[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20230529191416.53955-1-mathieu.desnoyers@efficios.com>
Date: Mon, 29 May 2023 15:14:12 -0400
From: Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: linux-kernel@...r.kernel.org, Thomas Gleixner <tglx@...utronix.de>,
"Paul E . McKenney" <paulmck@...nel.org>,
Boqun Feng <boqun.feng@...il.com>,
"H . Peter Anvin" <hpa@...or.com>, Paul Turner <pjt@...gle.com>,
linux-api@...r.kernel.org, Christian Brauner <brauner@...nel.org>,
Florian Weimer <fw@...eb.enyo.de>, David.Laight@...LAB.COM,
carlos@...hat.com, Peter Oskolkov <posk@...k.io>,
Alexander Mikhalitsyn <alexander@...alicyn.com>,
Chris Kennelly <ckennelly@...gle.com>,
Ingo Molnar <mingo@...hat.com>,
Darren Hart <dvhart@...radead.org>,
Davidlohr Bueso <dave@...olabs.net>,
André Almeida <andrealmeid@...lia.com>,
libc-alpha@...rceware.org, Steven Rostedt <rostedt@...dmis.org>,
Jonathan Corbet <corbet@....net>,
Noah Goldstein <goldstein.w.n@...il.com>,
Daniel Colascione <dancol@...gle.com>, longman@...hat.com,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
Subject: [RFC PATCH v2 0/4] Extend rseq with sched_state_ptr field
This prototype extends struct rseq with a new sched_state_ptr field,
which points to a structure containing a "on-cpu" flag kept up-to-date
by the scheduler.
It is meant to be used by userspace adaptative mutexes to decide between
busy-wait and futex wait system call (releasing the CPU) behaviors based
on the current state of the mutex owner.
The goal is to improve use-cases where the duration of the critical
sections for a given lock follows a multi-modal distribution, preventing
statistical guesses from doing a good job at choosing between busy-wait
and futex wait behavior.
This is in response to the LWN coverage of 2023 Open Source Summit North
America (https://lwn.net/Articles/931789/) unscheduled slot "Adaptive
spinning in user space" presented by André Almeida.
New in this v2:
- Introduce a "struct rseq_sched_state", which contains the on-cpu
scheduler flag and a thread ID field. This eliminates false sharing
on the struct rseq cache lines caused by busy-waiting.
I have favored adding a "thread ID" field to struct rseq_sched_state
rather than adding stores of owner pointer in addition to a
compare-and-swap and store on a uint32_t for lock state to minimize the
number of stores to perform on the fast-path.
Feedback is welcome!
Mathieu
Mathieu Desnoyers (4):
rseq: Add sched_state field to struct rseq
selftests/rseq: Add sched_state rseq field and getter
selftests/rseq: Implement sched state test program
selftests/rseq: Implement rseq_mutex test program
include/linux/sched.h | 16 +++
include/uapi/linux/rseq.h | 41 ++++++
kernel/rseq.c | 43 +++++++
tools/testing/selftests/rseq/.gitignore | 2 +
tools/testing/selftests/rseq/Makefile | 3 +-
tools/testing/selftests/rseq/rseq-abi.h | 42 ++++++
tools/testing/selftests/rseq/rseq.c | 13 ++
tools/testing/selftests/rseq/rseq.h | 5 +
tools/testing/selftests/rseq/rseq_mutex.c | 120 ++++++++++++++++++
.../testing/selftests/rseq/sched_state_test.c | 72 +++++++++++
10 files changed, 356 insertions(+), 1 deletion(-)
create mode 100644 tools/testing/selftests/rseq/rseq_mutex.c
create mode 100644 tools/testing/selftests/rseq/sched_state_test.c
--
2.25.1
Powered by blists - more mailing lists