[PATCH v5 0/2] hfsplus: prevent b-tree allocator corruption

[Shardul Bankar]

Hi all,

This series addresses a Syzkaller-reported vulnerability where fuzzed
HFS+ images mark the B-tree Header Node (Node 0) as free in the
allocation bitmap. This violates a core filesystem invariant and
leads to allocator corruption and kernel panics.

To fix this safely and cleanly, the series is split into two parts:

Patch 1 introduces a unified API for B-tree map record access
(struct hfs_bmap_ctx, hfs_bmap_test_bit, and hfs_bmap_clear_bit) and
refactors the boilerplate kmap_local_page() logic out of
hfs_bmap_alloc() and hfs_bmap_free().
Patch 2 utilizes this new API to perform a mount-time validation of
Node 0, forcing a safe read-only mount if structural or bit-level
corruption is detected.

Note on Allocator Optimization: Following discussions in v4, there is
a recognized opportunity to optimize hfs_bmap_alloc() from a first-fit
to a next-fit allocator by caching an in-core allocation hint (roving
pointer) and bounding the scan with tree->node_count. To keep the
scope of this series strictly aligned with the Syzkaller corruption
fix, that architectural optimization is deferred to a separate,
follow-up patchset/thread.

Link: https://lore.kernel.org/all/20260226091235.927749-1-shardul.b@xxxxxxxxxxxxxxxxxx/

v5:
 - API Encapsulation: Introduced struct hfs_bmap_ctx to cleanly bundle
   offset, length, and page index state instead of passing multiple
   pointers, addressing reviewer feedback.

 - Bit-Level Helpers: Added hfs_bmap_test_bit() and hfs_bmap_clear_bit()
   to safely encapsulate mapping/unmapping for single-bit accesses
   (like the mount-time check and node freeing).

 - Performance Retention: Retained the page-level mapping approach for
   the linear scan inside hfs_bmap_alloc() to prevent the severe
   performance regression of mapping/unmapping on a per-byte basis,
   while refactoring it to use the new ctx struct.

 - Hexagon Overflow Fix: Fixed a 0-day Kernel Test Robot warning on
   architectures with 256KB page sizes by upgrading the offset variables
   in the new struct hfs_bmap_ctx to unsigned int, preventing 16-bit shift
   overflows.
   Link: https://lore.kernel.org/all/202602270310.eBmeD8VX-lkp@xxxxxxxxx/

 - Map Record Spanning: Added a byte_offset parameter to the page mapper
   to correctly handle large map records that span across multiple 4KB
   pages.

 - Loop Mask Revert: Reverted the 0x80 bitmask in the alloc() inner loop
   back to its original state (and dropped the HFSPLUS_BTREE_NODE0_BIT
   macro), as it represents a generic sliding mask, not specifically
   Node 0.

 - String Array Cleanup: Replaced the verbose switch(id) block in the
   mount validation with a clean static array of constant strings for
   the CNID names, per reviewer feedback.

v4:
 - Split the changes into a 2-patch series (Refactoring + Bug Fix).
 - Extracted map node traversal into a generic helper (hfs_bmap_get_map_page)
   as per Slava's feedback, replacing manual offset/page management.
 - Added node-type validation (HFS_NODE_HEADER vs HFS_NODE_MAP) inside the
   helper to defend against structurally corrupted linkages.
 - Replaced hardcoded values with named macros (HFSPLUS_BTREE_NODE0_BIT, etc).
 - Handled invalid map offsets/lengths as corruption, continuing the mount
   as SB_RDONLY instead of failing it completely to preserve data recovery.

v3:
  - Moved validation logic inline into hfs_btree_open() to allow
    reporting the specific corrupted tree ID.
  - Replaced custom offset calculations with existing hfs_bnode_find()
    and hfs_brec_lenoff() infrastructure to handle node sizes and
    page boundaries correctly.
  - Removed temporary 'btree_bitmap_corrupted' superblock flag; setup
    SB_RDONLY directly upon detection.
  - Moved logging to hfs_btree_open() to include the specific tree ID in
    the warning message
  - Used explicit bitwise check (&) instead of test_bit() to ensure
    portability. test_bit() bit-numbering is architecture-dependent
    (e.g., bit 0 vs bit 7 can swap meanings on BE vs LE), whereas
    masking 0x80 consistently targets the MSB required by the HFS+
    on-disk format.

v2:
  - Fix compiler warning about comparing u16 bitmap_off with PAGE_SIZE which
can exceed u16 maximum on some architectures
  - Cast bitmap_off to unsigned int for the PAGE_SIZE comparison to avoid
tautological constant-out-of-range comparison warning.
  - Link: https://lore.kernel.org/oe-kbuild-all/202601251011.kJUhBF3P-lkp@xxxxxxxxx/

Shardul Bankar (2):
  hfsplus: refactor b-tree map page access and add node-type validation
  hfsplus: validate b-tree node 0 bitmap at mount time

 fs/hfsplus/btree.c         | 222 +++++++++++++++++++++++++++++--------
 include/linux/hfs_common.h |   2 +
 2 files changed, 177 insertions(+), 47 deletions(-)

-- 
2.34.1