Ext2 write step 6: Truncation, metadata updates, and crash consistency

Summary

Implement file truncation, metadata timestamp updates, crash consistency ordering, and the RO_COMPAT feature check. This is the final step for ext2 write support.

This is step 6 of ext2 write support (#40). Depends on step 5 (#55).

Changes

1. Implement Filesystem::truncate

• Shrinking: Free blocks beyond the new size. Zero the partial last block if the new size is not block-aligned. Walk and free indirect blocks that are no longer needed.
• Growing: Allocate new blocks (or leave sparse holes with zero block pointers).
• Update inode size, blocks, and mtime. Write inode back to disk.

2. Metadata timestamp updates

• Update mtime on write and truncate
• Update ctime on metadata changes (chmod, link count changes)
• atime on read is optional — can be deferred or skipped (most modern systems use noatime or relatime)
• Timestamp writes can be batched and written lazily (pairs well with write coalescing #38)

3. Crash consistency ordering

Enforce a write ordering discipline to minimize corruption on crash:

1. Allocate in bitmaps and write bitmap blocks
2. Write data blocks
3. Write/update inode
4. Update directory entry

This ensures:

• Crash after bitmap update but before inode update → leaked blocks (fixable by fsck, no dangling references)
• Crash after inode update but before directory entry → orphaned inode (also fixable by fsck)

4. Implement Filesystem::sync

Flush all pending metadata writes to disk. Ensures all superblock, group descriptor, and inode updates are persisted.

5. RO_COMPAT feature check

When mounting read-write, check feature_ro_compat in the superblock. If any unsupported RO_COMPAT features are set, refuse to mount read-write (fall back to read-only). This prevents corrupting filesystems that use features the driver does not understand.

Key files

• panda-kernel/src/vfs/ext2/mod.rs — truncate, sync, RO_COMPAT check, timestamp updates
• panda-kernel/src/vfs/ext2/bitmap.rs — block freeing during truncation

Testing

• Truncate a file to 0, verify reads return empty and blocks are freed
• Truncate a file to a smaller size, verify data beyond new size is gone
• Truncate a file to a larger size, verify reads beyond old size return zeros (sparse)
• Verify stat reflects updated size and timestamps after truncate
• Verify sync does not error on a clean filesystem
• Verify RO_COMPAT check: create an ext2 image with an unsupported feature flag, verify write operations are refused
• Run all existing ext2 tests (read and write) to verify no regressions

Documentation

• Document the truncation algorithm (block freeing walk)
• Document the crash consistency guarantees and their limitations (no journal — fsck required after crash)
• Document the write ordering discipline with a diagram or numbered steps
• Document which RO_COMPAT features are recognized and which cause read-only fallback
• Update docs/ if any ext2 documentation exists

Part of #40

Start by reading panda-kernel/src/vfs/ext2/mod.rs and panda-kernel/src/vfs/ext2/bitmap.rs, then run the existing ext2 tests. Done means truncation, timestamps, sync, crash-ordering behavior, and the RO_COMPAT check meet the listed cases, with ext2 documentation updated and no regressions in read or write tests.