The target VCN being sought was not found within runs[0], causing
run_lookup() to return false. This causes run_lookup_entry() to return
false, which in turn results in a len value of 0, and the new parameter
passed to attr_data_get_block() is NULL. Collectively, these factors
ultimately cause attr_data_get_block_locked() to exit prematurely without
initializing lcn, thereby triggering [1].
To prevent [1], the clen check within ni_seek_data_or_hole() has been
moved to occur before the lcn check.
[1]
BUG: KMSAN: uninit-value in ni_seek_data_or_hole+0x24f/0x5f0 fs/ntfs3/frecord.c:2862
ni_seek_data_or_hole+0x24f/0x5f0 fs/ntfs3/frecord.c:2862
ntfs_llseek+0x22a/0x4a0 fs/ntfs3/file.c:1530
vfs_llseek fs/read_write.c:391 [inline]
Fixes: c613269677 ("fs/ntfs3: implement llseek SEEK_DATA/SEEK_HOLE by scanning data runs")
Reported-by: syzbot+c2cfe997245202e46f10@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=c2cfe997245202e46f10
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The ntfs3 specific -Wmaybe-uninitialized flag found one more false-postive,
this time with gcc-10 on s390:
fs/ntfs3/frecord.c: In function 'ni_expand_list':
fs/ntfs3/frecord.c:1370:16: error: 'ins_attr' may be used uninitialized in this function [-Werror=maybe-uninitialized]
Add an explicit NULL pointer check before using the pointer, and
initialize it to NULL.
Fixes: 48d9b57b16 ("fs/ntfs3: add a subset of W=1 warnings for stricter checks")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
format code according to .clang-format, add useful comments and remove
non-useful comments.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Handle non-data/hole seeks through generic_file_llseek_size() and return
-ENXIO immediately when SEEK_DATA or SEEK_HOLE is requested at or past
EOF. Handle compressed files in such cases properly as well.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
It is safe to call _ntfs_bad_inode on live inodes since:
commit 519b078998 ("fs/ntfs3: Exclude call make_bad_inode for live nodes.")
The WARN_ON was added when it wasn't safe by:
commit d99208b919 ("fs/ntfs3: cancle set bad inode after removing name fails")
Replace the WARN_ON with a call to _ntfs_bad_inode() to prevent further
operations on the inconsistent inode.
Reported-by: syzbot+4d8e30dbafb5c1260479@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=4d8e30dbafb5c1260479
Fixes: 519b078998 ("fs/ntfs3: Exclude call make_bad_inode for live nodes.")
Signed-off-by: Helen Koike <koike@igalia.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Initialize err in ni_allocate_da_blocks_locked() and correct the
pre_alloc condition in attr_allocate_clusters().
Suggested-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
ntfs_lock_new_page() currently returns a struct page * but it primarily
operates on folios via __filemap_get_folio(). Convert it to return a
struct folio * and use folio_alloc() + __folio_set_locked() for the
temporary page used to avoid data corruption during decompression.
When the cached folio is not uptodate, preserve the existing behavior by
using folio_file_page() and converting the returned page back to a
folio.
Update ni_readpage_cmpr() and ni_decompress_file() to handle the new
return type while keeping the existing struct page * array and the
unlock_page()/put_page() cleanup paths unchanged.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202602072013.jwrURE2e-lkp@intel.com/
Closes: https://lore.kernel.org/oe-kbuild-all/202602071921.nGIiI1J5-lkp@intel.com/
Signed-off-by: Sun Jian <sun.jian.kdev@gmail.com>
[almaz.alexandrovich@paragon-software.com: removed ni_fiemap function,
added reported-by and closes tags to commit]
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
This patch implements delayed allocation (delalloc) in ntfs3 driver.
It introduces an in-memory delayed-runlist (run_da) and the helpers to
track, reserve and later convert those delayed reservations into real
clusters at writeback time. The change keeps on-disk formats untouched and
focuses on pagecache integration, correctness and safe interaction with
fallocate, truncate, and dio/iomap paths.
Key points:
- add run_da (delay-allocated run tree) and bookkeeping for delayed clusters.
- mark ranges as delalloc (DELALLOC_LCN) instead of immediately allocating.
Actual allocation performed later (writeback / attr_set_size_ex / explicit
flush paths).
- direct i/o / iomap paths updated to avoid dio collisions with
delalloc: dio falls back or forces allocation of delayed blocks before
proceeding.
- punch/collapse/truncate/fallocate check and cancel delay-alloc reservations.
Sparse/compressed files handled specially.
- free-space checks updated (ntfs_check_free_space) to account for reserved
delalloc clusters and MFT record budgeting.
- delayed allocations are committed on last writer (file release) and on
explicit allocation flush paths.
Tested-by: syzbot@syzkaller.appspotmail.com
Reported-by: syzbot+2bd8e813c7f767aa9bb1@syzkaller.appspotmail.com
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
This patch modifies the ntfs3 driver by replacing the buffer_head-based
operations with the iomap ones.
Implementation details:
- Implements core iomap operations (ntfs_iomap_begin/end) for block mapping:
Proper handling of resident attributes via IOMAP_INLINE.
Support for sparse files through IOMAP_HOLE semantics.
Correct unwritten extent handling for zeroing operations.
- Replaces custom implementations with standardized iomap helpers:
Converts buffered reads to use iomap_read_folio and iomap_readahead.
Implements iomap_file_buffered_write for write operations.
Uses iomap_dio_rw for direct I/O paths.
Migrates zero range operations to iomap_zero_range.
- Preserves special handling paths for compressed files
- Implements proper EOF/valid data size management during writes
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Syzbot reported a task hung in ni_readpage_cmpr (now ni_read_folio_cmpr).
This is caused by a lock inversion deadlock involving the inode mutex
(ni_lock) and page locks.
Scenario:
1. Task A enters ntfs_read_folio() for page X. It acquires ni_lock.
2. Task A calls ni_read_folio_cmpr(), which attempts to lock all pages in
the compressed frame (including page Y).
3. Concurrently, Task B (e.g., via readahead) has locked page Y and
calls ntfs_read_folio().
4. Task B waits for ni_lock (held by A).
5. Task A waits for page Y lock (held by B).
-> DEADLOCK.
The fix is to restructure locking: do not take ni_lock in ntfs_read_folio().
Instead, acquire ni_lock inside ni_read_folio_cmpr() ONLY AFTER all required
page locks for the frame have been successfully acquired. This restores the
correct lock ordering (Page Lock -> ni_lock) consistent with VFS.
Reported-by: syzbot+5af33dd272b913b65880@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=5af33dd272b913b65880
Fixes: f35590ee26 ("fs/ntfs3: remove ntfs_bio_pages and use page cache for compressed I/O")
Signed-off-by: Szymon Wilczek <swilczek.lx@gmail.com>
[almaz.alexandrovich@paragon-software.com: ni_readpage_cmpr was renamed to ni_read_folio_cmpr]
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The generic llseek implementation does not understand ntfs data runs,
sparse regions, or compression semantics, and therefore cannot correctly
locate data or holes in files.
Add a filesystem-specific llseek handler that scans attribute data runs
to find the next data or hole starting at the given offset. Handle
resident attributes, sparse runs, compressed holes, and the implicit hole
at end-of-file.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Some xfstests expect fsync() on a file or directory to also persist
directory metadata up the parent chain. Using generic_file_fsync() is not
sufficient for ntfs, because parent directories are not explicitly
written out.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The old "readpage" naming is still used in ni_readpage_cmpr(), even though
the vfs has transitioned to the folio-based read_folio() API.
This patch performs a straightforward renaming of the helper:
ni_readpage_cmpr() -> ni_read_folio_cmpr().
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
If a rb node with the same ino already exists in the rb tree, the newly
alloced mft_inode in ni_add_subrecord() will not have its memory cleaned
up, which leads to the memory leak issue reported by syzbot.
The best option to avoid this issue is to put the newly alloced mft node
when a rb node with the same ino already exists in the rb tree and return
the rb node found in the rb tree to the parent layer.
syzbot reported:
BUG: memory leak
unreferenced object 0xffff888110bef280 (size 128):
backtrace (crc 126a088f):
ni_add_subrecord+0x31/0x180 fs/ntfs3/frecord.c:317
ntfs_look_free_mft+0xf0/0x790 fs/ntfs3/fsntfs.c:715
BUG: memory leak
unreferenced object 0xffff888109093400 (size 1024):
backtrace (crc 7197c55e):
mi_init+0x2b/0x50 fs/ntfs3/record.c:105
mi_format_new+0x40/0x220 fs/ntfs3/record.c:422
Fixes: 4342306f0f ("fs/ntfs3: Add file operations and implementation")
Reported-by: syzbot+3932ccb896e06f7414c9@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
After ntfs_look_free_mft() executes successfully, all subsequent code
that fails to execute must put mi.
Fixes: 4342306f0f ("fs/ntfs3: Add file operations and implementation")
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Replace the use of ntfs_bio_pages with the disk page cache for reading and
writing compressed files. This slightly improves performance when reading
compressed data and simplifies the I/O logic.
When an XPRESS or LZX compressed file is opened for writing, it is now
decompressed into a normal file before modification. A new argument (`int copy`)
is added to ni_read_frame() to handle writing of decompressed and mapped data.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The call to kmalloc() to allocate the attribute list buffer is given a
size of al_aligned(rs). This size can be larger than the data
subsequently copied into the buffer, leaving trailing bytes uninitialized.
This can trigger a KMSAN "uninit-value" warning if that memory is
later accessed.
Fix this by using kzalloc() instead, which ensures the entire
allocated buffer is zero-initialized, preventing the warning.
Reported-by: syzbot+83c9dd5c0dcf6184fdbf@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=83c9dd5c0dcf6184fdbf
Signed-off-by: Nirbhay Sharma <nirbhay.lkd@gmail.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
When reading a compressed file, we may read several pages in addition to
the one requested. The current code will overwrite pages in the page
cache with the data from disc which can definitely result in changes
that have been made being lost.
For example if we have four consecutie pages ABCD in the file compressed
into a single extent, on first access, we'll bring in ABCD. Then we
write to page B. Memory pressure results in the eviction of ACD.
When we attempt to write to page C, we will overwrite the data in page
B with the data currently on disk.
I haven't investigated the decompression code to check whether it's
OK to overwrite a clean page or whether it might be possible to see
corrupt data. Out of an abundance of caution, decline to overwrite
uptodate pages, not just dirty pages.
Fixes: 4342306f0f (fs/ntfs3: Add file operations and implementation)
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: stable@vger.kernel.org
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
These pages are accessed through vmap; they are not accessed
by calling page_address(), so they do not need to be kmapped.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
These pages are accessed through DMA and vmap; they are not accessed
by calling page_address(), so they do not need to be kmapped.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Reformatted the driver code according to the current .clang-format rules
and updated description of used terminology. No functional changes
intended.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Use ntfs_inode field 'ni_bad' to mark inode as bad (if something went wrong)
and to avoid any operations
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The reproducer uses a file0 on a ntfs3 file system with a corrupted i_link.
When renaming, the file0's inode is marked as a bad inode because the file
name cannot be deleted.
The underlying bug is that make_bad_inode() is called on a live inode.
In some cases it's "icache lookup finds a normal inode, d_splice_alias()
is called to attach it to dentry, while another thread decides to call
make_bad_inode() on it - that would evict it from icache, but we'd already
found it there earlier".
In some it's outright "we have an inode attached to dentry - that's how we
got it in the first place; let's call make_bad_inode() on it just for shits
and giggles".
Fixes: 78ab59fee0 ("fs/ntfs3: Rework file operations")
Reported-by: syzbot+1aa90f0eb1fc3e77d969@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=1aa90f0eb1fc3e77d969
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The symlinks created in windows will be broken in linux by ntfs3,
the patch fixes it.
Signed-off-by: Rong Zhang <ulin0208@gmail.com>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Remove all the code related to changing compression on the fly because
it's not safe and not maintainable.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
ni_load_attr() was added in 2021 by
commit 4342306f0f ("fs/ntfs3: Add file operations and implementation")
but hasn't been used.
Remove it.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Update inode->i_mapping->a_ops when the compression state changes to
ensure correct address space operations.
Clear ATTR_FLAG_SPARSED/FILE_ATTRIBUTE_SPARSE_FILE when enabling
compression to prevent flag conflicts.
v2:
Additionally, ensure that all dirty pages are flushed and concurrent access
to the page cache is blocked.
Fixes: 6b39bfaeec ("fs/ntfs3: Add support for the compression attribute")
Reported-by: Kun Hu <huk23@m.fudan.edu.cn>, Jiaji Qin <jjtan24@m.fudan.edu.cn>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Extended the `mi_enum_attr()` function interface with an additional
parameter, `struct ntfs_inode *ni`, to allow marking the inode
as bad as soon as an error is detected.
Reported-by: syzbot+73d8fc29ec7cba8286fa@syzkaller.appspotmail.com
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Convert the first page passed to ni_write_frame() to a folio and use
folio_pos() on that instead of open-coding the access to folio->index,
cast & shift.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Use local runs_tree instead of cached. This way excludes rw_semaphore lock.
Reported-by: syzbot+1c25748a40fe79b8a119@syzkaller.appspotmail.com
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Checking of NTFS_FLAGS_LOG_REPLAYING added to prevent access to
uninitialized bitmap during replay process.
Reported-by: syzbot+3bfd2cc059ab93efcdb4@syzkaller.appspotmail.com
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
The interface of fiemap_fill_next_extent_k() was modified
to eliminate the sparse warning.
Fixes: d57431c6f5 ("fs/ntfs3: Do copy_to_user out of run_lock")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202406271920.hndE8N6D-lkp@intel.com/
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
In order not to call copy_to_user (from fiemap_fill_next_extent)
we allocate memory in the kernel, fill it and copy it to user memory
after up_read(run_lock).
Reported-by: syzbot+36bb70085ef6edc2ebb9@syzkaller.appspotmail.com
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
This involves converting all users of offs_page to offs_folio, but
it's worth it because we get rid of a lot of hidden calls to
compound_head(). We continue to use order-0 folios here, and convert
back to a struct page to call ntfs_bio_pages().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
We still use an array of pages for the decompression, but this removes
a few calls to compound_head().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Nobody checks the error flag on ntfs3 folios, so stop setting and
clearing it.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
COMPRESSION_UNIT and NTFS_LZNT_CUNIT mean the same thing
(1u<<NTFS_LZNT_CUNIT) determines the size for compression (in clusters).
COMPRESS_MAX_CLUSTER is not used in the code.
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
When CONFIG_NTFS3_LZX_XPRESS is not set then we get the following build
error:
fs/ntfs3/frecord.c:2460:16: error: unused variable ‘i_size’
Signed-off-by: Mark O'Donovan <shiftee@posteo.net>
Fixes: 4fd6c08a16 ("fs/ntfs3: Use i_size_read and i_size_write")
Tested-by: Chris Clayton <chris2553@googlemail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRAhzRXHqcMeLMyaSiRxhvAZXjcogUCZTppYgAKCRCRxhvAZXjc
okIHAP9anLz1QDyMLH12ASuHjgBc0Of3jcB6NB97IWGpL4O21gEA46ohaD+vcJuC
YkBLU3lXqQ87nfu28ExFAzh10hG2jwM=
=m4pB
-----END PGP SIGNATURE-----
Merge tag 'vfs-6.7.ctime' of gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs
Pull vfs inode time accessor updates from Christian Brauner:
"This finishes the conversion of all inode time fields to accessor
functions as discussed on list. Changing timestamps manually as we
used to do before is error prone. Using accessors function makes this
robust.
It does not contain the switch of the time fields to discrete 64 bit
integers to replace struct timespec and free up space in struct inode.
But after this, the switch can be trivially made and the patch should
only affect the vfs if we decide to do it"
* tag 'vfs-6.7.ctime' of gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs: (86 commits)
fs: rename inode i_atime and i_mtime fields
security: convert to new timestamp accessors
selinux: convert to new timestamp accessors
apparmor: convert to new timestamp accessors
sunrpc: convert to new timestamp accessors
mm: convert to new timestamp accessors
bpf: convert to new timestamp accessors
ipc: convert to new timestamp accessors
linux: convert to new timestamp accessors
zonefs: convert to new timestamp accessors
xfs: convert to new timestamp accessors
vboxsf: convert to new timestamp accessors
ufs: convert to new timestamp accessors
udf: convert to new timestamp accessors
ubifs: convert to new timestamp accessors
tracefs: convert to new timestamp accessors
sysv: convert to new timestamp accessors
squashfs: convert to new timestamp accessors
server: convert to new timestamp accessors
client: convert to new timestamp accessors
...
