kernel/linux.git/fs/ntfs3/inode.c, branch linux-7.1.y

Merge tag 'ntfs3_for_7.1' of https://github.com/Paragon-Software-Group/linux-ntfs3

2026-04-20T17:59:47+00:00

Pull ntfs3 updates from Konstantin Komarov: "New: - reject inodes with zero non-DOS link count - return folios from ntfs_lock_new_page() - subset of W=1 warnings for stricter checks - work around -Wmaybe-uninitialized warnings - buffer boundary checks to run_unpack() - terminate the cached volume label after UTF-8 conversion Fixes: - check return value of indx_find to avoid infinite loop - prevent uninitialized lcn caused by zero len - increase CLIENT_REC name field size to prevent buffer overflow - missing run load for vcn0 in attr_data_get_block_locked() - memory leak in indx_create_allocate() - OOB write in attr_wof_frame_info() - mount failure on volumes with fragmented MFT bitmap - integer overflow in run_unpack() volume boundary check - validate rec->used in journal-replay file record check Updates: - resolve compare function in public index APIs - $LXDEV xattr lookup - potential double iput on d_make_root() failure - initialize err in ni_allocate_da_blocks_locked() - correct the pre_alloc condition in attr_allocate_clusters()" * tag 'ntfs3_for_7.1' of https://github.com/Paragon-Software-Group/linux-ntfs3: fs/ntfs3: fix Smatch warnings fs/ntfs3: validate rec->used in journal-replay file record check fs/ntfs3: terminate the cached volume label after UTF-8 conversion fs/ntfs3: fix potential double iput on d_make_root() failure ntfs3: fix integer overflow in run_unpack() volume boundary check ntfs3: add buffer boundary checks to run_unpack() ntfs3: fix mount failure on volumes with fragmented MFT bitmap fs/ntfs3: fix $LXDEV xattr lookup ntfs3: fix OOB write in attr_wof_frame_info() ntfs3: fix memory leak in indx_create_allocate() ntfs3: work around false-postive -Wmaybe-uninitialized warnings fs/ntfs3: fix missing run load for vcn0 in attr_data_get_block_locked() fs/ntfs3: increase CLIENT_REC name field size fs/ntfs3: prevent uninitialized lcn caused by zero len fs/ntfs3: add a subset of W=1 warnings for stricter checks fs/ntfs3: return folios from ntfs_lock_new_page() fs/ntfs3: resolve compare function in public index APIs ntfs3: reject inodes with zero non-DOS link count

Merge tag 'ntfs-for-7.1-rc1-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/ntfs

2026-04-17T23:35:49+00:00

Pull ntfs resurrection from Namjae Jeon: "Ever since Kari Argillander’s 2022 report [1] regarding the state of the ntfs3 driver, I have spent the last 4 years working to provide full write support and current trends (iomap, no buffer head, folio), enhanced performance, stable maintenance, utility support including fsck for NTFS in Linux. This new implementation is built upon the clean foundation of the original read-only NTFS driver, adding: - Write support: Implemented full write support based on the classic read-only NTFS driver. Added delayed allocation to improve write performance through multi-cluster allocation and reduced fragmentation of the cluster bitmap. - iomap conversion: Switched buffered IO (reads/writes), direct IO, file extent mapping, readpages, and writepages to use iomap. - Remove buffer_head: Completely removed buffer_head usage by converting to folios. As a result, the dependency on CONFIG_BUFFER_HEAD has been removed from Kconfig. - Stability improvements: The new ntfs driver passes 326 xfstests, compared to 273 for ntfs3. All tests passed by ntfs3 are a complete subset of the tests passed by this implementation. Added support for fallocate, idmapped mounts, permissions, and more. xfstests Results report: Total tests run: 787 Passed : 326 Failed : 38 Skipped : 423 Failed tests breakdown: - 34 tests require metadata journaling - 4 other tests: 094: No unwritten extent concept in NTFS on-disk format 563: cgroup v2 aware writeback accounting not supported 631: RENAME_WHITEOUT support required 787: NFS delegation test" Link: https://lore.kernel.org/all/da20d32b-5185-f40b-48b8-2986922d8b25@stargateuniverse.net/ [1] [ Let's see if this undead filesystem ends up being of the "Easter miracle" kind, or the "Nosferatu of filesystems" kind... ] * tag 'ntfs-for-7.1-rc1-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/ntfs: (46 commits) ntfs: remove redundant out-of-bound checks ntfs: add bound checking to ntfs_external_attr_find ntfs: add bound checking to ntfs_attr_find ntfs: fix ignoring unreachable code warnings ntfs: fix inconsistent indenting warnings ntfs: fix variable dereferenced before check warnings ntfs: prefer IS_ERR_OR_NULL() over manual NULL check ntfs: harden ntfs_listxattr against EA entries ntfs: harden ntfs_ea_lookup against malformed EA entries ntfs: check $EA query-length in ntfs_ea_get ntfs: validate WSL EA payload sizes ntfs: fix WSL ea restore condition ntfs: add missing newlines to pr_err() messages ntfs: fix pointer/integer casting warnings ntfs: use ->mft_no instead of ->i_ino in prints ntfs: change mft_no type to u64 ntfs: select FS_IOMAP in Kconfig ntfs: add MODULE_ALIAS_FS ntfs: reduce stack usage in ntfs_write_mft_block() ntfs: fix sysctl table registration and path ...

Merge tag 'vfs-7.1-rc1.bh.metadata' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

2026-04-13T19:46:42+00:00

Pull vfs buffer_head updates from Christian Brauner: "This cleans up the mess that has accumulated over the years in metadata buffer_head tracking for inodes. It moves the tracking into dedicated structure in filesystem-private part of the inode (so that we don't use private_list, private_data, and private_lock in struct address_space), and also moves couple other users of private_data and private_list so these are removed from struct address_space saving 3 longs in struct inode for 99% of inodes" * tag 'vfs-7.1-rc1.bh.metadata' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (42 commits) fs: Drop i_private_list from address_space fs: Drop mapping_metadata_bhs from address space ext4: Track metadata bhs in fs-private inode part minix: Track metadata bhs in fs-private inode part udf: Track metadata bhs in fs-private inode part fat: Track metadata bhs in fs-private inode part bfs: Track metadata bhs in fs-private inode part affs: Track metadata bhs in fs-private inode part ext2: Track metadata bhs in fs-private inode part fs: Provide functions for handling mapping_metadata_bhs directly fs: Switch inode_has_buffers() to take mapping_metadata_bhs fs: Make bhs point to mapping_metadata_bhs fs: Move metadata bhs tracking to a separate struct fs: Fold fsync_buffers_list() into sync_mapping_buffers() fs: Drop osync_buffers_list() kvm: Use private inode list instead of i_private_list fs: Remove i_private_data aio: Stop using i_private_data and i_private_lock hugetlbfs: Stop using i_private_data fs: Stop using i_private_data for metadata bh tracking ...

ntfs3: Drop pointless sync_mapping_buffers() and invalidate_inode_buffers() calls

2026-03-26T14:03:27+00:00

ntfs3 never calls mark_buffer_dirty_inode() and thus its metadata buffers list is always empty. Drop the pointless sync_mapping_buffers() and invalidate_inode_buffers() calls. CC: Konstantin Komarov CC: ntfs3@lists.linux.dev Signed-off-by: Jan Kara Link: https://patch.msgid.link/20260326095354.16340-45-jack@suse.cz Signed-off-by: Christian Brauner

ntfs3: remove copy and pasted iomap code

2026-03-10T09:29:03+00:00

ntfs3 copied the iomap code without attribution or talking to the maintainers, to hook into the bio completion for (unexplained) zeroing. Fix this by just overriding the bio completion handler in the submit handler. Signed-off-by: Christoph Hellwig Link: https://patch.msgid.link/20260223132021.292832-13-hch@lst.de Reviewed-by: "Darrick J. Wong" Signed-off-by: Christian Brauner

iomap: only call into ->submit_read when there is a read_ctx

2026-03-10T09:29:03+00:00

Move the NULL check into the callers to simplify the callees. Fuse was missing this before, but has a constant read_ctx that is never NULL or changed, so no change here either. Signed-off-by: Christoph Hellwig Link: https://patch.msgid.link/20260223132021.292832-11-hch@lst.de Tested-by: Anuj Gupta Reviewed-by: "Darrick J. Wong" Signed-off-by: Christian Brauner

iomap: pass the iomap_iter to ->submit_read

2026-03-10T09:29:03+00:00

This provides additional context for file systems. Rename the fuse instance to match the method name while we're at it. Signed-off-by: Christoph Hellwig Link: https://patch.msgid.link/20260223132021.292832-10-hch@lst.de Tested-by: Anuj Gupta Reviewed-by: "Darrick J. Wong" Signed-off-by: Christian Brauner

fs/ntfs3: prevent uninitialized lcn caused by zero len

2026-03-04T09:28:42+00:00

syzbot reported a uninit-value in ntfs_iomap_begin [1]. Since runs was not touched yet, run_lookup_entry() immediately fails and returns false, which makes the value of "*len" 0. Simultaneously, the new value and err value are also 0, causing the logic in attr_data_get_block_locked() to jump directly to ok, ultimately resulting in *lcn being triggered before it is set [1]. In ntfs_iomap_begin(), the check for a 0 value in clen is moved forward to before updating lcn to avoid this [1]. [1] BUG: KMSAN: uninit-value in ntfs_iomap_begin+0x8c0/0x1460 fs/ntfs3/inode.c:825 ntfs_iomap_begin+0x8c0/0x1460 fs/ntfs3/inode.c:825 iomap_iter+0x9b7/0x1540 fs/iomap/iter.c:110 Local variable lcn created at: ntfs_iomap_begin+0x15d/0x1460 fs/ntfs3/inode.c:786 Fixes: 10d7c95af043 ("fs/ntfs3: add delayed-allocation (delalloc) support") Reported-by: syzbot+7be88937363ac7ab7bb0@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=7be88937363ac7ab7bb0 Tested-by: syzbot+7be88937363ac7ab7bb0@syzkaller.appspotmail.com Signed-off-by: Edward Adam Davis Signed-off-by: Konstantin Komarov

ntfs3: reject inodes with zero non-DOS link count

2026-02-24T19:03:14+00:00

ntfs_read_mft() counts file name attributes into two variables: names (all names including DOS 8.3) and links (non-DOS names only). The validation at line 424 checks names but set_nlink() at line 436 uses links. A corrupted NTFS image where all file name attributes have type FILE_NAME_DOS passes the names check but results in set_nlink(inode, 0). When such an inode is loaded via a code path that passes name=NULL to ntfs_iget5() and the nlink=0 inode enters the VFS. The subsequent unlink, rmdir, or rename targeting this inode calls drop_nlink() which triggers WARN_ON(inode->i_nlink == 0) in fs/inode.c. An all-DOS-name MFT record cannot exist on a valid NTFS volume. Reject such records by checking for links == 0 before calling set_nlink(). Signed-off-by: Ziyi Guo Signed-off-by: Konstantin Komarov

ntfs3: remove legacy ntfs driver support

2026-02-19T12:50:58+00:00

Reverts the following commits that introduced legacy ntfs driver alias and related support code: 74871791ffa9 ntfs3: serve as alias for the legacy ntfs driver 1ff2e956608c fs/ntfs3: Redesign legacy ntfs support 9b872cc50daa ntfs3: add legacy ntfs file operations d55f90e9b243 ntfs3: enforce read-only when used as legacy ntfs driver The legacy ntfs driver has been remade as a new implementation, so the alias and related codes in ntfs3 are no longer needed. Acked-by: Christoph Hellwig Signed-off-by: Namjae Jeon