On Wed, 9 Aug 2023, Christoph Hellwig wrote: > Please do not add a new ->direct_IO method. I'm currently working hard > on removing it, just set FMODE_CAN_ODIRECT and handle the fallback in > your read_iter/write_iter methods. Thanks for the input, I'd missed that FMODE_CAN_ODIRECT development. I can see why you would surely prefer not to have a .direct_IO added. But whether that's right for tmpfs at this time, I'll let you and all decide: I've tried and tested the v2 patch now, and will send it out shortly; but it has to add a shmem_file_write_iter(), where shmem was doing fine with generic_file_write_iter() + direct_IO() stub before. So my own feeling is that the v1 patch with shmem_direct_IO() was better, duplicating less code; but whatever, you can all decide between them. > > But if we just start claiming direct I/O support for file systems that > don't actually support it, I'm starting to seriously wonder why we > bother with the flag at all and don't just allow O_DIRECT opens > to always succeed.. Yes, I've wondered that way too, but don't have a strong opinion on it. Hugh
