On Wed, Mar 05, 2025 at 12:45:52AM +0000, Pavel Begunkov wrote: > It's not something recent. After some digging I think the > one I remember is > > https://lore.kernel.org/all/20190717150445.1131-2-axboe@xxxxxxxxx/ > > Remove by > > commit 7b6620d7db566a46f49b4b9deab9fa061fd4b59b > Author: Jens Axboe <axboe@xxxxxxxxx> > Date: Thu Aug 15 11:09:16 2019 -0600 > > block: remove REQ_NOWAIT_INLINE That does indeed look pretty broken, mostly because it doesn't actually return the errors inline as return values but tries to emulate it. > > If you don't want to do synchronous wouldblock errors that's your > > only option. I think it would suck badly, but it's certainly easier > > to backport. > > Is there some intrinsic difference of iomap from the block file > in block/fops.c? Or is that one broken? block/fops.c has roughly the same issue, but it's not anywhere near as sever, as block/fops.c doesn't have any file system state that gets confused by the async BLK_STS_AGAIN. But unless io_uring knows how to rewind the state for that case it probably also doesn't work properly. It might be worth to look into testcases that actually exercise the nowait I/O on block devices with annoyingly small limits or that split a lot (like RAID0).
