From: Christoph Hellwig <hch@xxxxxx> [ Upstream commit 7a9d43eace888a0ee6095035997bb138425844d3 ] When direct I/O completions invalidates the page cache it holds neither the i_rwsem nor the invalidate_lock so it can be racing with iomap_write_delalloc_release. If the search for the end of the region that contains data returns the start offset we hit such a race and just need to look for the end of the newly created hole instead. Signed-off-by: Christoph Hellwig <hch@xxxxxx> Link: https://lore.kernel.org/r/20240910043949.3481298-2-hch@xxxxxx Reviewed-by: Darrick J. Wong <djwong@xxxxxxxxxx> Signed-off-by: Christian Brauner <brauner@xxxxxxxxxx> Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx> --- fs/iomap/buffered-io.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c index 6b89b5589ba28..9cac1ba6bb523 100644 --- a/fs/iomap/buffered-io.c +++ b/fs/iomap/buffered-io.c @@ -1177,7 +1177,15 @@ static int iomap_write_delalloc_release(struct inode *inode, error = data_end; goto out_unlock; } - WARN_ON_ONCE(data_end <= start_byte); + + /* + * If we race with post-direct I/O invalidation of the page cache, + * there might be no data left at start_byte. + */ + if (data_end == start_byte) + continue; + + WARN_ON_ONCE(data_end < start_byte); WARN_ON_ONCE(data_end > scan_end_byte); error = iomap_write_delalloc_scan(inode, &punch_start_byte, -- 2.43.0
