kernel/linux.git/fs/ntfs, branch master

ntfs: use page allocation for resident attribute inline data

2026-04-22T10:05:07+00:00

The current kmemdup() based allocation for IOMAP_INLINE can result in inline_data pointer having a non-zero page offset. This causes iomap_inline_data_valid() to fail the check: iomap->length <= PAGE_SIZE - offset_in_page(iomap->inline_data) and triggers the kernel BUG at fs/iomap/buffered-io.c:1061. This particularly affects workloads with frequent small file access (e.g. Firefox Nightly profile on NTFS with bind mount) when using the new ntfs. This fix this by allocating a full page with alloc_page() so that page_address() always returns a page-aligned address. Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: fix mmap_prepare writable check for shared mappings

2026-04-22T09:33:15+00:00

Linus pointed out that checking only VMA_WRITE_BIT is incorrect. Private writable mappings (MAP_PRIVATE) set VM_WRITE but do not write back to the filesystem. Also, mappings that can become writable via mprotect() (VM_MAYWRITE) must be handled. Use vma_desc_test_all(VMA_SHARED_BIT, VMA_MAYWRITE_BIT) instead, which matches what other filesystems do. Suggested-by: Linus Torvalds Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: fix potential 32-bit truncation in ntfs_write_cb()

2026-04-18T02:33:21+00:00

Smatch warned that the bitwise negation in ntfs_write_cb() might lead to unintended truncation. Casting the block size to loff_t before bitwise negation prevents the upper 32 bits of pos from being incorrectly zeroed out during the calculation of new_vcn. Signed-off-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: fix uninitialized variable in ntfs_map_runlist_nolock

2026-04-18T02:33:19+00:00

Smatch reported that ctx_needs_reset could be used uninitialized if ntfs_map_runlist_nolock() fails early when a search context is provided. Specifically, if the function returns -EIO because the attribute is resident, the code jumps to err_out. This initializes ctx_needs_reset to false to satisfy the static checker. Reported-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: delete dead code

2026-04-18T02:33:17+00:00

We know "ret2" is zero so there is no need to check. Delete the if statement. Signed-off-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: add missing error code in ntfs_mft_record_alloc()

2026-04-18T02:33:13+00:00

Return -ENOMEM if the kmalloc() fails. Don't return success. Signed-off-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: fix uninitialized variables in ntfs_ea_set_wsl_inode()

2026-04-18T02:33:09+00:00

Smatch reported uninitialized symbol warnings in ntfs_ea_set_wsl_inode() and __ntfs_create(). In ntfs_ea_set_wsl_inode(), the err variable could be returned without initialization if no flags are set and rdev is zero. Additionally, ea_size might remain uninitialized from the caller's perspective if no EA operations are performed. While these cases might not be triggered under current logic, we initialize them to zero to satisfy the static checker. Reported-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: fix uninitialized pointer in ntfs_write_mft_block

2026-04-18T02:33:07+00:00

Smatch reported that the variable rl could be used uninitialized in ntfs_write_mft_block(). After analyzing the code, when vol->cluster_size == NTFS_BLOCK_SIZE (512), it is smaller than folio_size, so rl is guaranteed to be initialized. If vol->cluster_size is larger, the condition to access rl becomes false, so a runtime error is not expected to occur. However, to make the static checker happy, this patch initializes rl to NULL and adds an explicit check before its usage. Reported-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: fix uninitialized variable in ntfs_write_simple_iomap_begin_non_resident

2026-04-18T02:33:05+00:00

Smatch reported that err could be used uninitialized if the code path does not enter the first ntfs_zero_range() block. Reported-by: Dan Carpenter Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon

ntfs: remove noop_direct_IO from address_space_operations

2026-04-18T02:33:03+00:00

Since commit a2ad63daa88b ("VFS: add FMODE_CAN_ODIRECT file flag"), noop_direct_io is not required. Signed-off-by: Hyunchul Lee Signed-off-by: Namjae Jeon