Before any write operation that would involve caching, ask
the cache controler whether writing would make it exceed its memory
consumption. If so, queue the write operation for later execution.
In case the write operation can wait, then, the FSD operation will be
halted until the write is allowed.
I could test it successfully by copying huge files from a FAT volume to
another. The write is halted until some portions of the file is written
to the disk.
I could also properly install Qt (SDK) on ReactOS with this and less than 1GB RAM:
- https://www.heisspiter.net/~Pierre/rostests/Qt_OS.png
- https://www.heisspiter.net/~Pierre/rostests/Qt_OS2.png
CORE-12081
CORE-14582
CORE-14313
This is a PoC of what it's possible to realize thanks to an
already existing hack in ntoskrnl :-).
With this extension, on the kdb:> prompt, you're able to type
in commands that will be handled by the FastFAT driver and not
by the kernel, allowing internal debug, not possible otherwise.
So far, three commands exist:
- ?fat.vols: lists all the mounted volumes by FastFAT
- ?fat.files: lists all the files on a specific volume (with their attributes)
- ?fat.setdbgfile: allows watching on specifics files lifetime
This is obviously only the begin and could be greatly improved.
For instance, this is what allowed to debug CORE-14557
Once a directory is crossed (opened or a child is opened), associated
FCB structure is created in FastFAT, but also a stream FO for caching.
Up to now, due to an extra reference taken by the stream file object,
even when the directory was no longer used, the directory was kept in
memory: the FCB was never deleted, the file object was never dereferenced,
and the cache never released.
The immediate effect of this bug is that our FAT driver was leaking every
directory that was used affecting the whole OS situation. In case of
directories intensive operation (like extraction the ReactOS source code
in ReactOS ;-)), we were just killin the whole OS RAM without any way to
release it and recover.
The other side effects: IOs were faster as half of the FS was always
permanant in RAM.
This commit fixes the issue by forcing the FSD to release the FO,
and the cache when a directory is no longer used, leading to its
destruction in RAM.
Downside: on IO intensive operation, expect slowdowns, obviously,
there's less caching now. But more efficient!
CORE-14557
This makes the function return the error instead of continuing and
performing use-after-frees operations.
This is likely to be a forgotten return!
CID 1434254, 1434268
is set before locking is actually attempted for Cc worker thread (lazy write/read ahead).
So in case locking fails, the top level IRP is set and never unset, and latter
call will hit the assert where it expects the top level IRP to be NULL.
Should be reported upstream (tm).
CORE-14315
This is assorted with misc fixes to make this code closer to what we currently have in FastFAT.
This also allows fixing a memory leak in case of single entry return.
CORE-13367
leaves volumes in half-(dis)mounted state and thus they get remounted while not
completely dismounted.
This can lead to major breakage and data corruption.
This requires deeper fixes (let's just drop that driver!).
CORE-14124
CORE-14126
CORE-14133
This avoids issues when these partitions are formatted to FAT for setup.
For now, this commit doesn't change anything, but once IopParseDevice hack
gets removed, this will make a difference!
CORE-6305
It fixes assertion failure in vfatDestroyFCB() where we would have release parent before child.
This is still not perfect, but less bug prone...
With this commits (and ENABLE_SWAPOUT defined), ReactOS seems to unmount FAT volumes quite nice! :-)
(Tried with fsutil volume dismount X:)
AddNewMftEntry() - Zero the (larger) buffer for the bitmap prior to reading the bitmap.
BrowseIndexEntries() - Check sub-nodes before checking an index entry. Read and use the index bitmap when checking sub-nodes.
+BrowseSubNodeIndexEntries() - Called for sub-nodes of an index when browsing index entries.
+NtfsDumpData() - Diagnostic function which allows for visualizing a series of bytes.
svn path=/branches/GSoC_2016/NTFS/; revision=75817
+AddBitmap() - adds a $BITMAP attribute to a file record.
+AddIndexAllocation() - adds an $INDEX_ALLOCATION attribute to a file record.
+CountBTreeKeys() - Counts the number of linked B-Tree keys.
CreateIndexBufferFromBTreeNode() - Set INDEX_NODE_LARGE if the node has sub-nodes.
CreateIndexRootFromBTree() - Simplify the usage and math of MaxIndexSize; make it only account for the cumulative size of the index entries.
+DemoteBTreeRoot() - Replaces the contents of an index root with a dummy key, and puts those contents in a new node, which is made a child of the dummy key. This is done when an index root grows too large.
+GetIndexEntryVCN() - Retrieves the VCN from an index entry.
NtfsAddFilenameToDirectory() - Fix math for MaxIndexRootSize.
NtfsInsertKey() - Add support for splitting a B-Tree node. Don't check size of index root (that will be handled later).
+SplitBTreeNode() - Called when a B-Tree node grows too large.
UpdateIndexAllocation() - Create an $I30 index allocation attribute and bitmap attribute if needed.
UpdateIndexNode() - Update children before updating the current node. Store VCN of child nodes in the index entries of their respective keys.
svn path=/branches/GSoC_2016/NTFS/; revision=75707
IncreaseMftSize() - Add some fixes. Write blank records to newly-allocated mft entries, and update $MFTMirr when finished; these changes are needed for chkdsk. Increase size by 64 records instead of 8.
+UpdateMftMirror() - Backs up the first ~4 master file table entries to the $MFTMirr file.
svn path=/branches/GSoC_2016/NTFS/; revision=75694
+AddIndexRoot() - Creates an $INDEX_ROOT attribute and adds it to a file record.
AddNewMftEntry() - Make sure the buffer used by RtlInitializeBitmap() is ULONG-aligned, and a ULONG-multiple in size, per MSDN.
AllocateIndexNode() - Calculate BytesNeeded correctly. Read $BITMAP attribute before increasing its length, in anticipation of a future commit that will check for a free bit before assigning a new index record to the end of the allocation. Use appropriate Set*AttributeDataLength() function, as $BITMAP can be resident or non-resident.
B_TREE_FILENAME_NODE - Give two members more accurate names: change "ExistsOnDisk" member to "HasValidVCN" and rename "NodeNumber" member "VCN."
+CreateEmptyBTree() - Creates a B-Tree to represent an empty directory (for AddIndexRoot).
+NtfsCreateEmptyFileRecord() - Creates an empty file record in memory, with no attributes.
CreateIndexRootFromBTree() - Fix TotalSizeOfEntries calculation.
+NtfsCreateDirectory() - Creates a file record for an empty directory and adds it to the mft.
svn path=/branches/GSoC_2016/NTFS/; revision=75692
-Don't add a preceding backslash when creating files on root.
-Use NTFS_FILE_NAME_POSIX name type if CaseSensitive option is specified.
-Don't try to create a file when a folder is requested (folder creation is still TODO).
svn path=/branches/GSoC_2016/NTFS/; revision=75671
+AllocateIndexNode() - Allocates a new index record in an index allocation.
+CreateDummyKey() - Creates the final B_TREE_KEY for a B_TREE_FILENAME_NODE. Also creates the associated index entry.
GetSizeOfIndexEntries() - Sums the size of each index entry in every key in a B-Tree node.
+SetIndexEntryVCN() - Sets the VCN of a given IndexEntry.
NtfsInsertKey() - Handle instance when the index root grows too large. If it does, add its contents to a new sub-node, and replace contents with a dummy-key whose child is the new node.
UpdateIndexAllocation() - Update index entry if a key has just been assigned a child allocation.
UpdateIndexNode() - Make sure the node exists on disk, and allocate an index record for it if it doesn't.
svn path=/branches/GSoC_2016/NTFS/; revision=75557
AddRun() - Allow for resizing the size of the data runs when the attribute isn't the last in the file record. Fix some comments.
CreateIndexBufferFromBTreeNode(), CreateIndexRootFromBTree - Fix math of IndexSize when checking if the index buffer is too large.
InternalSetResidentAttributeLength() - Allow changing the length of an attribute in the middle of a file record. Adjust the position of every attribute after the one being resized.
+MoveAttributes() - Moves a block of attributes to a new location in the file Record.
PrintAllVCNs() - Add consideration for an index allocation with a size of 0.
WriteAttribute() - Add optional parameter for a pointer to the file record being written to. If passed a file record, WriteAttribute() will skip reading the file record from disk, and will update the file record in memory before returning. This helps callers that use the file record after writing an attribute to stay in-sync with what's on disk.
svn path=/branches/GSoC_2016/NTFS/; revision=75554
-PrepareAttributeContext() - update to allocate memory for pRecord. Don't assume allocations are succeeding.
-ReleaseAttributeContext() - update to free memory for pRecord.
-InternalSetResidentAttributeLength() - Increase size of AttrContext->pRecord as needed. Update to return an NTSTATUS.
-SetResidentAttributeDataLength() - Fix bug that could occur when migrating resident attributes to non-resident if AttrContext->pRecord is too small for the new attribute.
-AddRun() - Fix a bug by reallocating AttrContext->pRecord if the record needs to be enlarged.
svn path=/branches/GSoC_2016/NTFS/; revision=75493
