDirectoryOpen:
case UserFileOpen:
DebugTrace(0, Dbg, "Close UserFileOpen/UserDirectoryOpen\n", 0);
//
// Uninitialize the cache map if we no longer need to use it
//
if ((NodeType(Fcb) == FAT_NTC_DCB) &&
IsListEmpty(&Fcb->Specific.Dcb.ParentDcbQueue) &&
(Fcb->OpenCount == 1) &&
(Fcb->Specific.Dcb.DirectoryFile != NULL)) {
PFILE_OBJECT DirectoryFileObject = Fcb->Specific.Dcb.DirectoryFile;
DebugTrace(0, Dbg, "Uninitialize the stream file object\n", 0);
CcUninitializeCacheMap( DirectoryFileObject, NULL, NULL );
//
// Dereference the directory file. This may cause a close
// Irp to be processed, so we need to do this before we destroy
// the Fcb.
//
Fcb->Specific.Dcb.DirectoryFile = NULL;
ObDereferenceObject( DirectoryFileObject );
}
Fcb->OpenCount -= 1;
Vcb->OpenFileCount -= 1;
if (FlagOn(Ccb->Flags, CCB_FLAG_READ_ONLY)) { Vcb->ReadOnlyCount -= 1; }
FatDeleteCcb( &IrpContext, &Ccb );
break;
default:
#ifdef _MSC_VER
#pragma prefast( suppress: 28159, "if the type of open is unknown, we seriously messed up." )
#endif
FatBugCheck( TypeOfOpen, 0, 0 );
}
//
// At this point we've cleaned up any on-disk structure that needs
// to be done, and we can now update the in-memory structures.
// Now if this is an unreferenced FCB or if it is
// an unreferenced DCB (not the root) then we can remove
// the fcb and set our ParentDcb to non null.
//
if (((NodeType(Fcb) == FAT_NTC_FCB) &&
(Fcb->OpenCount == 0))
||
((NodeType(Fcb) == FAT_NTC_DCB) &&
(IsListEmpty(&Fcb->Specific.Dcb.ParentDcbQueue)) &&
(Fcb->OpenCount == 0) &&
(Fcb->Specific.Dcb.DirectoryFileOpenCount == 0))) {
ParentDcb = Fcb->ParentDcb;
SetFlag( Vcb->VcbState, VCB_STATE_FLAG_DELETED_FCB );
FatDeleteFcb( &IrpContext, &Fcb );
//
// Uninitialize our parent's cache map if we no longer need
// to use it.
//
while ((NodeType(ParentDcb) == FAT_NTC_DCB) &&
IsListEmpty(&ParentDcb->Specific.Dcb.ParentDcbQueue) &&
(ParentDcb->OpenCount == 0) &&
(ParentDcb->Specific.Dcb.DirectoryFile != NULL)) {
PFILE_OBJECT DirectoryFileObject;
DirectoryFileObject = ParentDcb->Specific.Dcb.DirectoryFile;
DebugTrace(0, Dbg, "Uninitialize our parent Stream Cache Map\n", 0);
CcUninitializeCacheMap( DirectoryFileObject, NULL, NULL );
ParentDcb->Specific.Dcb.DirectoryFile = NULL;
ObDereferenceObject( DirectoryFileObject );
//
// Now, if the ObDereferenceObject() caused the final close
// to come in, then blow away the Fcb and continue up,
// otherwise wait for Mm to to dereference its file objects
// and stop here..
//
if ( ParentDcb->Specific.Dcb.DirectoryFileOpenCount == 0) {
PDCB CurrentDcb;
CurrentDcb = ParentDcb;
ParentDcb = CurrentDcb->ParentDcb;
SetFlag( Vcb->VcbState, VCB_STATE_FLAG_DELETED_FCB );
FatDeleteFcb( &IrpContext, &CurrentDcb );
} else {
break;
}
}
}
Status = STATUS_SUCCESS;
try_exit: NOTHING;
} _SEH2_FINALLY {
DebugUnwind( FatCommonClose );
//
// We are done processing the close. If we are the top of the close
// chain, see if the VCB can go away. We have biased the open count by
// one, so we need to take that into account.
//
if (!RecursiveClose) {
//
// See if there is only one open left. If so, it is ours. We only want
// to check for a dismount if a dismount is not already in progress.
// We also only do this if the Vcb condition is not VcbGood and the
// caller can handle the VCB going away. This is determined by whether
// they passed in the VcbDeleted argument. This request also needs
// to be top level.
//
if (Vcb->OpenFileCount == 1 &&
Vcb->VcbCondition != VcbGood &&
!FlagOn( Vcb->VcbState, VCB_STATE_FLAG_DISMOUNT_IN_PROGRESS ) &&
ARGUMENT_PRESENT( VcbDeleted ) &&
TopLevel) {
//
// We need the global lock, which must be acquired before the
// VCB. Since we already have the VCB, we have to drop and
// reacquire here. Note that we always want to wait from this
// point on. Note that the VCB cannot go away, since we have
// biased the open file count.
//
FatReleaseVcb( &IrpContext,
Vcb );
SetFlag( IrpContext.Flags, IRP_CONTEXT_FLAG_WAIT );
#ifdef _MSC_VER
#pragma prefast( suppress: 28137, "prefast wants the wait parameter in this macro expansion to be a constant, unfortunately this is not possible" )
#endif
FatAcquireExclusiveGlobal( &IrpContext );
FatAcquireExclusiveVcb( &IrpContext,
Vcb );
//
// We have our locks in the correct order. Remove our
// extra open and check for a dismount. Note that if
// something changed while we dropped the lock, it will
// not matter, since the dismount code does the correct
// checks to make sure the volume can really go away.
//
Vcb->OpenFileCount -= 1;
LocalVcbDeleted = FatCheckForDismount( &IrpContext,
Vcb,
FALSE );
FatReleaseGlobal( &IrpContext );
//
// Let the caller know what happened, if they want this information.
//
if (ARGUMENT_PRESENT( VcbDeleted )) {
*VcbDeleted = LocalVcbDeleted;
}
} else {
//
// The volume cannot go away now. Just remove our extra reference.
//
Vcb->OpenFileCount -= 1;
}
//
// If the VCB is still around, clear our recursion flag.
//
if (!LocalVcbDeleted) {
ClearFlag( Vcb->VcbState, VCB_STATE_FLAG_CLOSE_IN_PROGRESS );
}
}
//
// Only release the VCB if it did not go away.
//
if (!LocalVcbDeleted) {
FatReleaseVcb( &IrpContext, Vcb );
}
DebugTrace(-1, Dbg, "FatCommonClose -> %08lx\n", Status);
} _SEH2_END;
return Status;
}
Reference in a new issue View git blame Copy permalink