Re: [PATCH v3 4/4] io_uring: add support for zone-append

[Kanchan Joshi]

On Fri, Jul 10, 2020 at 7:13 PM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
>
> On Fri, Jul 10, 2020 at 06:59:45PM +0530, Kanchan Joshi wrote:
> > > block doesn't work for the case of writes to files that don't have
> > > to be aligned in any way.  And that I think is the more broadly
> > > applicable use case than zone append on block devices.
> >
> > But when can it happen that we do zone-append on a file (zonefs I
> > asssume), and device returns a location (write-pointer essentially)
> > which is not in multiple of 512b?
>
> All the time.  You open a file with O_APPEND.  You write a record to
> it of any kind of size, then the next write will return the position
> it got written at, which can be anything.
I understand if this is about cached write and we are talking about
O_APPEND in general.
But for direct block I/O write and ZoneFS writes, page-cache is not
used, so write(and zone-append result) will be aligned to underlying
block size.
Even though this patchset uses O_APPEND, it filters regular files and
non zoned-block devices by using new FMODE_ZONE_APPEND flag.

-- 
Joshi