On 11/4/19 3:57 PM, Christoph Hellwig wrote: > On Mon, Nov 04, 2019 at 03:55:41PM -0700, Jens Axboe wrote: >>> All existing ones I'm aware of are 128k, so 4k aligned, but if the LBA >>> format is 512B, you could start a 4k IO at a 126k offset to straddle the >>> boundary. Hm, maybe we don't care about the split penalty in that case >>> since unaligned access is already going to be slower for other reasons ... >> >> Yeah, not sure that's a huge concern for that particular case. If you >> think it's a real world issue, it should be possible to set aside a >> queue bit for this and always have them hit the slower split path. > > Well, we use that field for zoned devices also, in which case it is an > issue. I think I'll need to send a patch to skip the fast path if > we have chunk_sectors set. Yes, let's do that, makes sense. -- Jens Axboe
