On Wed, Jul 30, 2025 at 12:28:41PM +0200, Jan Kara wrote:
Commit d279c80e0bac ("iomap: inline iomap_dio_bio_opflags()") has broken the logic in iomap_dio_bio_iter() in a way that when the device does support FUA (or has no writeback cache) and the direct IO happens to freshly allocated or unwritten extents, we will *not* issue fsync after completing direct IO O_SYNC / O_DSYNC write because the IOMAP_DIO_WRITE_THROUGH flag stays mistakenly set.
Uh-oh.
Looks good:
Reviewed-by: Christoph Hellwig hch@lst.de