reactos/ntoskrnl/config/cmapi.c

2847 lines
85 KiB
C
Raw Normal View History

/*
* PROJECT: ReactOS Kernel
* LICENSE: GPL - See COPYING in the top level directory
* FILE: ntoskrnl/config/cmapi.c
* PURPOSE: Configuration Manager - Internal Registry APIs
* PROGRAMMERS: Alex Ionescu (alex.ionescu@reactos.org)
* Eric Kohl
* Aleksandar Andrejevic <theflash AT sdf DOT lonestar DOT org>
*/
/* INCLUDES ******************************************************************/
#include "ntoskrnl.h"
#define NDEBUG
#include "debug.h"
/* FUNCTIONS *****************************************************************/
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
BOOLEAN
NTAPI
CmpIsHiveAlreadyLoaded(IN HANDLE KeyHandle,
IN POBJECT_ATTRIBUTES SourceFile,
OUT PCMHIVE *CmHive)
{
NTSTATUS Status;
PCM_KEY_BODY KeyBody;
PCMHIVE Hive;
BOOLEAN Loaded = FALSE;
PAGED_CODE();
/* Sanity check */
CMP_ASSERT_EXCLUSIVE_REGISTRY_LOCK();
/* Reference the handle */
Status = ObReferenceObjectByHandle(KeyHandle,
0,
CmpKeyObjectType,
KernelMode,
(PVOID)&KeyBody,
NULL);
if (!NT_SUCCESS(Status)) return Loaded;
/* Don't touch deleted KCBs */
if (KeyBody->KeyControlBlock->Delete) return Loaded;
Hive = CONTAINING_RECORD(KeyBody->KeyControlBlock->KeyHive, CMHIVE, Hive);
/* Must be the root key */
if (!(KeyBody->KeyControlBlock->Flags & KEY_HIVE_ENTRY) ||
!(Hive->FileUserName.Buffer))
{
/* It isn't */
ObDereferenceObject(KeyBody);
return Loaded;
}
/* Now compare the name of the file */
if (!RtlCompareUnicodeString(&Hive->FileUserName,
SourceFile->ObjectName,
TRUE))
{
/* Same file found */
Loaded = TRUE;
*CmHive = Hive;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* If the hive is frozen, not sure what to do */
if (Hive->Frozen)
{
/* FIXME: TODO */
UNIMPLEMENTED_DBGBREAK("ERROR: Hive is frozen\n");
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
}
/* Dereference and return result */
ObDereferenceObject(KeyBody);
return Loaded;
}
BOOLEAN
NTAPI
CmpDoFlushAll(IN BOOLEAN ForceFlush)
{
PLIST_ENTRY NextEntry;
PCMHIVE Hive;
NTSTATUS Status;
BOOLEAN Result = TRUE;
/* Make sure that the registry isn't read-only now */
if (CmpNoWrite) return TRUE;
/* Otherwise, acquire the hive list lock and disable force flush */
CmpForceForceFlush = FALSE;
ExAcquirePushLockShared(&CmpHiveListHeadLock);
/* Loop the hive list */
NextEntry = CmpHiveListHead.Flink;
while (NextEntry != &CmpHiveListHead)
{
/* Get the hive */
Hive = CONTAINING_RECORD(NextEntry, CMHIVE, HiveList);
if (!(Hive->Hive.HiveFlags & HIVE_NOLAZYFLUSH))
{
/* Acquire the flusher lock */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
CmpLockHiveFlusherExclusive(Hive);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check for illegal state */
if (ForceFlush && Hive->UseCount)
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* Registry needs to be locked down */
CMP_ASSERT_EXCLUSIVE_REGISTRY_LOCK();
UNIMPLEMENTED_DBGBREAK("FIXME: Hive is damaged and needs fixup\n");
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Only sync if we are forced to or if it won't cause a hive shrink */
if (ForceFlush || !HvHiveWillShrink(&Hive->Hive))
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* Do the sync */
Status = HvSyncHive(&Hive->Hive);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* If something failed - set the flag and continue looping */
if (!NT_SUCCESS(Status))
Result = FALSE;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
else
{
/* We won't flush if the hive might shrink */
Result = FALSE;
CmpForceForceFlush = TRUE;
}
/* Release the flusher lock */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
CmpUnlockHiveFlusher(Hive);
}
/* Try the next entry */
NextEntry = NextEntry->Flink;
}
/* Release lock and return */
ExReleasePushLock(&CmpHiveListHeadLock);
return Result;
}
NTSTATUS
NTAPI
CmpSetValueKeyNew(IN PHHIVE Hive,
IN PCM_KEY_NODE Parent,
IN PUNICODE_STRING ValueName,
IN ULONG Index,
IN ULONG Type,
IN PVOID Data,
IN ULONG DataSize,
IN ULONG StorageType,
IN ULONG SmallData)
{
PCELL_DATA CellData;
HCELL_INDEX ValueCell;
NTSTATUS Status;
/* Check if we already have a value list */
if (Parent->ValueList.Count)
{
/* Then make sure it's valid and dirty it */
ASSERT(Parent->ValueList.List != HCELL_NIL);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (!HvMarkCellDirty(Hive, Parent->ValueList.List, FALSE))
{
/* Fail if we're out of space for log changes */
return STATUS_NO_LOG_SPACE;
}
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Allocate a value cell */
ValueCell = HvAllocateCell(Hive,
FIELD_OFFSET(CM_KEY_VALUE, Name) +
CmpNameSize(Hive, ValueName),
StorageType,
HCELL_NIL);
if (ValueCell == HCELL_NIL) return STATUS_INSUFFICIENT_RESOURCES;
/* Get the actual data for it */
CellData = HvGetCell(Hive, ValueCell);
ASSERT(CellData);
/* Now we can release it, make sure it's also dirty */
HvReleaseCell(Hive, ValueCell);
ASSERT(HvIsCellDirty(Hive, ValueCell));
/* Set it up and copy the name */
CellData->u.KeyValue.Signature = CM_KEY_VALUE_SIGNATURE;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
_SEH2_TRY
{
/* This can crash since the name is coming from user-mode */
CellData->u.KeyValue.NameLength = CmpCopyName(Hive,
CellData->u.KeyValue.Name,
ValueName);
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
/* Fail */
DPRINT1("Invalid user data!\n");
HvFreeCell(Hive, ValueCell);
_SEH2_YIELD(return _SEH2_GetExceptionCode());
}
_SEH2_END;
/* Check for compressed name */
if (CellData->u.KeyValue.NameLength < ValueName->Length)
{
/* This is a compressed name */
CellData->u.KeyValue.Flags = VALUE_COMP_NAME;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
else
{
/* No flags to set */
CellData->u.KeyValue.Flags = 0;
}
/* Check if this is a normal key */
if (DataSize > CM_KEY_VALUE_SMALL)
{
/* Build a data cell for it */
Status = CmpSetValueDataNew(Hive,
Data,
DataSize,
StorageType,
ValueCell,
&CellData->u.KeyValue.Data);
if (!NT_SUCCESS(Status))
{
/* We failed, free the cell */
HvFreeCell(Hive, ValueCell);
return Status;
}
/* Otherwise, set the data length, and make sure the data is dirty */
CellData->u.KeyValue.DataLength = DataSize;
ASSERT(HvIsCellDirty(Hive, CellData->u.KeyValue.Data));
}
else
{
/* This is a small key, set the data directly inside */
CellData->u.KeyValue.DataLength = DataSize + CM_KEY_VALUE_SPECIAL_SIZE;
CellData->u.KeyValue.Data = SmallData;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Set the type now */
CellData->u.KeyValue.Type = Type;
/* Add this value cell to the child list */
Status = CmpAddValueToList(Hive,
ValueCell,
Index,
StorageType,
&Parent->ValueList);
/* If we failed, free the entire cell, including the data */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (!NT_SUCCESS(Status))
{
/* Overwrite the status with a known one */
CmpFreeValue(Hive, ValueCell);
Status = STATUS_INSUFFICIENT_RESOURCES;
}
/* Return Status */
return Status;
}
NTSTATUS
NTAPI
CmpSetValueKeyExisting(IN PHHIVE Hive,
IN HCELL_INDEX OldChild,
IN PCM_KEY_VALUE Value,
IN ULONG Type,
IN PVOID Data,
IN ULONG DataSize,
IN ULONG StorageType,
IN ULONG TempData)
{
HCELL_INDEX DataCell, NewCell;
PCELL_DATA CellData;
ULONG Length;
BOOLEAN WasSmall, IsSmall;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Registry writes must be blocked */
CMP_ASSERT_FLUSH_LOCK(Hive);
/* Mark the old child cell dirty */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (!HvMarkCellDirty(Hive, OldChild, FALSE)) return STATUS_NO_LOG_SPACE;
/* See if this is a small or normal key */
WasSmall = CmpIsKeyValueSmall(&Length, Value->DataLength);
/* See if our new data can fit in a small key */
IsSmall = (DataSize <= CM_KEY_VALUE_SMALL) ? TRUE: FALSE;
/* Big keys are unsupported */
ASSERT_VALUE_BIG(Hive, Length);
ASSERT_VALUE_BIG(Hive, DataSize);
/* Mark the old value dirty */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (!CmpMarkValueDataDirty(Hive, Value)) return STATUS_NO_LOG_SPACE;
/* Check if we have a small key */
if (IsSmall)
{
/* Check if we had a normal key with some data in it */
if (!(WasSmall) && (Length > 0))
{
/* Free the previous data */
CmpFreeValueData(Hive, Value->Data, Length);
}
/* Write our data directly */
Value->DataLength = DataSize + CM_KEY_VALUE_SPECIAL_SIZE;
Value->Data = TempData;
Value->Type = Type;
return STATUS_SUCCESS;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* We have a normal key. Was the old cell also normal and had data? */
if (!(WasSmall) && (Length > 0))
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Get the current data cell and actual data inside it */
DataCell = Value->Data;
ASSERT(DataCell != HCELL_NIL);
CellData = HvGetCell(Hive, DataCell);
if (!CellData) return STATUS_INSUFFICIENT_RESOURCES;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Immediately release the cell */
HvReleaseCell(Hive, DataCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Make sure that the data cell actually has a size */
ASSERT(HvGetCellSize(Hive, CellData) > 0);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check if the previous data cell could fit our new data */
if (DataSize <= (ULONG)(HvGetCellSize(Hive, CellData)))
{
/* Re-use it then */
NewCell = DataCell;
}
else
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Otherwise, re-allocate the current data cell */
NewCell = HvReallocateCell(Hive, DataCell, DataSize);
if (NewCell == HCELL_NIL) return STATUS_INSUFFICIENT_RESOURCES;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
else
{
/* This was a small key, or a key with no data, allocate a cell */
NewCell = HvAllocateCell(Hive, DataSize, StorageType, HCELL_NIL);
if (NewCell == HCELL_NIL) return STATUS_INSUFFICIENT_RESOURCES;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Now get the actual data for our data cell */
CellData = HvGetCell(Hive, NewCell);
ASSERT(CellData);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release it immediately */
HvReleaseCell(Hive, NewCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Copy our data into the data cell's buffer, and set up the value */
RtlCopyMemory(CellData, Data, DataSize);
Value->Data = NewCell;
Value->DataLength = DataSize;
Value->Type = Type;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Return success */
ASSERT(HvIsCellDirty(Hive, NewCell));
return STATUS_SUCCESS;
}
NTSTATUS
NTAPI
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
CmpQueryKeyData(IN PHHIVE Hive,
IN PCM_KEY_NODE Node,
IN KEY_INFORMATION_CLASS KeyInformationClass,
IN OUT PVOID KeyInformation,
IN ULONG Length,
IN OUT PULONG ResultLength)
{
NTSTATUS Status;
ULONG Size, SizeLeft, MinimumSize, Offset;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
PKEY_INFORMATION Info = (PKEY_INFORMATION)KeyInformation;
USHORT NameLength;
PVOID ClassData;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check if the value is compressed */
if (Node->Flags & KEY_COMP_NAME)
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Get the compressed name size */
NameLength = CmpCompressedNameSize(Node->Name, Node->NameLength);
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
else
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Get the real size */
NameLength = Node->NameLength;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check what kind of information is being requested */
switch (KeyInformationClass)
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Basic information */
case KeyBasicInformation:
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* This is the size we need */
Size = FIELD_OFFSET(KEY_BASIC_INFORMATION, Name) + NameLength;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* And this is the minimum we can work with */
MinimumSize = FIELD_OFFSET(KEY_BASIC_INFORMATION, Name);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Let the caller know and assume success */
*ResultLength = Size;
Status = STATUS_SUCCESS;
/* Check if the bufer we got is too small */
if (Length < MinimumSize)
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Let the caller know and fail */
Status = STATUS_BUFFER_TOO_SMALL;
break;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Copy the basic information */
Info->KeyBasicInformation.LastWriteTime = Node->LastWriteTime;
Info->KeyBasicInformation.TitleIndex = 0;
Info->KeyBasicInformation.NameLength = NameLength;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Only the name is left */
SizeLeft = Length - MinimumSize;
Size = NameLength;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check if we don't have enough space for the name */
if (SizeLeft < Size)
{
/* Truncate the name we'll return, and tell the caller */
Size = SizeLeft;
Status = STATUS_BUFFER_OVERFLOW;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check if this is a compressed key */
if (Node->Flags & KEY_COMP_NAME)
{
/* Copy the compressed name */
CmpCopyCompressedName(Info->KeyBasicInformation.Name,
SizeLeft,
Node->Name,
Node->NameLength);
}
else
{
/* Otherwise, copy the raw name */
RtlCopyMemory(Info->KeyBasicInformation.Name,
Node->Name,
Size);
}
break;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Node information */
case KeyNodeInformation:
/* Calculate the size we need */
Size = FIELD_OFFSET(KEY_NODE_INFORMATION, Name) +
NameLength +
Node->ClassLength;
/* And the minimum size we can support */
MinimumSize = FIELD_OFFSET(KEY_NODE_INFORMATION, Name);
/* Return the size to the caller and assume succes */
*ResultLength = Size;
Status = STATUS_SUCCESS;
/* Check if the caller's buffer is too small */
if (Length < MinimumSize)
{
/* Let them know, and fail */
Status = STATUS_BUFFER_TOO_SMALL;
break;
}
/* Copy the basic information */
Info->KeyNodeInformation.LastWriteTime = Node->LastWriteTime;
Info->KeyNodeInformation.TitleIndex = 0;
Info->KeyNodeInformation.ClassLength = Node->ClassLength;
Info->KeyNodeInformation.NameLength = NameLength;
/* Now the name is left */
SizeLeft = Length - MinimumSize;
Size = NameLength;
/* Check if the name can fit entirely */
if (SizeLeft < Size)
{
/* It can't, we'll have to truncate. Tell the caller */
Size = SizeLeft;
Status = STATUS_BUFFER_OVERFLOW;
}
/* Check if the key node name is compressed */
if (Node->Flags & KEY_COMP_NAME)
{
/* Copy the compressed name */
CmpCopyCompressedName(Info->KeyNodeInformation.Name,
SizeLeft,
Node->Name,
Node->NameLength);
}
else
{
/* It isn't, so copy the raw name */
RtlCopyMemory(Info->KeyNodeInformation.Name,
Node->Name,
Size);
}
/* Check if the node has a class */
if (Node->ClassLength > 0)
{
/* Set the class offset */
Offset = FIELD_OFFSET(KEY_NODE_INFORMATION, Name) + NameLength;
Offset = ALIGN_UP_BY(Offset, sizeof(ULONG));
Info->KeyNodeInformation.ClassOffset = Offset;
/* Get the class data */
ClassData = HvGetCell(Hive, Node->Class);
if (ClassData == NULL)
{
Status = STATUS_INSUFFICIENT_RESOURCES;
break;
}
/* Check if we can copy anything */
if (Length > Offset)
{
/* Copy the class data */
RtlCopyMemory((PUCHAR)Info + Offset,
ClassData,
min(Node->ClassLength, Length - Offset));
}
/* Check if the buffer was large enough */
if (Length < Offset + Node->ClassLength)
{
Status = STATUS_BUFFER_OVERFLOW;
}
/* Release the class cell */
HvReleaseCell(Hive, Node->Class);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
else
{
/* It doesn't, so set offset to -1, not 0! */
Info->KeyNodeInformation.ClassOffset = 0xFFFFFFFF;
}
break;
/* Full information requsted */
case KeyFullInformation:
/* This is the size we need */
Size = FIELD_OFFSET(KEY_FULL_INFORMATION, Class) +
Node->ClassLength;
/* This is what we can work with */
MinimumSize = FIELD_OFFSET(KEY_FULL_INFORMATION, Class);
/* Return it to caller and assume success */
*ResultLength = Size;
Status = STATUS_SUCCESS;
/* Check if the caller's buffer is to small */
if (Length < MinimumSize)
{
/* Let them know and fail */
Status = STATUS_BUFFER_TOO_SMALL;
break;
}
/* Now copy all the basic information */
Info->KeyFullInformation.LastWriteTime = Node->LastWriteTime;
Info->KeyFullInformation.TitleIndex = 0;
Info->KeyFullInformation.ClassLength = Node->ClassLength;
Info->KeyFullInformation.SubKeys = Node->SubKeyCounts[Stable] +
Node->SubKeyCounts[Volatile];
Info->KeyFullInformation.Values = Node->ValueList.Count;
Info->KeyFullInformation.MaxNameLen = Node->MaxNameLen;
Info->KeyFullInformation.MaxClassLen = Node->MaxClassLen;
Info->KeyFullInformation.MaxValueNameLen = Node->MaxValueNameLen;
Info->KeyFullInformation.MaxValueDataLen = Node->MaxValueDataLen;
/* Check if we have a class */
if (Node->ClassLength > 0)
{
/* Set the class offset */
Offset = FIELD_OFFSET(KEY_FULL_INFORMATION, Class);
Info->KeyFullInformation.ClassOffset = Offset;
/* Get the class data */
ClassData = HvGetCell(Hive, Node->Class);
if (ClassData == NULL)
{
Status = STATUS_INSUFFICIENT_RESOURCES;
break;
}
/* Copy the class data */
ASSERT(Length >= Offset);
RtlCopyMemory(Info->KeyFullInformation.Class,
ClassData,
min(Node->ClassLength, Length - Offset));
/* Check if the buffer was large enough */
if (Length < Offset + Node->ClassLength)
{
Status = STATUS_BUFFER_OVERFLOW;
}
/* Release the class cell */
HvReleaseCell(Hive, Node->Class);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
else
{
/* We don't have a class, so set offset to -1, not 0! */
Info->KeyFullInformation.ClassOffset = 0xFFFFFFFF;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
break;
/* Any other class that got sent here is invalid! */
default:
/* Set failure code */
Status = STATUS_INVALID_PARAMETER;
break;
}
/* Return status */
return Status;
}
NTSTATUS
NTAPI
CmSetValueKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN PUNICODE_STRING ValueName,
IN ULONG Type,
IN PVOID Data,
IN ULONG DataLength)
{
PHHIVE Hive = NULL;
PCM_KEY_NODE Parent;
PCM_KEY_VALUE Value = NULL;
HCELL_INDEX CurrentChild, Cell;
NTSTATUS Status;
BOOLEAN Found, Result;
ULONG Count, ChildIndex, SmallData, Storage;
VALUE_SEARCH_RETURN_TYPE SearchResult;
BOOLEAN FirstTry = TRUE, FlusherLocked = FALSE;
HCELL_INDEX ParentCell = HCELL_NIL, ChildCell = HCELL_NIL;
/* Acquire hive and KCB lock */
CmpLockRegistry();
CmpAcquireKcbLockShared(Kcb);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Sanity check */
ASSERT(sizeof(ULONG) == CM_KEY_VALUE_SMALL);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Don't touch deleted KCBs */
DoAgain:
if (Kcb->Delete)
{
/* Fail */
Status = STATUS_KEY_DELETED;
goto Quickie;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Don't let anyone mess with symlinks */
if ((Kcb->Flags & KEY_SYM_LINK) &&
((Type != REG_LINK) ||
!(ValueName) ||
!(RtlEqualUnicodeString(&CmSymbolicLinkValueName, ValueName, TRUE))))
{
/* Invalid modification of a symlink key */
Status = STATUS_ACCESS_DENIED;
goto Quickie;
}
/* Check if this is the first attempt */
if (FirstTry)
{
/* Search for the value in the cache */
SearchResult = CmpCompareNewValueDataAgainstKCBCache(Kcb,
ValueName,
Type,
Data,
DataLength);
if (SearchResult == SearchNeedExclusiveLock)
{
/* Try again with the exclusive lock */
CmpConvertKcbSharedToExclusive(Kcb);
goto DoAgain;
}
else if (SearchResult == SearchSuccess)
{
/* We don't actually need to do anything! */
Status = STATUS_SUCCESS;
goto Quickie;
}
/* We need the exclusive KCB lock now */
if (!CmpIsKcbLockedExclusive(Kcb) &&
!CmpTryToConvertKcbSharedToExclusive(Kcb))
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* Acquire exclusive lock */
CmpConvertKcbSharedToExclusive(Kcb);
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Cache lookup failed, so don't try it next time */
FirstTry = FALSE;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Now grab the flush lock since the key will be modified */
ASSERT(FlusherLocked == FALSE);
CmpLockHiveFlusherShared((PCMHIVE)Kcb->KeyHive);
FlusherLocked = TRUE;
goto DoAgain;
}
else
{
/* Get pointer to key cell */
Hive = Kcb->KeyHive;
Cell = Kcb->KeyCell;
/* Get the parent */
Parent = (PCM_KEY_NODE)HvGetCell(Hive, Cell);
ASSERT(Parent);
ParentCell = Cell;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Prepare to scan the key node */
Count = Parent->ValueList.Count;
Found = FALSE;
if (Count > 0)
{
/* Try to find the existing name */
Result = CmpFindNameInList(Hive,
&Parent->ValueList,
ValueName,
&ChildIndex,
&CurrentChild);
if (!Result)
{
/* Fail */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
/* Check if we found something */
if (CurrentChild != HCELL_NIL)
{
/* Release existing child */
if (ChildCell != HCELL_NIL)
{
HvReleaseCell(Hive, ChildCell);
ChildCell = HCELL_NIL;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Get its value */
Value = (PCM_KEY_VALUE)HvGetCell(Hive, CurrentChild);
if (!Value)
{
/* Fail */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
/* Remember that we found it */
ChildCell = CurrentChild;
Found = TRUE;
}
}
else
{
/* No child list, we'll need to add it */
ChildIndex = 0;
}
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Should only get here on the second pass */
ASSERT(FirstTry == FALSE);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* The KCB must be locked exclusive at this point */
CMP_ASSERT_KCB_LOCK(Kcb);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Mark the cell dirty */
if (!HvMarkCellDirty(Hive, Cell, FALSE))
{
/* Not enough log space, fail */
Status = STATUS_NO_LOG_SPACE;
goto Quickie;
}
/* Get the storage type */
Storage = HvGetCellType(Cell);
/* Check if this is small data */
SmallData = 0;
if ((DataLength <= CM_KEY_VALUE_SMALL) && (DataLength > 0))
{
/* Need SEH because user data may be invalid */
_SEH2_TRY
{
/* Copy it */
RtlCopyMemory(&SmallData, Data, DataLength);
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
/* Return failure code */
Status = _SEH2_GetExceptionCode();
_SEH2_YIELD(goto Quickie);
}
_SEH2_END;
}
/* Check if we didn't find a matching key */
if (!Found)
{
/* Call the internal routine */
Status = CmpSetValueKeyNew(Hive,
Parent,
ValueName,
ChildIndex,
Type,
Data,
DataLength,
Storage,
SmallData);
}
else
{
/* Call the internal routine */
Status = CmpSetValueKeyExisting(Hive,
CurrentChild,
Value,
Type,
Data,
DataLength,
Storage,
SmallData);
}
/* Check for success */
if (NT_SUCCESS(Status))
{
/* Check if the maximum value name length changed */
ASSERT(Parent->MaxValueNameLen == Kcb->KcbMaxValueNameLen);
if (Parent->MaxValueNameLen < ValueName->Length)
{
/* Set the new values */
Parent->MaxValueNameLen = ValueName->Length;
Kcb->KcbMaxValueNameLen = ValueName->Length;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check if the maximum data length changed */
ASSERT(Parent->MaxValueDataLen == Kcb->KcbMaxValueDataLen);
if (Parent->MaxValueDataLen < DataLength)
{
/* Update it */
Parent->MaxValueDataLen = DataLength;
Kcb->KcbMaxValueDataLen = Parent->MaxValueDataLen;
}
/* Save the write time */
KeQuerySystemTime(&Parent->LastWriteTime);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
Kcb->KcbLastWriteTime = Parent->LastWriteTime;
/* Check if the cell is cached */
if (Found && CMP_IS_CELL_CACHED(Kcb->ValueCache.ValueList))
{
/* Shouldn't happen */
ASSERT(FALSE);
}
else
{
/* Cleanup the value cache */
CmpCleanUpKcbValueCache(Kcb);
/* Sanity checks */
ASSERT(!CMP_IS_CELL_CACHED(Kcb->ValueCache.ValueList));
ASSERT(!(Kcb->ExtFlags & CM_KCB_SYM_LINK_FOUND));
/* Set the value cache */
Kcb->ValueCache.Count = Parent->ValueList.Count;
Kcb->ValueCache.ValueList = Parent->ValueList.List;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Notify registered callbacks */
CmpReportNotify(Kcb,
Hive,
Kcb->KeyCell,
REG_NOTIFY_CHANGE_LAST_SET);
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release the cells */
Quickie:
if ((ParentCell != HCELL_NIL) && Hive) HvReleaseCell(Hive, ParentCell);
if ((ChildCell != HCELL_NIL) && Hive) HvReleaseCell(Hive, ChildCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release the locks */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (FlusherLocked) CmpUnlockHiveFlusher((PCMHIVE)Hive);
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmDeleteValueKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN UNICODE_STRING ValueName)
{
NTSTATUS Status = STATUS_OBJECT_NAME_NOT_FOUND;
PHHIVE Hive;
PCM_KEY_NODE Parent;
HCELL_INDEX ChildCell, Cell;
PCHILD_LIST ChildList;
PCM_KEY_VALUE Value = NULL;
ULONG ChildIndex;
BOOLEAN Result;
/* Acquire hive lock */
CmpLockRegistry();
/* Lock KCB exclusively */
CmpAcquireKcbLockExclusive(Kcb);
/* Don't touch deleted keys */
if (Kcb->Delete)
{
/* Undo everything */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return STATUS_KEY_DELETED;
}
/* Get the hive and the cell index */
Hive = Kcb->KeyHive;
Cell = Kcb->KeyCell;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Lock flushes */
CmpLockHiveFlusherShared((PCMHIVE)Hive);
/* Get the parent key node */
Parent = (PCM_KEY_NODE)HvGetCell(Hive, Cell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
ASSERT(Parent);
/* Get the value list and check if it has any entries */
ChildList = &Parent->ValueList;
if (ChildList->Count)
{
/* Try to find this value */
Result = CmpFindNameInList(Hive,
ChildList,
&ValueName,
&ChildIndex,
&ChildCell);
if (!Result)
{
/* Fail */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
/* Value not found, return error */
if (ChildCell == HCELL_NIL) goto Quickie;
/* We found the value, mark all relevant cells dirty */
if (!(HvMarkCellDirty(Hive, Cell, FALSE) &&
HvMarkCellDirty(Hive, Parent->ValueList.List, FALSE) &&
HvMarkCellDirty(Hive, ChildCell, FALSE)))
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* Not enough log space, fail */
Status = STATUS_NO_LOG_SPACE;
goto Quickie;
}
/* Get the key value */
2018-10-13 23:20:31 +00:00
Value = (PCM_KEY_VALUE)HvGetCell(Hive, ChildCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
ASSERT(Value);
/* Mark it and all related data as dirty */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (!CmpMarkValueDataDirty(Hive, Value))
{
/* Not enough log space, fail */
Status = STATUS_NO_LOG_SPACE;
goto Quickie;
}
/* Sanity checks */
ASSERT(HvIsCellDirty(Hive, Parent->ValueList.List));
ASSERT(HvIsCellDirty(Hive, ChildCell));
/* Remove the value from the child list */
Status = CmpRemoveValueFromList(Hive, ChildIndex, ChildList);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (!NT_SUCCESS(Status))
{
/* Set known error */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
/* Remove the value and its data itself */
if (!CmpFreeValue(Hive, ChildCell))
{
/* Failed to free the value, fail */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
/* Set the last write time */
KeQuerySystemTime(&Parent->LastWriteTime);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
Kcb->KcbLastWriteTime = Parent->LastWriteTime;
/* Sanity check */
ASSERT(Parent->MaxValueNameLen == Kcb->KcbMaxValueNameLen);
ASSERT(Parent->MaxValueDataLen == Kcb->KcbMaxValueDataLen);
ASSERT(HvIsCellDirty(Hive, Cell));
/* Check if the value list is empty now */
if (!Parent->ValueList.Count)
{
/* Then clear key node data */
Parent->MaxValueNameLen = 0;
Parent->MaxValueDataLen = 0;
Kcb->KcbMaxValueNameLen = 0;
Kcb->KcbMaxValueDataLen = 0;
}
/* Cleanup the value cache */
CmpCleanUpKcbValueCache(Kcb);
/* Sanity checks */
ASSERT(!CMP_IS_CELL_CACHED(Kcb->ValueCache.ValueList));
ASSERT(!(Kcb->ExtFlags & CM_KCB_SYM_LINK_FOUND));
/* Set the value cache */
Kcb->ValueCache.Count = ChildList->Count;
Kcb->ValueCache.ValueList = ChildList->List;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Notify registered callbacks */
CmpReportNotify(Kcb, Hive, Cell, REG_NOTIFY_CHANGE_LAST_SET);
/* Change default Status to success */
Status = STATUS_SUCCESS;
}
Quickie:
/* Release the parent cell, if any */
if (Parent) HvReleaseCell(Hive, Cell);
/* Check if we had a value */
if (Value)
{
/* Release the child cell */
ASSERT(ChildCell != HCELL_NIL);
HvReleaseCell(Hive, ChildCell);
}
/* Release locks */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
CmpUnlockHiveFlusher((PCMHIVE)Hive);
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmQueryValueKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN UNICODE_STRING ValueName,
IN KEY_VALUE_INFORMATION_CLASS KeyValueInformationClass,
IN PVOID KeyValueInformation,
IN ULONG Length,
IN PULONG ResultLength)
{
NTSTATUS Status;
PCM_KEY_VALUE ValueData;
ULONG Index;
BOOLEAN ValueCached = FALSE;
PCM_CACHED_VALUE *CachedValue;
HCELL_INDEX CellToRelease;
VALUE_SEARCH_RETURN_TYPE Result;
PHHIVE Hive;
PAGED_CODE();
/* Acquire hive lock */
CmpLockRegistry();
/* Lock the KCB shared */
CmpAcquireKcbLockShared(Kcb);
/* Don't touch deleted keys */
DoAgain:
if (Kcb->Delete)
{
/* Undo everything */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return STATUS_KEY_DELETED;
}
/* We don't deal with this yet */
if (Kcb->ExtFlags & CM_KCB_SYM_LINK_FOUND)
{
/* Shouldn't happen */
ASSERT(FALSE);
}
/* Get the hive */
Hive = Kcb->KeyHive;
/* Find the key value */
Result = CmpFindValueByNameFromCache(Kcb,
&ValueName,
&CachedValue,
&Index,
&ValueData,
&ValueCached,
&CellToRelease);
if (Result == SearchNeedExclusiveLock)
{
/* Check if we need an exclusive lock */
ASSERT(CellToRelease == HCELL_NIL);
ASSERT(ValueData == NULL);
/* Try with exclusive KCB lock */
CmpConvertKcbSharedToExclusive(Kcb);
goto DoAgain;
}
if (Result == SearchSuccess)
{
/* Sanity check */
ASSERT(ValueData != NULL);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* User data, protect against exceptions */
_SEH2_TRY
{
/* Query the information requested */
Result = CmpQueryKeyValueData(Kcb,
CachedValue,
ValueData,
ValueCached,
KeyValueInformationClass,
KeyValueInformation,
Length,
ResultLength,
&Status);
if (Result == SearchNeedExclusiveLock)
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release the value cell */
if (CellToRelease != HCELL_NIL)
{
HvReleaseCell(Hive, CellToRelease);
CellToRelease = HCELL_NIL;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Try with exclusive KCB lock */
CmpConvertKcbSharedToExclusive(Kcb);
_SEH2_YIELD(goto DoAgain);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
Status = _SEH2_GetExceptionCode();
}
_SEH2_END;
}
else
{
/* Failed to find the value */
Status = STATUS_OBJECT_NAME_NOT_FOUND;
}
/* If we have a cell to release, do so */
if (CellToRelease != HCELL_NIL) HvReleaseCell(Hive, CellToRelease);
/* Release locks */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmEnumerateValueKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN ULONG Index,
IN KEY_VALUE_INFORMATION_CLASS KeyValueInformationClass,
IN PVOID KeyValueInformation,
IN ULONG Length,
IN PULONG ResultLength)
{
NTSTATUS Status;
PHHIVE Hive;
PCM_KEY_NODE Parent;
HCELL_INDEX CellToRelease = HCELL_NIL, CellToRelease2 = HCELL_NIL;
VALUE_SEARCH_RETURN_TYPE Result;
BOOLEAN IndexIsCached, ValueIsCached = FALSE;
PCELL_DATA CellData;
PCM_CACHED_VALUE *CachedValue;
PCM_KEY_VALUE ValueData = NULL;
PAGED_CODE();
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Acquire hive lock */
CmpLockRegistry();
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Lock the KCB shared */
CmpAcquireKcbLockShared(Kcb);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Don't touch deleted keys */
DoAgain:
if (Kcb->Delete)
{
/* Undo everything */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return STATUS_KEY_DELETED;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Get the hive and parent */
Hive = Kcb->KeyHive;
Parent = (PCM_KEY_NODE)HvGetCell(Hive, Kcb->KeyCell);
ASSERT(Parent);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* FIXME: Lack of cache? */
if (Kcb->ValueCache.Count != Parent->ValueList.Count)
{
DPRINT1("HACK: Overriding value cache count\n");
Kcb->ValueCache.Count = Parent->ValueList.Count;
}
/* Make sure the index is valid */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
if (Index >= Kcb->ValueCache.Count)
{
/* Release the cell and fail */
HvReleaseCell(Hive, Kcb->KeyCell);
Status = STATUS_NO_MORE_ENTRIES;
goto Quickie;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* We don't deal with this yet */
if (Kcb->ExtFlags & CM_KCB_SYM_LINK_FOUND)
{
/* Shouldn't happen */
ASSERT(FALSE);
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Find the value list */
Result = CmpGetValueListFromCache(Kcb,
&CellData,
&IndexIsCached,
&CellToRelease);
if (Result == SearchNeedExclusiveLock)
{
/* Check if we need an exclusive lock */
ASSERT(CellToRelease == HCELL_NIL);
HvReleaseCell(Hive, Kcb->KeyCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Try with exclusive KCB lock */
CmpConvertKcbSharedToExclusive(Kcb);
goto DoAgain;
}
else if (Result != SearchSuccess)
{
/* Sanity check */
ASSERT(CellData == NULL);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release the cell and fail */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Now get the key value */
Result = CmpGetValueKeyFromCache(Kcb,
CellData,
Index,
&CachedValue,
&ValueData,
IndexIsCached,
&ValueIsCached,
&CellToRelease2);
if (Result == SearchNeedExclusiveLock)
{
/* Cleanup state */
ASSERT(CellToRelease2 == HCELL_NIL);
if (CellToRelease)
{
HvReleaseCell(Hive, CellToRelease);
CellToRelease = HCELL_NIL;
}
HvReleaseCell(Hive, Kcb->KeyCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Try with exclusive KCB lock */
CmpConvertKcbSharedToExclusive(Kcb);
goto DoAgain;
}
else if (Result != SearchSuccess)
{
/* Sanity check */
ASSERT(ValueData == NULL);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release the cells and fail */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* User data, need SEH */
_SEH2_TRY
{
/* Query the information requested */
Result = CmpQueryKeyValueData(Kcb,
CachedValue,
ValueData,
ValueIsCached,
KeyValueInformationClass,
KeyValueInformation,
Length,
ResultLength,
&Status);
if (Result == SearchNeedExclusiveLock)
{
/* Cleanup state */
if (CellToRelease2) HvReleaseCell(Hive, CellToRelease2);
HvReleaseCell(Hive, Kcb->KeyCell);
if (CellToRelease) HvReleaseCell(Hive, CellToRelease);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Try with exclusive KCB lock */
CmpConvertKcbSharedToExclusive(Kcb);
_SEH2_YIELD(goto DoAgain);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
/* Get exception code */
Status = _SEH2_GetExceptionCode();
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
_SEH2_END;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
Quickie:
/* If we have a cell to release, do so */
if (CellToRelease != HCELL_NIL) HvReleaseCell(Hive, CellToRelease);
/* Release the parent cell */
HvReleaseCell(Hive, Kcb->KeyCell);
/* If we have a cell to release, do so */
if (CellToRelease2 != HCELL_NIL) HvReleaseCell(Hive, CellToRelease2);
/* Release locks */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
static
NTSTATUS
CmpQueryKeyDataFromCache(
_In_ PCM_KEY_CONTROL_BLOCK Kcb,
_Out_ PKEY_CACHED_INFORMATION KeyCachedInfo,
_In_ ULONG Length,
_Out_ PULONG ResultLength)
{
PCM_KEY_NODE Node;
PHHIVE KeyHive;
HCELL_INDEX KeyCell;
USHORT NameLength;
PAGED_CODE();
/* Get the hive and cell index */
KeyHive = Kcb->KeyHash.KeyHive;
KeyCell = Kcb->KeyHash.KeyCell;
#if DBG
/* Get the cell node */
Node = (PCM_KEY_NODE)HvGetCell(KeyHive, KeyCell);
if (Node != NULL)
{
ULONG SubKeyCount;
ASSERT(Node->ValueList.Count == Kcb->ValueCache.Count);
if (!(Kcb->ExtFlags & CM_KCB_INVALID_CACHED_INFO))
{
SubKeyCount = Node->SubKeyCounts[0] + Node->SubKeyCounts[1];
if (Kcb->ExtFlags & CM_KCB_NO_SUBKEY)
{
ASSERT(SubKeyCount == 0);
}
else if (Kcb->ExtFlags & CM_KCB_SUBKEY_ONE)
{
ASSERT(SubKeyCount == 1);
}
else if (Kcb->ExtFlags & CM_KCB_SUBKEY_HINT)
{
ASSERT(SubKeyCount == Kcb->IndexHint->Count);
}
else
{
ASSERT(SubKeyCount == Kcb->SubKeyCount);
}
}
ASSERT(Node->LastWriteTime.QuadPart == Kcb->KcbLastWriteTime.QuadPart);
ASSERT(Node->MaxNameLen == Kcb->KcbMaxNameLen);
ASSERT(Node->MaxValueNameLen == Kcb->KcbMaxValueNameLen);
ASSERT(Node->MaxValueDataLen == Kcb->KcbMaxValueDataLen);
/* Release the cell */
HvReleaseCell(KeyHive, KeyCell);
}
#endif // DBG
/* Make sure we have a name block */
if (Kcb->NameBlock == NULL)
{
return STATUS_INSUFFICIENT_RESOURCES;
}
/* Check for compressed name */
if (Kcb->NameBlock->Compressed)
{
/* Calculate the name size */
NameLength = CmpCompressedNameSize(Kcb->NameBlock->NameHash.Name,
Kcb->NameBlock->NameHash.NameLength);
}
else
{
/* Use the stored name size */
NameLength = Kcb->NameBlock->NameHash.NameLength;
}
/* Validate buffer length (we do not copy the name!) */
*ResultLength = sizeof(*KeyCachedInfo);
if (Length < *ResultLength)
{
return STATUS_BUFFER_TOO_SMALL;
}
/* Fill the structure */
KeyCachedInfo->LastWriteTime = Kcb->KcbLastWriteTime;
KeyCachedInfo->TitleIndex = 0;
KeyCachedInfo->NameLength = NameLength;
KeyCachedInfo->Values = Kcb->ValueCache.Count;
KeyCachedInfo->MaxNameLen = Kcb->KcbMaxNameLen;
KeyCachedInfo->MaxValueNameLen = Kcb->KcbMaxValueNameLen;
KeyCachedInfo->MaxValueDataLen = Kcb->KcbMaxValueDataLen;
/* Check the ExtFlags for what we have */
if (Kcb->ExtFlags & CM_KCB_INVALID_CACHED_INFO)
{
/* Cache is not valid, do a full lookup */
DPRINT1("Kcb cache incoherency detected, kcb = %p\n", Kcb);
/* Get the cell node */
Node = (PCM_KEY_NODE)HvGetCell(KeyHive, KeyCell);
if (Node == NULL)
{
return STATUS_INSUFFICIENT_RESOURCES;
}
/* Calculate number of subkeys */
KeyCachedInfo->SubKeys = Node->SubKeyCounts[0] + Node->SubKeyCounts[1];
/* Release the cell */
HvReleaseCell(KeyHive, KeyCell);
}
else if (Kcb->ExtFlags & CM_KCB_NO_SUBKEY)
{
/* There are no subkeys */
KeyCachedInfo->SubKeys = 0;
}
else if (Kcb->ExtFlags & CM_KCB_SUBKEY_ONE)
{
/* There is exactly one subley */
KeyCachedInfo->SubKeys = 1;
}
else if (Kcb->ExtFlags & CM_KCB_SUBKEY_HINT)
{
/* Get the number of subkeys from the subkey hint */
KeyCachedInfo->SubKeys = Kcb->IndexHint->Count;
}
else
{
/* No subkey hint, use the key count field */
KeyCachedInfo->SubKeys = Kcb->SubKeyCount;
}
return STATUS_SUCCESS;
}
static
NTSTATUS
CmpQueryFlagsInformation(
_In_ PCM_KEY_CONTROL_BLOCK Kcb,
_Out_ PKEY_USER_FLAGS_INFORMATION KeyFlagsInfo,
_In_ ULONG Length,
_In_ PULONG ResultLength)
{
/* Validate the buffer size */
*ResultLength = sizeof(*KeyFlagsInfo);
if (Length < *ResultLength)
{
return STATUS_BUFFER_TOO_SMALL;
}
/* Copy the user flags */
KeyFlagsInfo->UserFlags = Kcb->KcbUserFlags;
return STATUS_SUCCESS;
}
static
NTSTATUS
CmpQueryNameInformation(
_In_ PCM_KEY_CONTROL_BLOCK Kcb,
_Out_opt_ PKEY_NAME_INFORMATION KeyNameInfo,
_In_ ULONG Length,
_Out_ PULONG ResultLength)
{
ULONG NeededLength;
PCM_KEY_CONTROL_BLOCK CurrentKcb;
NeededLength = 0;
CurrentKcb = Kcb;
/* Count the needed buffer size */
while (CurrentKcb)
{
if (CurrentKcb->NameBlock->Compressed)
NeededLength += CmpCompressedNameSize(CurrentKcb->NameBlock->Name, CurrentKcb->NameBlock->NameLength);
else
NeededLength += CurrentKcb->NameBlock->NameLength;
NeededLength += sizeof(OBJ_NAME_PATH_SEPARATOR);
CurrentKcb = CurrentKcb->ParentKcb;
}
_SEH2_TRY
{
*ResultLength = FIELD_OFFSET(KEY_NAME_INFORMATION, Name) + NeededLength;
if (Length < RTL_SIZEOF_THROUGH_FIELD(KEY_NAME_INFORMATION, NameLength))
_SEH2_YIELD(return STATUS_BUFFER_TOO_SMALL);
if (Length < *ResultLength)
{
KeyNameInfo->NameLength = NeededLength;
_SEH2_YIELD(return STATUS_BUFFER_OVERFLOW);
}
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
_SEH2_YIELD(return _SEH2_GetExceptionCode());
}
_SEH2_END;
/* Do the real copy */
KeyNameInfo->NameLength = 0;
CurrentKcb = Kcb;
_SEH2_TRY
{
while (CurrentKcb)
{
ULONG NameLength;
if (CurrentKcb->NameBlock->Compressed)
{
NameLength = CmpCompressedNameSize(CurrentKcb->NameBlock->Name, CurrentKcb->NameBlock->NameLength);
/* Copy the compressed name */
CmpCopyCompressedName(&KeyNameInfo->Name[(NeededLength - NameLength)/sizeof(WCHAR)],
NameLength,
CurrentKcb->NameBlock->Name,
CurrentKcb->NameBlock->NameLength);
}
else
{
NameLength = CurrentKcb->NameBlock->NameLength;
/* Otherwise, copy the raw name */
RtlCopyMemory(&KeyNameInfo->Name[(NeededLength - NameLength)/sizeof(WCHAR)],
CurrentKcb->NameBlock->Name,
NameLength);
}
NeededLength -= NameLength;
NeededLength -= sizeof(OBJ_NAME_PATH_SEPARATOR);
/* Add path separator */
KeyNameInfo->Name[NeededLength/sizeof(WCHAR)] = OBJ_NAME_PATH_SEPARATOR;
KeyNameInfo->NameLength += NameLength + sizeof(OBJ_NAME_PATH_SEPARATOR);
CurrentKcb = CurrentKcb->ParentKcb;
}
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
_SEH2_YIELD(return _SEH2_GetExceptionCode());
}
_SEH2_END;
/* Make sure we copied everything */
ASSERT(NeededLength == 0);
ASSERT(KeyNameInfo->Name[0] == OBJ_NAME_PATH_SEPARATOR);
/* We're done */
return STATUS_SUCCESS;
}
NTSTATUS
NTAPI
CmQueryKey(_In_ PCM_KEY_CONTROL_BLOCK Kcb,
_In_ KEY_INFORMATION_CLASS KeyInformationClass,
_Out_opt_ PVOID KeyInformation,
_In_ ULONG Length,
_Out_ PULONG ResultLength)
{
NTSTATUS Status;
PHHIVE Hive;
PCM_KEY_NODE Parent;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
HV_TRACK_CELL_REF CellReferences = {0};
/* Acquire hive lock */
CmpLockRegistry();
/* Lock KCB shared */
CmpAcquireKcbLockShared(Kcb);
/* Don't touch deleted keys */
if (Kcb->Delete)
{
/* Fail */
Status = STATUS_KEY_DELETED;
goto Quickie;
}
/* Data can be user-mode, use SEH */
_SEH2_TRY
{
/* Check what class we got */
switch (KeyInformationClass)
{
/* Typical information */
case KeyFullInformation:
case KeyBasicInformation:
case KeyNodeInformation:
{
/* Get the hive and parent */
Hive = Kcb->KeyHive;
Parent = (PCM_KEY_NODE)HvGetCell(Hive, Kcb->KeyCell);
ASSERT(Parent);
/* Track cell references */
if (!HvTrackCellRef(&CellReferences, Hive, Kcb->KeyCell))
{
/* Not enough memory to track references */
Status = STATUS_INSUFFICIENT_RESOURCES;
}
else
{
/* Call the internal API */
Status = CmpQueryKeyData(Hive,
Parent,
KeyInformationClass,
KeyInformation,
Length,
ResultLength);
}
break;
}
case KeyCachedInformation:
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* Call the internal API */
Status = CmpQueryKeyDataFromCache(Kcb,
KeyInformation,
Length,
ResultLength);
break;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
case KeyFlagsInformation:
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* Call the internal API */
Status = CmpQueryFlagsInformation(Kcb,
KeyInformation,
Length,
ResultLength);
break;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
case KeyNameInformation:
{
/* Call the internal API */
Status = CmpQueryNameInformation(Kcb,
KeyInformation,
Length,
ResultLength);
break;
}
/* Illegal classes */
default:
{
/* Print message and fail */
DPRINT1("Unsupported class: %d!\n", KeyInformationClass);
Status = STATUS_INVALID_INFO_CLASS;
break;
}
}
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
/* Fail with exception code */
Status = _SEH2_GetExceptionCode();
_SEH2_YIELD(goto Quickie);
}
_SEH2_END;
Quickie:
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release references */
HvReleaseFreeCellRefArray(&CellReferences);
/* Release locks */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmEnumerateKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN ULONG Index,
IN KEY_INFORMATION_CLASS KeyInformationClass,
IN PVOID KeyInformation,
IN ULONG Length,
IN PULONG ResultLength)
{
NTSTATUS Status;
PHHIVE Hive;
PCM_KEY_NODE Parent, Child;
HCELL_INDEX ChildCell;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
HV_TRACK_CELL_REF CellReferences = {0};
/* Acquire hive lock */
CmpLockRegistry();
/* Lock the KCB shared */
CmpAcquireKcbLockShared(Kcb);
/* Don't touch deleted keys */
if (Kcb->Delete)
{
/* Undo everything */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
Status = STATUS_KEY_DELETED;
goto Quickie;
}
/* Get the hive and parent */
Hive = Kcb->KeyHive;
Parent = (PCM_KEY_NODE)HvGetCell(Hive, Kcb->KeyCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
ASSERT(Parent);
/* Get the child cell */
ChildCell = CmpFindSubKeyByNumber(Hive, Parent, Index);
/* Release the parent cell */
HvReleaseCell(Hive, Kcb->KeyCell);
/* Check if we found the child */
if (ChildCell == HCELL_NIL)
{
/* We didn't, fail */
Status = STATUS_NO_MORE_ENTRIES;
goto Quickie;
}
/* Now get the actual child node */
Child = (PCM_KEY_NODE)HvGetCell(Hive, ChildCell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
ASSERT(Child);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Track references */
if (!HvTrackCellRef(&CellReferences, Hive, ChildCell))
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Can't allocate memory for tracking */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Quickie;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Data can be user-mode, use SEH */
_SEH2_TRY
{
/* Query the data requested */
Status = CmpQueryKeyData(Hive,
Child,
KeyInformationClass,
KeyInformation,
Length,
ResultLength);
}
_SEH2_EXCEPT(EXCEPTION_EXECUTE_HANDLER)
{
/* Fail with exception code */
Status = _SEH2_GetExceptionCode();
_SEH2_YIELD(goto Quickie);
}
_SEH2_END;
Quickie:
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release references */
HvReleaseFreeCellRefArray(&CellReferences);
/* Release locks */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmDeleteKey(IN PCM_KEY_BODY KeyBody)
{
NTSTATUS Status;
PHHIVE Hive;
PCM_KEY_NODE Node, Parent;
HCELL_INDEX Cell, ParentCell;
PCM_KEY_CONTROL_BLOCK Kcb;
/* Acquire hive lock */
CmpLockRegistry();
/* Get the kcb */
Kcb = KeyBody->KeyControlBlock;
/* Don't allow deleting the root */
if (!Kcb->ParentKcb)
{
/* Fail */
CmpUnlockRegistry();
return STATUS_CANNOT_DELETE;
}
/* Lock parent and child */
CmpAcquireTwoKcbLocksExclusiveByKey(Kcb->ConvKey, Kcb->ParentKcb->ConvKey);
/* Check if we're already being deleted */
if (Kcb->Delete)
{
/* Don't do it twice */
Status = STATUS_SUCCESS;
2018-10-13 23:20:31 +00:00
goto Quickie;
}
/* Get the hive and node */
Hive = Kcb->KeyHive;
Cell = Kcb->KeyCell;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Lock flushes */
CmpLockHiveFlusherShared((PCMHIVE)Hive);
/* Get the key node */
Node = (PCM_KEY_NODE)HvGetCell(Hive, Cell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
ASSERT(Node);
/* Sanity check */
ASSERT(Node->Flags == Kcb->Flags);
/* Check if we don't have any children */
if (!(Node->SubKeyCounts[Stable] + Node->SubKeyCounts[Volatile]) &&
!(Node->Flags & KEY_NO_DELETE))
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Send notification to registered callbacks */
CmpReportNotify(Kcb, Hive, Cell, REG_NOTIFY_CHANGE_NAME);
/* Get the parent and free the cell */
ParentCell = Node->Parent;
Status = CmpFreeKeyByCell(Hive, Cell, TRUE);
if (NT_SUCCESS(Status))
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Flush any notifications */
CmpFlushNotifiesOnKeyBodyList(Kcb, FALSE);
/* Clean up information we have on the subkey */
CmpCleanUpSubKeyInfo(Kcb->ParentKcb);
/* Get the parent node */
Parent = (PCM_KEY_NODE)HvGetCell(Hive, ParentCell);
if (Parent)
{
/* Update the maximum name length */
Kcb->ParentKcb->KcbMaxNameLen = (USHORT)Parent->MaxNameLen;
/* Make sure we're dirty */
ASSERT(HvIsCellDirty(Hive, ParentCell));
/* Update the write time */
KeQuerySystemTime(&Parent->LastWriteTime);
Kcb->ParentKcb->KcbLastWriteTime = Parent->LastWriteTime;
/* Release the cell */
HvReleaseCell(Hive, ParentCell);
}
/* Set the KCB in delete mode and remove it */
Kcb->Delete = TRUE;
CmpRemoveKeyControlBlock(Kcb);
/* Clear the cell */
Kcb->KeyCell = HCELL_NIL;
}
}
else
{
/* Fail */
Status = STATUS_CANNOT_DELETE;
}
/* Release the cell */
HvReleaseCell(Hive, Cell);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release flush lock */
CmpUnlockHiveFlusher((PCMHIVE)Hive);
/* Release the KCB locks */
2018-10-13 23:20:31 +00:00
Quickie:
CmpReleaseTwoKcbLockByKey(Kcb->ConvKey, Kcb->ParentKcb->ConvKey);
/* Release hive lock */
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmFlushKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN BOOLEAN ExclusiveLock)
{
PCMHIVE CmHive;
#if DBG
CM_CHECK_REGISTRY_STATUS CheckStatus;
#endif
NTSTATUS Status = STATUS_SUCCESS;
PHHIVE Hive;
/* Ignore flushes until we're ready */
if (CmpNoWrite) return STATUS_SUCCESS;
/* Get the hives */
Hive = Kcb->KeyHive;
CmHive = (PCMHIVE)Hive;
/* Check if this is the master hive */
if (CmHive == CmiVolatileHive)
{
/* Flush all the hives instead */
CmpDoFlushAll(FALSE);
}
else
{
#if DBG
/* Make sure the registry hive we're going to flush is OK */
CheckStatus = CmCheckRegistry(CmHive, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
#endif
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Don't touch the hive */
CmpLockHiveFlusherExclusive(CmHive);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
ASSERT(CmHive->ViewLock);
KeAcquireGuardedMutex(CmHive->ViewLock);
CmHive->ViewLockOwner = KeGetCurrentThread();
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Will the hive shrink? */
if (HvHiveWillShrink(Hive))
{
/* I don't believe the current Hv does shrinking */
ASSERT(FALSE);
// CMP_ASSERT_EXCLUSIVE_REGISTRY_LOCK_OR_LOADING(CmHive);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
else
{
/* Now we can release views */
ASSERT(CmHive->ViewLock);
// CMP_ASSERT_VIEW_LOCK_OWNED(CmHive);
ASSERT((CmpSpecialBootCondition == TRUE) ||
(CmHive->HiveIsLoading == TRUE) ||
(CmHive->ViewLockOwner == KeGetCurrentThread()) ||
(CmpTestRegistryLockExclusive() == TRUE));
CmHive->ViewLockOwner = NULL;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
KeReleaseGuardedMutex(CmHive->ViewLock);
}
/* Flush only this hive */
if (!HvSyncHive(Hive))
{
/* Fail */
Status = STATUS_REGISTRY_IO_FAILED;
}
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Release the flush lock */
CmpUnlockHiveFlusher(CmHive);
}
/* Return the status */
return Status;
}
NTSTATUS
NTAPI
CmLoadKey(IN POBJECT_ATTRIBUTES TargetKey,
IN POBJECT_ATTRIBUTES SourceFile,
IN ULONG Flags,
IN PCM_KEY_BODY KeyBody)
{
SECURITY_QUALITY_OF_SERVICE ServiceQos;
SECURITY_CLIENT_CONTEXT ClientSecurityContext;
HANDLE KeyHandle;
BOOLEAN Allocate = TRUE;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
PCMHIVE CmHive, LoadedHive;
NTSTATUS Status;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
CM_PARSE_CONTEXT ParseContext;
/* Check if we have a trust key */
if (KeyBody)
{
/* Fail */
DPRINT("Trusted classes not yet supported\n");
}
/* Build a service QoS for a security context */
ServiceQos.Length = sizeof(SECURITY_QUALITY_OF_SERVICE);
ServiceQos.ImpersonationLevel = SecurityImpersonation;
ServiceQos.ContextTrackingMode = SECURITY_DYNAMIC_TRACKING;
ServiceQos.EffectiveOnly = TRUE;
Status = SeCreateClientSecurity(PsGetCurrentThread(),
&ServiceQos,
FALSE,
&ClientSecurityContext);
if (!NT_SUCCESS(Status))
{
/* Fail */
DPRINT1("Security context failed\n");
return Status;
}
/* Open the target key */
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
RtlZeroMemory(&ParseContext, sizeof(ParseContext));
ParseContext.CreateOperation = FALSE;
Status = ObOpenObjectByName(TargetKey,
CmpKeyObjectType,
KernelMode,
NULL,
KEY_READ,
&ParseContext,
&KeyHandle);
if (!NT_SUCCESS(Status)) KeyHandle = NULL;
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Open the hive */
Status = CmpCmdHiveOpen(SourceFile,
&ClientSecurityContext,
&Allocate,
&CmHive,
CM_CHECK_REGISTRY_PURGE_VOLATILES);
/* Get rid of the security context */
SeDeleteClientSecurity(&ClientSecurityContext);
/* See if we failed */
if (!NT_SUCCESS(Status))
{
/* See if the target already existed */
if (KeyHandle)
{
/* Lock the registry */
CmpLockRegistryExclusive();
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Check if we are already loaded */
if (CmpIsHiveAlreadyLoaded(KeyHandle, SourceFile, &LoadedHive))
{
/* That's okay then */
ASSERT(LoadedHive);
Status = STATUS_SUCCESS;
}
/* Release the registry */
CmpUnlockRegistry();
}
/* Close the key handle if we had one */
if (KeyHandle) ZwClose(KeyHandle);
return Status;
}
/* Lock the registry shared */
CmpLockRegistry();
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Lock loading */
ExAcquirePushLockExclusive(&CmpLoadHiveLock);
/* Lock the hive to this thread */
CmHive->Hive.HiveFlags |= HIVE_IS_UNLOADING;
CmHive->CreatorOwner = KeGetCurrentThread();
/* Set flag */
if (Flags & REG_NO_LAZY_FLUSH) CmHive->Hive.HiveFlags |= HIVE_NOLAZYFLUSH;
/* Link the hive */
Status = CmpLinkHiveToMaster(TargetKey->ObjectName,
TargetKey->RootDirectory,
CmHive,
Allocate,
TargetKey->SecurityDescriptor);
if (NT_SUCCESS(Status))
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Add to HiveList key */
CmpAddToHiveFileList(CmHive);
/* Sync the hive if necessary */
if (Allocate)
{
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Sync it under the flusher lock */
CmpLockHiveFlusherExclusive(CmHive);
HvSyncHive(&CmHive->Hive);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
CmpUnlockHiveFlusher(CmHive);
}
/* Release the hive */
CmHive->Hive.HiveFlags &= ~HIVE_IS_UNLOADING;
CmHive->CreatorOwner = NULL;
}
else
{
DPRINT1("CmpLinkHiveToMaster failed, Status %lx\n", Status);
/* We're touching this hive, set the loading flag */
CmHive->HiveIsLoading = TRUE;
/* Close associated file handles */
CmpCloseHiveFiles(CmHive);
/* Cleanup its resources */
CmpDestroyHive(CmHive);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
}
/* Allow loads */
ExReleasePushLock(&CmpLoadHiveLock);
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
/* Is this first profile load? */
if (!CmpProfileLoaded && !CmpWasSetupBoot)
[NTOSKRNL/CONFIG] - Flusher lock fixes: wrong kind of lock,total mess (and the wrong kind of lock). Properly fixed throughout cmapi.c, but still missing in many other places. - Add support for detecting loading of an already loaded hive. - Start adding calls to CmpReportNotify to support registry callbacks. - Do work needed to flush notifications for a deleted node (but CmpFlushNotify not yet implemented). - Add support for adding each newly loaded hive to the HiveList key in the registry (but CmpAddHiveToFileList not yet implemented). - Add some ViewLock acquire/releases where needed. - Load the key in a faster way (Ob vs Zw) - Add checks everywhere for HvMarkCellDirty success. In future (when log/backup file is enabled), it can return FALSE (e.g. when we are out of space). - Change logic in CmpDoFlushAll to only flush a hive if it won't shrink (in the future, flushing may lead to hive shrinkage for efficiency). - Add SEH2 protection to all CmApis that may deal with user-mode data. - Add HvTrackCellRef/HvReleaseCellRef for tracking cell references in scenarios where we might need many GetCell/ReleaseCell calls. For now stubbed to only work with up to 4 static references. - Properly unlock/free in some failure paths in some of the CM APIs. - Add some missing HvReleaseCell in paths where it was missing. - Try to fix hack in enumerate key. - Fix wrong usage of KeQuerySystemTime. It was called twice to save it in 2 different places. Instead, there should be only one call, and then duplicate the value across. - Fix logic in CmpSetValueExistingData/Key. Tested with winetests and .NET framework 1.1 installation which fully completes. svn path=/trunk/; revision=46702
2010-04-03 20:22:32 +00:00
{
/* User is now logged on, set quotas */
CmpProfileLoaded = TRUE;
CmpSetGlobalQuotaAllowed();
}
/* Unlock the registry */
CmpUnlockRegistry();
/* Close handle and return */
if (KeyHandle) ZwClose(KeyHandle);
return Status;
}
static
BOOLEAN
NTAPI
CmpUnlinkHiveFromMaster(IN PCMHIVE CmHive,
IN HCELL_INDEX Cell)
{
PCELL_DATA CellData;
HCELL_INDEX LinkCell;
NTSTATUS Status;
DPRINT("CmpUnlinkHiveFromMaster()\n");
/* Get the cell data */
CellData = HvGetCell(&CmHive->Hive, Cell);
if (CellData == NULL)
return FALSE;
/* Get the link cell and release the current cell */
LinkCell = CellData->u.KeyNode.Parent;
HvReleaseCell(&CmHive->Hive, Cell);
/* Remove the link cell from the master hive */
CmpLockHiveFlusherExclusive(CmiVolatileHive);
Status = CmpFreeKeyByCell((PHHIVE)CmiVolatileHive,
LinkCell,
TRUE);
CmpUnlockHiveFlusher(CmiVolatileHive);
if (!NT_SUCCESS(Status))
{
DPRINT1("CmpFreeKeyByCell() failed (Status 0x%08lx)\n", Status);
return FALSE;
}
/* Remove the hive from the list */
ExAcquirePushLockExclusive(&CmpHiveListHeadLock);
RemoveEntryList(&CmHive->HiveList);
ExReleasePushLock(&CmpHiveListHeadLock);
return TRUE;
}
NTSTATUS
NTAPI
CmUnloadKey(
_In_ PCM_KEY_CONTROL_BLOCK Kcb,
_In_ ULONG Flags)
{
PHHIVE Hive;
PCMHIVE CmHive;
HCELL_INDEX Cell;
DPRINT("CmUnloadKey(%p, %lx)\n", Kcb, Flags);
/* Ensure the registry is locked exclusively for the calling thread */
CMP_ASSERT_EXCLUSIVE_REGISTRY_LOCK();
/* Get the hive */
Hive = Kcb->KeyHive;
Cell = Kcb->KeyCell;
CmHive = (PCMHIVE)Hive;
/* Fail if the key is not a hive root key */
if (Cell != Hive->BaseBlock->RootCell)
{
DPRINT1("Key is not a hive root key!\n");
return STATUS_INVALID_PARAMETER;
}
/* Fail if we try to unload the master hive */
if (CmHive == CmiVolatileHive)
{
DPRINT1("Do not try to unload the master hive!\n");
return STATUS_INVALID_PARAMETER;
}
/* Mark this hive as being unloaded */
Hive->HiveFlags |= HIVE_IS_UNLOADING;
/* Search for any opened keys in this hive, and take an appropriate action */
if (Kcb->RefCount > 1)
{
if (Flags != REG_FORCE_UNLOAD)
{
if (CmpEnumerateOpenSubKeys(Kcb, TRUE, FALSE) != 0)
{
/* There are open subkeys but we don't force hive unloading, fail */
Hive->HiveFlags &= ~HIVE_IS_UNLOADING;
return STATUS_CANNOT_DELETE;
}
}
else
{
if (CmpEnumerateOpenSubKeys(Kcb, TRUE, TRUE) != 0)
{
/* There are open subkeys that we cannot force to unload, fail */
Hive->HiveFlags &= ~HIVE_IS_UNLOADING;
return STATUS_CANNOT_DELETE;
}
}
}
/* Set the loading flag */
CmHive->HiveIsLoading = TRUE;
/* Flush the hive */
CmFlushKey(Kcb, TRUE);
/* Unlink the hive from the master hive */
if (!CmpUnlinkHiveFromMaster(CmHive, Cell))
{
DPRINT("CmpUnlinkHiveFromMaster() failed!\n");
/* Remove the unloading flag */
Hive->HiveFlags &= ~HIVE_IS_UNLOADING;
/* Reset the loading flag */
CmHive->HiveIsLoading = FALSE;
/* Return failure */
return STATUS_INSUFFICIENT_RESOURCES;
}
/* Flush any notifications if we force hive unloading */
if (Flags == REG_FORCE_UNLOAD)
CmpFlushNotifiesOnKeyBodyList(Kcb, TRUE); // Lock is already held
/* Clean up information we have on the subkey */
CmpCleanUpSubKeyInfo(Kcb->ParentKcb);
/* Set the KCB in delete mode and remove it */
Kcb->Delete = TRUE;
CmpRemoveKeyControlBlock(Kcb);
/* Release the hive loading lock */
ExReleasePushLockExclusive(&CmpLoadHiveLock);
/* Release hive lock */
CmpUnlockRegistry();
/* Close file handles */
CmpCloseHiveFiles(CmHive);
/* Remove the hive from the hive file list */
CmpRemoveFromHiveFileList(CmHive);
/**
** NOTE:
** The following code is mostly equivalent to what we "call" CmpDestroyHive()
**/
/* Destroy the security descriptor cache */
CmpDestroySecurityCache(CmHive);
/* Destroy the view list */
CmpDestroyHiveViewList(CmHive);
/* Delete the flusher lock */
ExDeleteResourceLite(CmHive->FlusherLock);
ExFreePoolWithTag(CmHive->FlusherLock, TAG_CMHIVE);
/* Delete the view lock */
ExFreePoolWithTag(CmHive->ViewLock, TAG_CMHIVE);
/* Free the hive storage */
HvFree(Hive);
/* Free the hive */
CmpFree(CmHive, TAG_CM);
return STATUS_SUCCESS;
}
ULONG
NTAPI
CmpEnumerateOpenSubKeys(
_In_ PCM_KEY_CONTROL_BLOCK RootKcb,
_In_ BOOLEAN RemoveEmptyCacheEntries,
_In_ BOOLEAN DereferenceOpenedEntries)
{
PCM_KEY_HASH Entry;
PCM_KEY_CONTROL_BLOCK CachedKcb;
PCM_KEY_CONTROL_BLOCK ParentKcb;
ULONG ParentKeyCount;
ULONG i, j;
ULONG SubKeys = 0;
DPRINT("CmpEnumerateOpenSubKeys() called\n");
/* Ensure the registry is locked exclusively for the calling thread */
CMP_ASSERT_EXCLUSIVE_REGISTRY_LOCK();
/* The root key is the only referenced key. There are no referenced sub keys. */
if (RootKcb->RefCount == 1)
{
DPRINT("Open sub keys: 0\n");
return 0;
}
/* Enumerate all hash lists */
for (i = 0; i < CmpHashTableSize; i++)
{
/* Get the first cache entry */
Entry = CmpCacheTable[i].Entry;
/* Enumerate all cache entries */
while (Entry)
{
/* Get the KCB of the current cache entry */
CachedKcb = CONTAINING_RECORD(Entry, CM_KEY_CONTROL_BLOCK, KeyHash);
/* Check keys only that are subkeys to our root key */
if (CachedKcb->TotalLevels > RootKcb->TotalLevels)
{
/* Calculate the number of parent keys to the root key */
ParentKeyCount = CachedKcb->TotalLevels - RootKcb->TotalLevels;
/* Find a parent key that could be the root key */
ParentKcb = CachedKcb;
for (j = 0; j < ParentKeyCount; j++)
{
ParentKcb = ParentKcb->ParentKcb;
}
/* Check whether the parent is the root key */
if (ParentKcb == RootKcb)
{
DPRINT("Found a sub key, RefCount = %u\n", CachedKcb->RefCount);
if (CachedKcb->RefCount > 0)
{
2018-10-27 08:04:22 +00:00
DPRINT("Found a sub key pointing to '%.*s', RefCount = %u\n",
CachedKcb->NameBlock->NameLength, CachedKcb->NameBlock->Name,
CachedKcb->RefCount);
/* If we dereference opened KCBs, don't touch read-only keys */
if (DereferenceOpenedEntries &&
!(CachedKcb->ExtFlags & CM_KCB_READ_ONLY_KEY))
{
/* Flush any notifications */
CmpFlushNotifiesOnKeyBodyList(CachedKcb, TRUE); // Lock is already held
/* Clean up information we have on the subkey */
CmpCleanUpSubKeyInfo(CachedKcb->ParentKcb);
/* Get and cache the next cache entry */
// Entry = Entry->NextHash;
Entry = CachedKcb->NextHash;
/* Set the KCB in delete mode and remove it */
CachedKcb->Delete = TRUE;
CmpRemoveKeyControlBlock(CachedKcb);
/* Clear the cell */
CachedKcb->KeyCell = HCELL_NIL;
/* Restart with the next cache entry */
continue;
}
/* Else, the key cannot be dereferenced, and we count it as in use */
/* Count the current hash entry if it is in use */
SubKeys++;
}
else if ((CachedKcb->RefCount == 0) && RemoveEmptyCacheEntries)
{
/* Remove the current key from the delayed close list */
CmpRemoveFromDelayedClose(CachedKcb);
/* Remove the current cache entry */
CmpCleanUpKcbCacheWithLock(CachedKcb, TRUE);
/* Restart, because the hash list has changed */
Entry = CmpCacheTable[i].Entry;
continue;
}
}
}
/* Get the next cache entry */
Entry = Entry->NextHash;
}
}
if (SubKeys > 0)
DPRINT1("Open sub keys: %u\n", SubKeys);
return SubKeys;
}
static
NTSTATUS
CmpDeepCopyKeyInternal(IN PHHIVE SourceHive,
IN HCELL_INDEX SrcKeyCell,
IN PHHIVE DestinationHive,
IN HCELL_INDEX Parent,
IN HSTORAGE_TYPE StorageType,
OUT PHCELL_INDEX DestKeyCell OPTIONAL)
{
NTSTATUS Status;
PCM_KEY_NODE SrcNode;
PCM_KEY_NODE DestNode = NULL;
HCELL_INDEX NewKeyCell = HCELL_NIL;
HCELL_INDEX NewClassCell = HCELL_NIL, NewSecCell = HCELL_NIL;
HCELL_INDEX SubKey, NewSubKey;
ULONG Index, SubKeyCount;
PAGED_CODE();
DPRINT("CmpDeepCopyKeyInternal(0x%08X, 0x%08X, 0x%08X, 0x%08X, 0x%08X, 0x%08X)\n",
SourceHive,
SrcKeyCell,
DestinationHive,
Parent,
StorageType,
DestKeyCell);
/* Get the source cell node */
SrcNode = (PCM_KEY_NODE)HvGetCell(SourceHive, SrcKeyCell);
ASSERT(SrcNode);
/* Sanity check */
ASSERT(SrcNode->Signature == CM_KEY_NODE_SIGNATURE);
/* Create a simple copy of the source key */
NewKeyCell = CmpCopyCell(SourceHive,
SrcKeyCell,
DestinationHive,
StorageType);
if (NewKeyCell == HCELL_NIL)
{
/* Not enough storage space */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Cleanup;
}
/* Get the destination cell node */
DestNode = (PCM_KEY_NODE)HvGetCell(DestinationHive, NewKeyCell);
ASSERT(DestNode);
/* Set the parent and copy the flags */
DestNode->Parent = Parent;
DestNode->Flags = (SrcNode->Flags & KEY_COMP_NAME); // Keep only the single permanent flag
if (Parent == HCELL_NIL)
{
/* This is the new root node */
DestNode->Flags |= KEY_HIVE_ENTRY | KEY_NO_DELETE;
}
/* Copy the class cell */
if (SrcNode->ClassLength > 0)
{
NewClassCell = CmpCopyCell(SourceHive,
SrcNode->Class,
DestinationHive,
StorageType);
if (NewClassCell == HCELL_NIL)
{
/* Not enough storage space */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Cleanup;
}
DestNode->Class = NewClassCell;
DestNode->ClassLength = SrcNode->ClassLength;
}
else
{
DestNode->Class = HCELL_NIL;
DestNode->ClassLength = 0;
}
/* Copy the security cell (FIXME: HACKish poor-man version) */
if (SrcNode->Security != HCELL_NIL)
{
NewSecCell = CmpCopyCell(SourceHive,
SrcNode->Security,
DestinationHive,
StorageType);
if (NewSecCell == HCELL_NIL)
{
/* Not enough storage space */
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Cleanup;
}
}
DestNode->Security = NewSecCell;
/* Copy the value list */
Status = CmpCopyKeyValueList(SourceHive,
&SrcNode->ValueList,
DestinationHive,
&DestNode->ValueList,
StorageType);
if (!NT_SUCCESS(Status))
goto Cleanup;
/* Clear the invalid subkey index */
DestNode->SubKeyCounts[Stable] = DestNode->SubKeyCounts[Volatile] = 0;
DestNode->SubKeyLists[Stable] = DestNode->SubKeyLists[Volatile] = HCELL_NIL;
/* Calculate the total number of subkeys */
SubKeyCount = SrcNode->SubKeyCounts[Stable] + SrcNode->SubKeyCounts[Volatile];
/* Loop through all the subkeys */
for (Index = 0; Index < SubKeyCount; Index++)
{
/* Get the subkey */
SubKey = CmpFindSubKeyByNumber(SourceHive, SrcNode, Index);
ASSERT(SubKey != HCELL_NIL);
/* Call the function recursively for the subkey */
//
// FIXME: Danger!! Kernel stack exhaustion!!
//
Status = CmpDeepCopyKeyInternal(SourceHive,
SubKey,
DestinationHive,
NewKeyCell,
StorageType,
&NewSubKey);
if (!NT_SUCCESS(Status))
goto Cleanup;
/* Add the copy of the subkey to the new key */
if (!CmpAddSubKey(DestinationHive,
NewKeyCell,
NewSubKey))
{
/* Cleanup allocated cell */
HvFreeCell(DestinationHive, NewSubKey);
Status = STATUS_INSUFFICIENT_RESOURCES;
goto Cleanup;
}
}
/* Set success */
Status = STATUS_SUCCESS;
Cleanup:
/* Release the cells */
if (DestNode) HvReleaseCell(DestinationHive, NewKeyCell);
if (SrcNode) HvReleaseCell(SourceHive, SrcKeyCell);
/* Cleanup allocated cells in case of failure */
if (!NT_SUCCESS(Status))
{
if (NewSecCell != HCELL_NIL)
HvFreeCell(DestinationHive, NewSecCell);
if (NewClassCell != HCELL_NIL)
HvFreeCell(DestinationHive, NewClassCell);
if (NewKeyCell != HCELL_NIL)
HvFreeCell(DestinationHive, NewKeyCell);
NewKeyCell = HCELL_NIL;
}
/* Set the cell index if requested and return status */
if (DestKeyCell) *DestKeyCell = NewKeyCell;
return Status;
}
NTSTATUS
NTAPI
CmpDeepCopyKey(IN PHHIVE SourceHive,
IN HCELL_INDEX SrcKeyCell,
IN PHHIVE DestinationHive,
IN HSTORAGE_TYPE StorageType,
OUT PHCELL_INDEX DestKeyCell OPTIONAL)
{
/* Call the internal function */
return CmpDeepCopyKeyInternal(SourceHive,
SrcKeyCell,
DestinationHive,
HCELL_NIL,
StorageType,
DestKeyCell);
}
NTSTATUS
NTAPI
CmSaveKey(IN PCM_KEY_CONTROL_BLOCK Kcb,
IN HANDLE FileHandle,
IN ULONG Flags)
{
#if DBG
CM_CHECK_REGISTRY_STATUS CheckStatus;
PCMHIVE HiveToValidate = NULL;
#endif
NTSTATUS Status = STATUS_SUCCESS;
PCMHIVE KeyHive = NULL;
PAGED_CODE();
DPRINT("CmSaveKey(0x%08X, 0x%08X, %lu)\n", Kcb, FileHandle, Flags);
/* Lock the registry and KCB */
CmpLockRegistry();
CmpAcquireKcbLockShared(Kcb);
#if DBG
/* Get the hive for validation */
HiveToValidate = (PCMHIVE)Kcb->KeyHive;
#endif
if (Kcb->Delete)
{
/* The source key has been deleted, do nothing */
Status = STATUS_KEY_DELETED;
goto Cleanup;
}
if (Kcb->KeyHive == &CmiVolatileHive->Hive)
{
/* Keys that are directly in the master hive can't be saved */
Status = STATUS_ACCESS_DENIED;
goto Cleanup;
}
#if DBG
/* Make sure this control block has a sane hive */
CheckStatus = CmCheckRegistry(HiveToValidate, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
#endif
/* Create a new hive that will hold the key */
Status = CmpInitializeHive(&KeyHive,
HINIT_CREATE,
HIVE_VOLATILE,
HFILE_TYPE_PRIMARY,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES);
if (!NT_SUCCESS(Status)) goto Cleanup;
/* Copy the key recursively into the new hive */
Status = CmpDeepCopyKey(Kcb->KeyHive,
Kcb->KeyCell,
&KeyHive->Hive,
Stable,
&KeyHive->Hive.BaseBlock->RootCell);
if (!NT_SUCCESS(Status)) goto Cleanup;
/* Set the primary handle of the hive */
KeyHive->FileHandles[HFILE_TYPE_PRIMARY] = FileHandle;
/* Dump the hive into the file */
HvWriteHive(&KeyHive->Hive);
Cleanup:
/* Free the hive */
if (KeyHive) CmpDestroyHive(KeyHive);
#if DBG
if (NT_SUCCESS(Status))
{
/* Before we say goodbye, make sure the hive is still OK */
CheckStatus = CmCheckRegistry(HiveToValidate, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
}
#endif
/* Release the locks */
CmpReleaseKcbLock(Kcb);
CmpUnlockRegistry();
return Status;
}
NTSTATUS
NTAPI
CmSaveMergedKeys(IN PCM_KEY_CONTROL_BLOCK HighKcb,
IN PCM_KEY_CONTROL_BLOCK LowKcb,
IN HANDLE FileHandle)
{
#if DBG
CM_CHECK_REGISTRY_STATUS CheckStatus;
PCMHIVE LowHiveToValidate = NULL;
PCMHIVE HighHiveToValidate = NULL;
#endif
PCMHIVE KeyHive = NULL;
NTSTATUS Status = STATUS_SUCCESS;
PAGED_CODE();
DPRINT("CmSaveKey(%p, %p, %p)\n", HighKcb, LowKcb, FileHandle);
/* Lock the registry and the KCBs */
CmpLockRegistry();
CmpAcquireKcbLockShared(HighKcb);
CmpAcquireKcbLockShared(LowKcb);
#if DBG
/* Get the high and low hives for validation */
HighHiveToValidate = (PCMHIVE)HighKcb->KeyHive;
LowHiveToValidate = (PCMHIVE)LowKcb->KeyHive;
#endif
if (LowKcb->Delete || HighKcb->Delete)
{
/* The source key has been deleted, do nothing */
Status = STATUS_KEY_DELETED;
goto done;
}
#if DBG
/* Make sure that both the high and low precedence hives are OK */
CheckStatus = CmCheckRegistry(HighHiveToValidate, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
CheckStatus = CmCheckRegistry(LowHiveToValidate, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
#endif
/* Create a new hive that will hold the key */
Status = CmpInitializeHive(&KeyHive,
HINIT_CREATE,
HIVE_VOLATILE,
HFILE_TYPE_PRIMARY,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES);
if (!NT_SUCCESS(Status))
goto done;
/* Copy the low precedence key recursively into the new hive */
Status = CmpDeepCopyKey(LowKcb->KeyHive,
LowKcb->KeyCell,
&KeyHive->Hive,
Stable,
&KeyHive->Hive.BaseBlock->RootCell);
if (!NT_SUCCESS(Status))
goto done;
/* Copy the high precedence key recursively into the new hive */
Status = CmpDeepCopyKey(HighKcb->KeyHive,
HighKcb->KeyCell,
&KeyHive->Hive,
Stable,
&KeyHive->Hive.BaseBlock->RootCell);
if (!NT_SUCCESS(Status))
goto done;
/* Set the primary handle of the hive */
KeyHive->FileHandles[HFILE_TYPE_PRIMARY] = FileHandle;
/* Dump the hive into the file */
HvWriteHive(&KeyHive->Hive);
done:
/* Free the hive */
if (KeyHive)
CmpDestroyHive(KeyHive);
#if DBG
if (NT_SUCCESS(Status))
{
/* Check those hives again before we say goodbye */
CheckStatus = CmCheckRegistry(HighHiveToValidate, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
CheckStatus = CmCheckRegistry(LowHiveToValidate, CM_CHECK_REGISTRY_DONT_PURGE_VOLATILES | CM_CHECK_REGISTRY_VALIDATE_HIVE);
ASSERT(CM_CHECK_REGISTRY_SUCCESS(CheckStatus));
}
#endif
/* Release the locks */
CmpReleaseKcbLock(LowKcb);
CmpReleaseKcbLock(HighKcb);
CmpUnlockRegistry();
return Status;
}