From: Jens Axboe <axboe@kernel.dk>
To: "Christian Brauner" <brauner@kernel.org>,
"Méven Car" <meven29@gmail.com>
Cc: io-uring@vger.kernel.org, linux-fsdevel@vger.kernel.org,
"Alexander Viro" <viro@zeniv.linux.org.uk>,
"Jan Kara" <jack@suse.cz>, "Méven Car" <meven@kde.org>
Subject: Re: [RFC PATCH v3 0/2] io_uring: add IORING_OP_UTIMENSAT
Date: Fri, 20 Nov 2026 09:34:28 -0700 [thread overview]
Message-ID: <0460f99b-ab87-46f3-ba58-2daba2e821db@kernel.dk> (raw)
In-Reply-To: <20261120-vorfeld-formlos-nahbereich-00aebc4c5150@brauner>
On 11/20/26 3:33 AM, Christian Brauner wrote:
> On Sat, Nov 14, 2026 at 11:46:19AM +0100, M?ven Car wrote:
>> IORING_OP_FCHMOD was proposed in 2024 [1] and not merged. The same
>> questions apply to fchmod and fchown, which a copy also needs, and a
>> single setattr opcode could set the mode, owner and times in one
>> ->setattr() call. I started with the times because a copy always sets
>> them.
>
> To me it seems strange to grow an io_uring opcode per attribute. What
> makes the most sense to me is a common interface to set multiple
> attributes in one shot that is shared between io_uring and the regular
> system call path. IOW, an extensible struct based system call that takes
> the mode, owner and times and does a single ->setattr() and an io_uring
> opcode that uses the same struct. Then a copy sets everything at once
> and userspace that doesn't use io_uring gets the same thing and it would
> reduce the number of system calls for userspace and less work for
> io_uring as well.
>
> The other thing that this would allow is for filesystems to implement
> multi-attribute changes in one transaction.
Arrays of things is usually unhandy for io_uring, eg if the idea is to
have an array of structs to deal with. It's a lot more expensive for
persisting the data, as you know need to copy a bunch of it, rather than
just have the single item which will fit in the per-sqe data for
io_uring and not require any allocations. An example are things like
sendmmsg - that kind of construct just doesn't make sense for io_uring,
as you could've just done N SQEs instead.
It also becomes problematic with the non-blocking aspect to deal with
multiples at the time, dealing with "yep first N worked, N+1 would've
blocked" which then turns into do you blocking retry the latter bits, or
do you just return N done and let the app deal with it.
With that said, I need to actually set some time aside to look at this
patchset and render an informed opinion :-)
--
Jens Axboe
prev parent reply other threads:[~2026-11-20 16:34 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-14 10:46 [RFC PATCH v3 0/2] io_uring: add IORING_OP_UTIMENSAT Méven Car
2026-11-14 10:46 ` [RFC PATCH v3 1/2] fs: add filename_utimensat() Méven Car
2026-11-14 10:46 ` [RFC PATCH v3 2/2] io_uring: add support for utimensat Méven Car
2026-11-20 10:33 ` [RFC PATCH v3 0/2] io_uring: add IORING_OP_UTIMENSAT Christian Brauner
2026-11-20 16:34 ` Jens Axboe [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=0460f99b-ab87-46f3-ba58-2daba2e821db@kernel.dk \
--to=axboe@kernel.dk \
--cc=brauner@kernel.org \
--cc=io-uring@vger.kernel.org \
--cc=jack@suse.cz \
--cc=linux-fsdevel@vger.kernel.org \
--cc=meven29@gmail.com \
--cc=meven@kde.org \
--cc=viro@zeniv.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox