[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20250403.120200.1300877147444853564.fujita.tomonori@gmail.com>
Date: Thu, 03 Apr 2025 12:02:00 +0900 (JST)
From: FUJITA Tomonori <fujita.tomonori@...il.com>
To: boqun.feng@...il.com
Cc: fujita.tomonori@...il.com, a.hindborg@...nel.org, tglx@...utronix.de,
linux-kernel@...r.kernel.org, rust-for-linux@...r.kernel.org,
netdev@...r.kernel.org, andrew@...n.ch, hkallweit1@...il.com,
tmgross@...ch.edu, ojeda@...nel.org, alex.gaynor@...il.com,
gary@...yguo.net, bjorn3_gh@...tonmail.com, benno.lossin@...ton.me,
a.hindborg@...sung.com, aliceryhl@...gle.com, anna-maria@...utronix.de,
frederic@...nel.org, arnd@...db.de, jstultz@...gle.com, sboyd@...nel.org,
mingo@...hat.com, peterz@...radead.org, juri.lelli@...hat.com,
vincent.guittot@...aro.org, dietmar.eggemann@....com, rostedt@...dmis.org,
bsegall@...gle.com, mgorman@...e.de, vschneid@...hat.com,
tgunders@...hat.com, me@...enk.dev, david.laight.linux@...il.com
Subject: Re: [PATCH v11 6/8] MAINTAINERS: rust: Add new sections for
DELAY/SLEEP and TIMEKEEPING API
On Wed, 2 Apr 2025 17:51:41 -0700
Boqun Feng <boqun.feng@...il.com> wrote:
> On Thu, Apr 03, 2025 at 08:03:34AM +0900, FUJITA Tomonori wrote:
>> On Wed, 2 Apr 2025 09:29:18 -0700
>> Boqun Feng <boqun.feng@...il.com> wrote:
>>
>> > On Wed, Apr 02, 2025 at 11:16:27PM +0900, FUJITA Tomonori wrote:
>> >> On Mon, 31 Mar 2025 07:03:15 -0700
>> >> Boqun Feng <boqun.feng@...il.com> wrote:
>> >>
>> >> >> My recommendation would be to take all of `rust/kernel/time` under one
>> >> >> entry for now. I suggest the following, folding in the hrtimer entry as
>> >> >> well:
>> >> >>
>> >> >> DELAY, SLEEP, TIMEKEEPING, TIMERS [RUST]
>> >> >> M: Andreas Hindborg <a.hindborg@...nel.org>
>> >> >
>> >> > Given you're the one who would handle the patches, I think this make
>> >> > more sense.
>> >> >
>> >> >> R: Boqun Feng <boqun.feng@...il.com>
>> >> >> R: FUJITA Tomonori <fujita.tomonori@...il.com>
>> >> >
>> >> > Tomo, does this look good to you?
>> >>
>> >> Fine by me.
>> >>
>> >> So a single entry for all the Rust time stuff, which isn't aligned
>> >> with C's MAINTAINERS entries. It's just for now?
>> >>
>> >
>> > Given Andreas is the one who's going to handle the PRs, and he will put
>> > all the things in one branch. I think it's fine even for long term, and
>> > we got all relevant reviewers covered. If the Rust timekeeping + hrtimer
>> > community expands in the future, we can also add more entries. We don't
>> > necessarily need to copy all maintainer structures from C ;-)
>>
>> It seems I was mistaken. I had thought that the ideal goal was for the
>> same team to maintain both the C code and the corresponding Rust code.
>>
>
> Yeah, that was the ideal goal, but Frederic said in the hrtimer series:
>
> https://lore.kernel.org/rust-for-linux/Z8iLIyofy6KGOsq5@localhost.localdomain/
>
> , to me it's clear that hrtimer maintainers want hrtimer Rust patches to
> go to rust tree via Andreas, and given timekeeping maintainers are
> basically the same group of people, and Thomas explicitly asked to be
> added as reviewers:
>
> https://lore.kernel.org/rust-for-linux/87o6xu15m1.ffs@tglx/
>
> It's a clear signal that timekeeping and hrtimer Rust patches are
> preferred to go to rust tree, and Andreas had nicely accepted to handle
> timekeeping and sleep/delay patches, so it makes sense to use one entry
> if he prefers. Hope this clarifies things.
Yeah, I see your point.
>> >> >> I assume patch 1 will go through the sched/core tree, and then Miguel
>> >> >> can pick 7.
>> >> >>
>> >> >
>> >> > Patch 1 & 7 probably should go together, but we can decide it later.
>> >>
>> >> Since nothing has moved forward for quite a while, maybe it's time to
>> >> drop patch 1.
>> >
>> > No, I think we should keep it. Because otherwise we will use a macro
>>
>> Yeah, I know. the first version of this uses a macro.
>>
>>
>> > version of read_poll_timeout(), which is strictly worse. I'm happy to
>> > collect patch #1 and the cpu_relax() patch of patch #7, and send an PR
>> > to tip. Could you split them a bit:
>> >
>> > * Move the Rust might_sleep() in patch #7 to patch #1 and put it at
>> > kernel::task, also if we EXPORT_SYMBOL(__might_sleep_precision), we
>> > don't need the rust_helper for it.
>> >
>> > * Have a separate containing the cpu_relax() bit.
>> >
>> > * Also you may want to put #[inline] at cpu_relax() and might_resched().
>> >
>> > and we can start from there. Sounds good?
>>
>> I can do whatever but I don't think these matters. The problem is that
>
> Confused. I said I would do a PR, that means if no objection, the
> patches will get merged. Isn't this a way to move forward? Or you're
> against that I'm doing a PR?
I don't object to you doing a PR.
I meant that it's unclear whether we can move forward with the
approach, as we haven't received much response from the maintainers
and we don't know what the blockers are.
>> we haven't received a response from the scheduler maintainers for a
>> long time. We don't even know if the implementation is actually an
>> issue.
>>
>
> If there's an issue, I can fix it. After all, printk confirmed that
> ".*s" format works for this case:
>
> https://lore.kernel.org/rust-for-linux/ZyyAsjsz05AlkOBd@pathway.suse.cz/
>
> and Peter sort of confirmed he's not against the idea:
>
> https://lore.kernel.org/rust-for-linux/20250201121613.GC8256@noisy.programming.kicks-ass.net/
>
> I don't see any major issue blocking this. But of course, I'm happy to
> resolve if there is one.
I know but this patch adds a workaround to the C code for Rust’s sake,
I would understand if the maintainers reject it and prefer having Rust
work around the issue instead.
Anyway, let's try one more time. I'll modify the patch #1 as you
suggested and send a new patchset.
Thanks!
Powered by blists - more mailing lists