Pull exfat updates from Namjae Jeon:
- Implement FALLOC_FL_ALLOCATE_RANGE to add support for preallocating
clusters without zeroing, helping to reduce file fragmentation
- Add a unified block readahead helper for FAT chain conversion, bitmap
allocation, and directory entry lookups
- Optimize exfat_chain_cont_cluster() by caching buffer heads to
minimize mark_buffer_dirty() and mirroring overhead during
NO_FAT_CHAIN to FAT_CHAIN conversion
- Switch to truncate_inode_pages_final() in evict_inode() to prevent
BUG_ON caused by shadow entries during reclaim
- Fix a 32-bit truncation bug in directory entry calculations by
ensuring proper bitwise coercion
- Fix sb->s_maxbytes calculation to correctly reflect the maximum
possible volume size for a given cluster size, resolving xfstests
generic/213
- Introduced exfat_cluster_walk() helper to traverse FAT chains by a
specified step, handling both ALLOC_NO_FAT_CHAIN and ALLOC_FAT_CHAIN
modes
- Introduced exfat_chain_advance() helper to advance an exfat_chain
structure, updating both the current cluster and remaining size
- Remove dead assignments and fix Smatch warnings
* tag 'exfat-for-7.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/exfat:
exfat: use exfat_chain_advance helper
exfat: introduce exfat_chain_advance helper
exfat: remove NULL cache pointer case in exfat_ent_get
exfat: use exfat_cluster_walk helper
exfat: introduce exfat_cluster_walk helper
exfat: fix incorrect directory checksum after rename to shorter name
exfat: fix s_maxbytes
exfat: fix passing zero to ERR_PTR() in exfat_mkdir()
exfat: fix error handling for FAT table operations
exfat: optimize exfat_chain_cont_cluster with cached buffer heads
exfat: drop redundant sec parameter from exfat_mirror_bh
exfat: use readahead helper in exfat_get_dentry
exfat: use readahead helper in exfat_allocate_bitmap
exfat: add block readahead in exfat_chain_cont_cluster
exfat: add fallocate FALLOC_FL_ALLOCATE_RANGE support
exfat: Fix bitwise operation having different size
exfat: Drop dead assignment of num_clusters
exfat: use truncate_inode_pages_final() at evict_inode()
Introduce exfat_chain_advance() to walk a exfat_chain structure by a
given step, updating both ->dir and ->size fields atomically. This
helper handles both ALLOC_NO_FAT_CHAIN and ALLOC_FAT_CHAIN modes with
proper boundary checking.
Suggested-by: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Since exfat_get_next_cluster has been updated, no callers pass a NULL
pointer to exfat_ent_get, so remove the handling logic for this case.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Replace the custom exfat_walk_fat_chain() function and open-coded
FAT chain walking logic with the exfat_cluster_walk() helper across
exfat_find_location, __exfat_get_dentry_set, and exfat_map_cluster.
Suggested-by: Sungjong Seo <sj1557.seo@samsung.com>
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Introduce exfat_cluster_walk() to walk the FAT chain by a given step,
handling both ALLOC_NO_FAT_CHAIN and ALLOC_FAT_CHAIN modes. Also
redefine exfat_get_next_cluster as a thin wrapper around it for
backward compatibility.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
When renaming a file in-place to a shorter name, exfat_remove_entries
marks excess entries as DELETED, but es->num_entries is not updated
accordingly. As a result, exfat_update_dir_chksum iterates over the
deleted entries and computes an incorrect checksum.
This does not lead to persistent corruption because mark_inode_dirty()
is called afterward, and __exfat_write_inode later recomputes the
checksum using the correct num_entries value.
Fix by setting es->num_entries = num_entries in exfat_init_ext_entry.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
With fallocate support, xfstest unit generic/213 fails with
QA output created by 213
We should get: fallocate: No space left on device
Strangely, xfs_io sometimes says "Success" when something went wrong
-fallocate: No space left on device
+fallocate: File too large
because sb->s_maxbytes is set to the volume size.
To be in line with other non-extent-based filesystems, set to max volume
size possible with the cluster size of the volume.
Signed-off-by: David Timber <dxdt@dev.snart.me>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Detected by Smatch.
namei.c:890 exfat_mkdir() warn:
passing zero to 'ERR_PTR'
Signed-off-by: Yang Wen <anmuxixixi@gmail.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Fix three error handling issues in FAT table operations:
1. Fix exfat_update_bh() to properly return errors from sync_dirty_buffer
2. Fix exfat_end_bh() to properly return errors from exfat_update_bh()
and exfat_mirror_bh()
3. Fix ignored return values from exfat_chain_cont_cluster() in inode.c
and namei.c
These fixes ensure that FAT table write errors are properly propagated
to the caller instead of being silently ignored.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
When converting files from NO_FAT_CHAIN to FAT_CHAIN format, profiling
reveals significant time spent in mark_buffer_dirty() and exfat_mirror_bh()
operations. This overhead occurs because each FAT entry modification
triggers a full block dirty marking and mirroring operation.
For consecutive clusters that reside in the same block, optimize by caching
the buffer head and performing dirty marking only once at the end of the
block's modifications.
Performance improvements for converting a 30GB file:
| Cluster Size | Before Patch | After Patch | Speedup |
|--------------|--------------|-------------|---------|
| 512 bytes | 4.243s | 1.866s | 2.27x |
| 4KB | 0.863s | 0.236s | 3.66x |
| 32KB | 0.069s | 0.034s | 2.03x |
| 256KB | 0.012s | 0.006s | 2.00x |
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
The sector offset can be obtained from bh->b_blocknr, so drop the
redundant sec parameter from exfat_mirror_bh(). Also clean up the
function to use exfat_update_bh() helper.
No functional changes.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Replace the custom exfat_dir_readahead() function with the unified
exfat_blk_readahead() helper in exfat_get_dentry(). This removes
the duplicate readahead implementation and uses the common interface,
also reducing code complexity.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Use the newly added exfat_blk_readahead() helper in exfat_allocate_bitmap()
to simplify the code. This eliminates the duplicate inline readahead logic
and uses the unified readahead interface.
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
When a file cannot allocate contiguous clusters, exfat converts the file
from NO_FAT_CHAIN to FAT_CHAIN format. For large files, this conversion
process can take a significant amount of time.
Add simple readahead to read all the FAT blocks in advance, as these
blocks are consecutive, significantly improving the conversion performance.
Test in an empty exfat filesystem:
dd if=/dev/zero of=/mnt/file bs=1M count=30k
dd if=/dev/zero of=/mnt/file2 bs=1M count=1
time cat /mnt/file2 >> /mnt/file
| cluster size | before patch | after patch |
| ------------ | ------------ | ----------- |
| 512 | 47.667s | 4.316s |
| 4k | 6.436s | 0.541s |
| 32k | 0.758s | 0.071s |
| 256k | 0.117s | 0.011s |
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Currently, the Linux (ex)FAT drivers do not employ any cluster
allocation strategy to keep fragmentation at bay. As a result, when
multiple processes are competing for new clusters to expand files in
exfat filesystem on Linux simultaneously, the files end up heavily
fragmented. HDDs are most impacted, but this could also have some
negative impact on various forms of flash memory depending on the
type of underlying technology.
For instance, modern digital cameras produce multiple media files for a
single video stream. If the application does not take the fragmentation
issue into account or the system is under memory pressure, the kernel
end up allocating clusters in said files in a interleaved manner.
Demo script:
for (( i = 0; i < 4; i += 1 ));
do
dd if=/dev/urandom iflag=fullblock bs=1M count=64 of=frag-$i &
done
for (( i = 0; i < 4; i += 1 ));
do
wait
done
filefrag frag-*
Result - Linux kernel native exfat, async mount:
780 extents found
740 extents found
809 extents found
712 extents found
Result - Linux kernel native exfat, sync mount:
1852 extents found
1836 extents found
1846 extents found
1881 extents found
Result - Windows XP:
3 extents found
3 extents found
3 extents found
2 extents found
Windows kernel, on the other hand, regardless of the underlying storage
interface or the medium, seems to space out clusters for each file.
Similar strategy has to be employed by Linux fat filesystems for
efficient utilisation of storage backend.
In the meantime, userspace applications like rsync may
use fallocate to combat this issue.
This patch may introduce a regression-like behaviour to some niche
filesystem-agnostic applications that use fallocate and proceed to
non-sequentially write to the file. Examples:
- libtorrent's use of posix_fallocate() and the first fragment from a
peer is near the end of the file
- "Download accelerators" that do partial content requests(HTTP 206)
in multiple threads writing to the same file
The delay incurred in such use cases is documented in WinAPI. Patches
that add the ioctl equivalents to the WinAPI function
SetFileValidData() and `fsutil file queryvaliddata ...` will follow.
Signed-off-by: David Timber <dxdt@dev.snart.me>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
cpos has type loff_t (long long), while s_blocksize has type u32. The
inversion wil happen on u32, the coercion to s64 happens afterwards and
will do 0-left-paddding, resulting in the upper bits getting masked out.
Cast s_blocksize to loff_t before negating it.
Found by static code analysis using Klocwork.
Signed-off-by: Philipp Hahn <phahn-oss@avm.de>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
num_clusters is not used anywhere afterwards. Remove assignment.
Found by static code analysis using Klocwork.
Signed-off-by: Philipp Hahn <phahn-oss@avm.de>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Currently, exfat uses truncate_inode_pages() in exfat_evict_inode().
However, truncate_inode_pages() does not mark the mapping as exiting,
so reclaim may still install shadow entries for the mapping until
the inode teardown completes.
In older kernels like Linux 5.10, if shadow entries are present
at that point,clear_inode() can hit
BUG_ON(inode->i_data.nrexceptional);
To align with VFS eviction semantics and prevent this situation,
switch to truncate_inode_pages_final() in ->evict_inode().
Other filesystems were updated to use truncate_inode_pages_final()
in ->evict_inode() by commit 91b0abe36a ("mm + fs: store shadow
entries in page cache")'.
Signed-off-by: Yang Wen <anmuxixixi@gmail.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
This converts some of the visually simpler cases that have been split
over multiple lines. I only did the ones that are easy to verify the
resulting diff by having just that final GFP_KERNEL argument on the next
line.
Somebody should probably do a proper coccinelle script for this, but for
me the trivial script actually resulted in an assertion failure in the
middle of the script. I probably had made it a bit _too_ trivial.
So after fighting that far a while I decided to just do some of the
syntactically simpler cases with variations of the previous 'sed'
scripts.
The more syntactically complex multi-line cases would mostly really want
whitespace cleanup anyway.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This was done entirely with mindless brute force, using
git grep -l '\<k[vmz]*alloc_objs*(.*, GFP_KERNEL)' |
xargs sed -i 's/\(alloc_objs*(.*\), GFP_KERNEL)/\1)/'
to convert the new alloc_obj() users that had a simple GFP_KERNEL
argument to just drop that argument.
Note that due to the extreme simplicity of the scripting, any slightly
more complex cases spread over multiple lines would not be triggered:
they definitely exist, but this covers the vast bulk of the cases, and
the resulting diff is also then easier to check automatically.
For the same reason the 'flex' versions will be done as a separate
conversion.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
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>
Add a blank line after variable declarations in fatent.c and file.c.
This improves readability and makes code style more consistent
across the exfat subsystem.
Signed-off-by: William Hansen-Baird <william.hansen.baird@gmail.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Linus Torvalds
Chi ZhilingandNamjae Jeon
David TimberandNamjae Jeon
Jan KaraandChristian Brauner
Yang WenandNamjae Jeon
Philipp HahnandNamjae Jeon
Kees Cook
William Hansen-BairdandNamjae Jeon