On Sun, Oct 30, 2022 at 09:01:31AM +1100, Dave Chinner wrote: > Right, think of iomap_truncate_page() as having exactly the same > responsibilites as block_truncate_page() has for filesystems using > bufferheads. i.e. both functions need to ensure the disk contents > are correctly zeroed such that the caller can safely call > truncate_setsize() afterwards resulting in both the on-disk state > and in-memory state remaining coherent. XFS always had these kinds of hacks even when it was using block_truncate_page. That being said, I fully agree that handling it in iomap is the absolutely right thing to do and we should have done it earlier.
