Do not merge yet - My proof of concept for AOT compatible DiscUtils#54
Draft
devedse wants to merge 20 commits into
Draft
Do not merge yet - My proof of concept for AOT compatible DiscUtils#54devedse wants to merge 20 commits into
devedse wants to merge 20 commits into
Conversation
Member
There was a problem hiding this comment.
This seems to be wrong in several ways. It needs to implement reading from other locations of the hive file outside of current bin object, so _buffer cannot be used to get all data. There are also a few errors in offset calculations, and I am also not sure whether the total length calculation is really reliable. I will try to find out how this can be resolved in some way.
- Updated version to 0.17 in version.json - Added new SourceGenerator project with FactoryGenerator - Improved .gitignore to exclude additional build outputs - Modified Directory.Build.props for versioning and compiler settings - Registered new LogicalVolumeFactory and VirtualDiskTransport instances - Enhanced logging in various classes for better traceability - Created generate-packages.ps1 script for automated package generation
Modernize and refactor plugin/registration system for disk image and file system types: - Bump version to 1.0.85 - Add AoT-friendly registration path for DiskImageBuilder - Simplify and streamline FileSystemManager and VirtualDiskManager initialization - Remove console logging and exception handling from factory discovery - Make LogicalVolumeFactory internal - Make VolumeManager registration internal and direct - Add SetupCompleteAot for AoT environments - General cleanup for improved startup and reliability
devedse
force-pushed
the
LTRData.DiscUtils-initial
branch
from
277a7b4 to
5e3e244
Compare
…p to 1.0.104 During idle periods an iSCSI target can close the session due to the DefaultTime2Retain/DefaultTime2Wait timeout expiring while no traffic is flowing. This causes the next I/O to fail with a broken connection rather than a clean retry. Changes in Connection.cs: - SemaphoreSlim (_streamSemaphore) serialises all PDU traffic so the keepalive timer and the main Send/SendAsync/Logout paths never interleave on the NetworkStream. - Timer (_keepAliveTimer) fires every 10 s after login; skips ticks when a real command is in-flight. - SendNopOut: ITT=0xFFFFFFFF / TTT=0xFFFFFFFF so no NOP-In response is expected (RFC 3720 §10.18). - HandleNopIn: responds to target-initiated NOP-In (TTT!=0xFFFFFFFF) without advancing ExpectedStatusSequenceNumber (RFC 3720 §10.19). - ReadPdu/ReadPduAsync: loop over consecutive NOP-In PDUs (max 16) before returning the next real response PDU. Version bump: 1.0.103 -> 1.0.104
Resolved conflicts: - Directory.Build.props: keep fork version 1.0.104 (fork maintains its own NuGet version sequence, above upstream's 1.0.83). - Library/DiscUtils.Iscsi/Connection.cs: take upstream. Both sides implemented the same NOP-Out keepalive; upstream's is async (SendNopOutAsync) and its Send/SendAsync brace structure is self-consistent (the fork side would double-nest the for loop after auto-merge). Functionally equivalent keepalive; no fork logic lost. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…xed ri/li lists SubKeyIndirectListCell backs both "ri" index-root lists (entries are subordinate lists) and "li" leaf-index lists (entries are key nodes). The read paths (Count, EnumerateKeyNames, EnumerateKeys, DoFindKey, KeyFinder) already tolerate a single list whose entries are a mix of sublists and key nodes, by dispatching on the actual cell type. But LinkSubKey/UnlinkSubKey still branched on ListType and hard-cast every entry (GetCell<ListCell> for "ri", GetCell<KeyNodeCell> for "li"). On a real Windows SYSTEM hive whose "ri" list contains a direct key-node entry, deleting a subkey (UnlinkSubKey) threw: System.InvalidCastException: Unable to cast object of type 'DiscUtils.Registry.KeyNodeCell' to type 'DiscUtils.Registry.ListCell'. Reading/enumerating the same key worked (those paths were already hardened); only add/delete crashed. Fix: LinkSubKey/UnlinkSubKey now dispatch on the actual cell type per entry (GetCell<Cell> + "is ListCell"), exactly as the read paths do. Behaviour for homogeneous "ri"/"li" lists is unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Includes the cherry-picked SubKeyIndirectListCell fix (InvalidCastException on deleting/adding keys in mixed ri/li subkey lists). Triggers a fork build so DevePXEBoot can consume the registry delete fix before it lands upstream (PR LTRData#69). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…stream) Registry hive cells are 8-byte aligned. The check must reject sizes whose low 3 bits are set: (size & 0x7) != 0. Upstream LTRData.DiscUtils-initial had regressed this to (size % 0x7) != 0 (mod 7), which rejects ~6 of every 7 valid sizes - including the ~80-byte root key cell - so RegistryHive.Create() itself throws "Invalid cell size" and every registry/BootConfig test fails (and the build hangs). The fork's pre-merge code used (size % 8); this merge pulled in the regression. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ies (regressed in a551635) GetFileLength must honor the NtfsOptions.FileLengthFromDirectoryEntries option: when set (the default), a default-data-stream length request returns the directory entry's cached RealSize. Commit a551635 ("Bugfixes for FileExist and GetFileLength") removed that shortcut, so GetFileLength always looked up the data attribute and the option became a no-op - breaking NtfsFileSystemTest.GetFileLength (and the async variant), which expects a hardlink's stale directory-entry size (14325) by default and the live data-attribute size (50) only when the option is disabled. Restores the shortcut. The FileExists fix in that commit is correct and left intact. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
FindNextPresentSector/FindNextAbsentSector advance with pos += (8 - sectorInBlock & 0x7) * Sizes.Sector. C# precedence makes this (8 - sectorInBlock) & 0x7, not 8 - (sectorInBlock & 0x7): when sectorInBlock is byte-aligned and the block-bitmap byte is empty, the original (8 - sectorInBlock % 8) advanced by 8 sectors, but the rewrite yields 8 & 7 = 0, so pos never advances and the extent enumeration (LayerExtents/Extents and the dynamic VHD/VMDK builders) loops forever. Commit f7bf755 ("Minor code cleanups and optimizations") introduced this by rewriting % 8 to & 0x7 without parentheses. Parenthesized so the mask applies to sectorInBlock (the 7 - x & 0x7 mask forms are coincidentally equivalent but were parenthesized too for clarity). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…b5b75d) The synchronized wrapper kept its own Position auto-property and set content.Position = Position before each operation but never wrote the advanced position back afterward, so Position stayed frozen and reads/writes through the wrapper re-processed the same offset (e.g. ExFat ReadLongSparseLimited read the wrong data). Delegate Position to content.Position under the lock - correct for this single-shared- cursor synchronized wrapper and fixes every read/write path (sync and async) at once. Introduced by 4b5b75d ("Synchronized wrappers for disks and streams"). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…o LTRData.DiscUtils-initial
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Draft PR with some proof of concepts regarding AOT compatible disc utils