Re: [PATCH V4 6/6] loop: add hint for handling aio via IOCB_NOWAIT

[Ming Lei <ming.lei@redhat.com>]

On Mon, Oct 06, 2025 at 11:33:17PM -0700, Christoph Hellwig wrote:
> On Mon, Oct 06, 2025 at 10:18:12PM +0800, Ming Lei wrote:
> > On Fri, Oct 03, 2025 at 12:06:44AM -0700, Christoph Hellwig wrote:
> > > On Sun, Sep 28, 2025 at 09:29:25PM +0800, Ming Lei wrote:
> > > > - there isn't any queued blocking async WRITEs, because NOWAIT won't cause
> > > > contention with blocking WRITE, which often implies exclusive lock
> > > 
> > > Isn't this a generic thing we should be doing in core code so that
> > > it applies to io_uring I/O as well?
> > 
> > No.
> > 
> > It is just policy of using NOWAIT or not, so far:
> > 
> > - RWF_NOWAIT can be set from preadv/pwritev
> > 
> > - used for handling io_uring FS read/write
> > 
> > Even though loop's situation is similar with io-uring, however, both two are
> > different subsystem, and there is nothing `core code` for both, more importantly
> > it is just one policy: use it or not use it, each subsystem can make its
> > own decision based on subsystem internal.
> 
> I fail to parse what you say here.  You are encoding special magic
> about what underlying file systems do in an upper layer.  I'd much

NOWAIT is obviously interface provided by FS, here loop just wants to try
NOWAIT first in block layer dispatch context for avoiding the extra wq
schedule latency.

But for write on sparse file, trying NOWAIT first may bring extra retry
cost, that is why the hint is added. It is very coarse, but potential
regression can be avoided.

> rather have a flag similar FOP_DIO_PARALLEL_WRITE that makes this
> limitation clear rather then opencoding it in the loop driver while

What is the limitation?

> leabing the primary user of RWF_NOWAIT out in the cold.

FOP_DIO_PARALLEL_WRITE is one static FS feature, but here it is FS
runtime behavior, such as if the write can be blocked because of space
allocation, so it can't be done by one static flag.

io-uring shares nothing with loop in this area, it is just one policy wrt.
use NOWAIT or not. I don't understand why you insist on covering both
from FS internal...



Thanks,
Ming