Re: [PATCH 1/2] ntfs: keep the file size on a failed initialized-size extension

From: liubaolin

Date: Wed Oct 28 2026 - 04:42:02 EST




在 2026/10/28 16:06, Hongling Zeng 写道:
ntfs_extend_initialized_size() only extends the initialized size and
never changes the data size, but on a failure it calls
truncate_setsize() with the old initialized size, which shrinks the
VFS size down to it and drops the page cache beyond that point.

None of the callers rolls the shrunk size back. A buffered write
into a hole has nothing to roll back - the data size never changed -
so the file keeps the shrunk size even though the data size on disk
is intact. The mmap and named stream callers return without any
rollback either.

The sizes did not change, so nothing needs the shrink. Remove it.

Fixes: af0db57d4293 ("ntfs: update inode operations")
Signed-off-by: Hongling Zeng <zenghongling@xxxxxxxxxx>
---
fs/ntfs/inode.c | 2 --
1 file changed, 2 deletions(-)

diff --git a/fs/ntfs/inode.c b/fs/ntfs/inode.c
index 206db4e99ef1..ac13e8776909 100644
--- a/fs/ntfs/inode.c
+++ b/fs/ntfs/inode.c
@@ -2599,8 +2599,6 @@ int ntfs_extend_initialized_size(struct inode *vi, const loff_t offset,
mutex_lock(&mrec_ni->mrec_lock);
err = ntfs_attr_set_initialized_size(ni, new_size);
mutex_unlock(&mrec_ni->mrec_lock);
- if (err)
- truncate_setsize(vi, old_init_size);
out:
filemap_invalidate_unlock(vi->i_mapping);
return err;

Looks good to me.

Reviewed-by: Baolin Liu <liubaolin@xxxxxxxxxx>