[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <cover.1622558659.git.asml.silence@gmail.com>
Date: Tue, 1 Jun 2021 15:58:25 +0100
From: Pavel Begunkov <asml.silence@...il.com>
To: io-uring@...r.kernel.org
Cc: Jens Axboe <axboe@...nel.dk>, Andres Freund <andres@...razel.de>,
Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...hat.com>,
Peter Zijlstra <peterz@...radead.org>,
Darren Hart <dvhart@...radead.org>,
Davidlohr Bueso <dave@...olabs.net>,
linux-kernel@...r.kernel.org
Subject: [RFC 0/4] futex request support
Proof of concept for io_uring futex requests. The wake side does
FUTEX_WAKE_OP type of modify-compare operation but with a single
address. Wait reqs go through io-wq and so slow path.
Should be interesting for a bunch of people, so we should first
outline API and capabilities it should give. As I almost never
had to deal with futexes myself, would especially love to hear
use case, what might be lacking and other blind spots.
1) Do we need PI?
2) Do we need requeue? Anything else?
3) How hot waits are? May be done fully async avoiding io-wq, but
apparently requires more changes in futex code.
4) We can avoid all page locking/unlocking for shared futexes
with pre-registration. How much is it interesting?
For _very_ basic examples see [1]
[1] https://github.com/isilence/liburing/tree/futex_v1
Pavel Begunkov (4):
futex: add op wake for a single key
io_uring: frame out futex op
io_uring: support futex wake requests
io_uring: implement futex wait
fs/io_uring.c | 91 +++++++++++++++++++++++++++++++++++
include/linux/futex.h | 15 ++++++
include/uapi/linux/io_uring.h | 19 +++++++-
kernel/futex.c | 64 +++++++++++++++++++++++-
4 files changed, 186 insertions(+), 3 deletions(-)
--
2.31.1
Powered by blists - more mailing lists