From: Peter Zijlstra <peterz@infradead.org>
To: Namhyung Kim <namhyung@kernel.org>
Cc: rcu@vger.kernel.org, intel-gfx@lists.freedesktop.org,
Boqun Feng <boqun.feng@gmail.com>,
LKML <linux-kernel@vger.kernel.org>,
Steven Rostedt <rostedt@goodmis.org>,
Radoslaw Burny <rburny@google.com>,
Byungchul Park <byungchul.park@lge.com>,
"Paul E. McKenney" <paul.mckenney@linaro.org>,
Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
cgroups@vger.kernel.org, Tejun Heo <tj@kernel.org>,
Waiman Long <longman@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>,
Will Deacon <will@kernel.org>, Ingo Molnar <mingo@kernel.org>,
linux-btrfs@vger.kernel.org
Subject: Re: [Intel-gfx] [RFC 00/12] locking: Separate lock tracepoints from lockdep/lock_stat (v1)
Date: Wed, 9 Feb 2022 10:09:08 +0100 [thread overview]
Message-ID: <20220209090908.GK23216@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <20220208184208.79303-1-namhyung@kernel.org>
On Tue, Feb 08, 2022 at 10:41:56AM -0800, Namhyung Kim wrote:
> Eventually I'm mostly interested in the contended locks only and I
> want to reduce the overhead in the fast path. By moving that, it'd be
> easy to track contended locks with timing by using two tracepoints.
So why not put in two new tracepoints and call it a day?
Why muck about with all that lockdep stuff just to preserve the name
(and in the process continue to blow up data structures etc..). This
leaves distros in a bind, will they enable this config and provide
tracepoints while bloating the data structures and destroying things
like lockref (which relies on sizeof(spinlock_t)), or not provide this
at all.
Yes, the name is convenient, but it's just not worth it IMO. It makes
the whole proposition too much of a trade-off.
Would it not be possible to reconstruct enough useful information from
the lock callsite?
WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>, Will Deacon <will@kernel.org>,
Waiman Long <longman@redhat.com>,
Boqun Feng <boqun.feng@gmail.com>,
LKML <linux-kernel@vger.kernel.org>,
Thomas Gleixner <tglx@linutronix.de>,
Steven Rostedt <rostedt@goodmis.org>,
Byungchul Park <byungchul.park@lge.com>,
"Paul E. McKenney" <paul.mckenney@linaro.org>,
Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
Radoslaw Burny <rburny@google.com>, Tejun Heo <tj@kernel.org>,
rcu@vger.kernel.org, cgroups@vger.kernel.org,
linux-btrfs@vger.kernel.org, intel-gfx@lists.freedesktop.org
Subject: Re: [RFC 00/12] locking: Separate lock tracepoints from lockdep/lock_stat (v1)
Date: Wed, 9 Feb 2022 10:09:08 +0100 [thread overview]
Message-ID: <20220209090908.GK23216@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <20220208184208.79303-1-namhyung@kernel.org>
On Tue, Feb 08, 2022 at 10:41:56AM -0800, Namhyung Kim wrote:
> Eventually I'm mostly interested in the contended locks only and I
> want to reduce the overhead in the fast path. By moving that, it'd be
> easy to track contended locks with timing by using two tracepoints.
So why not put in two new tracepoints and call it a day?
Why muck about with all that lockdep stuff just to preserve the name
(and in the process continue to blow up data structures etc..). This
leaves distros in a bind, will they enable this config and provide
tracepoints while bloating the data structures and destroying things
like lockref (which relies on sizeof(spinlock_t)), or not provide this
at all.
Yes, the name is convenient, but it's just not worth it IMO. It makes
the whole proposition too much of a trade-off.
Would it not be possible to reconstruct enough useful information from
the lock callsite?
WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Namhyung Kim <namhyung@kernel.org>
Cc: rcu@vger.kernel.org, intel-gfx@lists.freedesktop.org,
Boqun Feng <boqun.feng@gmail.com>,
LKML <linux-kernel@vger.kernel.org>,
Steven Rostedt <rostedt@goodmis.org>,
Radoslaw Burny <rburny@google.com>,
Byungchul Park <byungchul.park@lge.com>,
"Paul E. McKenney" <paul.mckenney@linaro.org>,
Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
cgroups@vger.kernel.org, Tejun Heo <tj@kernel.org>,
Waiman Long <longman@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>,
Will Deacon <will@kernel.org>, Ingo Molnar <mingo@kernel.org>,
linux-btrfs@vger.kernel.org
Subject: Re: [RFC 00/12] locking: Separate lock tracepoints from lockdep/lock_stat (v1)
Date: Wed, 9 Feb 2022 10:09:08 +0100 [thread overview]
Message-ID: <20220209090908.GK23216@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <20220208184208.79303-1-namhyung@kernel.org>
On Tue, Feb 08, 2022 at 10:41:56AM -0800, Namhyung Kim wrote:
> Eventually I'm mostly interested in the contended locks only and I
> want to reduce the overhead in the fast path. By moving that, it'd be
> easy to track contended locks with timing by using two tracepoints.
So why not put in two new tracepoints and call it a day?
Why muck about with all that lockdep stuff just to preserve the name
(and in the process continue to blow up data structures etc..). This
leaves distros in a bind, will they enable this config and provide
tracepoints while bloating the data structures and destroying things
like lockref (which relies on sizeof(spinlock_t)), or not provide this
at all.
Yes, the name is convenient, but it's just not worth it IMO. It makes
the whole proposition too much of a trade-off.
Would it not be possible to reconstruct enough useful information from
the lock callsite?
next prev parent reply other threads:[~2022-02-09 9:09 UTC|newest]
Thread overview: 108+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-02-08 18:41 [RFC 00/12] locking: Separate lock tracepoints from lockdep/lock_stat (v1) Namhyung Kim
2022-02-08 18:41 ` Namhyung Kim
2022-02-08 18:41 ` [Intel-gfx] " Namhyung Kim
2022-02-08 18:41 ` [PATCH 01/12] locking: Pass correct outer wait type info Namhyung Kim
2022-02-08 18:41 ` [PATCH 02/12] cgroup: rstat: Make cgroup_rstat_cpu_lock name readable Namhyung Kim
2022-02-08 18:41 ` Namhyung Kim
2022-02-08 18:46 ` Tejun Heo
2022-02-08 19:16 ` Namhyung Kim
2022-02-08 19:16 ` Namhyung Kim
2022-02-08 23:51 ` Namhyung Kim
2022-02-08 23:51 ` Namhyung Kim
2022-02-08 18:41 ` [PATCH 03/12] timer: Protect lockdep functions with #ifdef Namhyung Kim
2022-02-08 19:36 ` Steven Rostedt
2022-02-08 20:29 ` Namhyung Kim
2022-02-08 21:19 ` Steven Rostedt
2022-02-08 18:42 ` [PATCH 04/12] workqueue: " Namhyung Kim
2022-02-08 18:48 ` Tejun Heo
2022-02-08 19:17 ` Namhyung Kim
2022-02-08 19:38 ` Steven Rostedt
2022-02-08 18:42 ` [PATCH 05/12] drm/i915: " Namhyung Kim
2022-02-08 18:42 ` [Intel-gfx] " Namhyung Kim
2022-02-08 18:51 ` Jani Nikula
2022-02-08 18:51 ` [Intel-gfx] " Jani Nikula
2022-02-08 19:22 ` Namhyung Kim
2022-02-08 19:22 ` [Intel-gfx] " Namhyung Kim
2022-02-09 13:49 ` Jani Nikula
2022-02-09 13:49 ` [Intel-gfx] " Jani Nikula
2022-02-09 16:27 ` Steven Rostedt
2022-02-09 16:27 ` [Intel-gfx] " Steven Rostedt
2022-02-09 19:28 ` Namhyung Kim
2022-02-09 19:28 ` [Intel-gfx] " Namhyung Kim
2022-02-08 18:42 ` [PATCH 06/12] btrfs: change lockdep class size check using ks->names Namhyung Kim
2022-02-08 19:03 ` David Sterba
2022-02-08 18:42 ` [PATCH 07/12] locking: Introduce CONFIG_LOCK_INFO Namhyung Kim
2022-02-08 18:42 ` [PATCH 08/12] locking/mutex: Init name properly w/ CONFIG_LOCK_INFO Namhyung Kim
2022-02-08 18:42 ` [PATCH 09/12] locking: Add more static lockdep init macros Namhyung Kim
2022-02-08 18:42 ` [PATCH 10/12] locking: Add CONFIG_LOCK_TRACEPOINTS option Namhyung Kim
2022-02-08 18:42 ` [PATCH 11/12] locking/mutex: Revive fast functions for CONFIG_LOCK_TRACEPOINTS Namhyung Kim
2022-02-09 8:40 ` Peter Zijlstra
2022-02-09 20:15 ` Namhyung Kim
2022-02-08 18:42 ` [PATCH 12/12] locking: Move lock_acquired() from the fast path Namhyung Kim
2022-02-08 19:14 ` [RFC 00/12] locking: Separate lock tracepoints from lockdep/lock_stat (v1) Namhyung Kim
2022-02-08 19:14 ` Namhyung Kim
2022-02-08 19:14 ` [Intel-gfx] " Namhyung Kim
2022-02-09 9:09 ` Peter Zijlstra [this message]
2022-02-09 9:09 ` Peter Zijlstra
2022-02-09 9:09 ` Peter Zijlstra
2022-02-09 18:19 ` [Intel-gfx] " Waiman Long
2022-02-09 18:19 ` Waiman Long
2022-02-09 18:19 ` Waiman Long
2022-02-09 18:29 ` Mathieu Desnoyers
2022-02-09 18:29 ` Mathieu Desnoyers
2022-02-09 18:29 ` [Intel-gfx] " Mathieu Desnoyers
2022-02-09 19:02 ` Waiman Long
2022-02-09 19:02 ` Waiman Long
2022-02-09 19:02 ` Waiman Long
2022-02-09 19:17 ` Mathieu Desnoyers
2022-02-09 19:17 ` Mathieu Desnoyers
2022-02-09 19:17 ` [Intel-gfx] " Mathieu Desnoyers
2022-02-09 19:37 ` Waiman Long
2022-02-09 19:37 ` Waiman Long
2022-02-09 19:37 ` Waiman Long
2022-02-09 19:22 ` [Intel-gfx] " Namhyung Kim
2022-02-09 19:22 ` Namhyung Kim
2022-02-09 19:22 ` Namhyung Kim
2022-02-09 19:28 ` Mathieu Desnoyers
2022-02-09 19:28 ` Mathieu Desnoyers
2022-02-09 19:28 ` [Intel-gfx] " Mathieu Desnoyers
2022-02-09 19:45 ` Namhyung Kim
2022-02-09 19:45 ` Namhyung Kim
2022-02-09 19:45 ` [Intel-gfx] " Namhyung Kim
2022-02-09 19:56 ` Mathieu Desnoyers
2022-02-09 19:56 ` Mathieu Desnoyers
2022-02-09 19:56 ` [Intel-gfx] " Mathieu Desnoyers
2022-02-09 20:17 ` Waiman Long
2022-02-09 20:17 ` Waiman Long
2022-02-09 20:17 ` [Intel-gfx] " Waiman Long
2022-02-10 0:27 ` Namhyung Kim
2022-02-10 0:27 ` Namhyung Kim
2022-02-10 0:27 ` Namhyung Kim
2022-02-10 2:12 ` Waiman Long
2022-02-10 2:12 ` Waiman Long
2022-02-10 2:12 ` [Intel-gfx] " Waiman Long
2022-02-10 9:33 ` Peter Zijlstra
2022-02-10 9:33 ` Peter Zijlstra
2022-02-10 9:33 ` [Intel-gfx] " Peter Zijlstra
2022-02-10 0:32 ` Namhyung Kim
2022-02-10 0:32 ` Namhyung Kim
2022-02-10 0:32 ` Namhyung Kim
2022-02-10 9:13 ` Peter Zijlstra
2022-02-10 9:13 ` Peter Zijlstra
2022-02-10 9:13 ` [Intel-gfx] " Peter Zijlstra
2022-02-10 19:14 ` Paul E. McKenney
2022-02-10 19:14 ` Paul E. McKenney
2022-02-10 19:27 ` Waiman Long
2022-02-10 19:27 ` Waiman Long
2022-02-10 19:27 ` [Intel-gfx] " Waiman Long
2022-02-10 20:10 ` Paul E. McKenney
2022-02-10 20:10 ` Paul E. McKenney
2022-02-11 5:57 ` Namhyung Kim
2022-02-11 5:57 ` Namhyung Kim
2022-02-11 5:57 ` [Intel-gfx] " Namhyung Kim
2022-02-11 5:55 ` Namhyung Kim
2022-02-11 5:55 ` Namhyung Kim
2022-02-11 5:55 ` [Intel-gfx] " Namhyung Kim
2022-02-11 10:39 ` Peter Zijlstra
2022-02-11 10:39 ` Peter Zijlstra
2022-02-11 10:39 ` [Intel-gfx] " Peter Zijlstra
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=20220209090908.GK23216@worktop.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=boqun.feng@gmail.com \
--cc=byungchul.park@lge.com \
--cc=cgroups@vger.kernel.org \
--cc=intel-gfx@lists.freedesktop.org \
--cc=linux-btrfs@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=longman@redhat.com \
--cc=mathieu.desnoyers@efficios.com \
--cc=mingo@kernel.org \
--cc=namhyung@kernel.org \
--cc=paul.mckenney@linaro.org \
--cc=rburny@google.com \
--cc=rcu@vger.kernel.org \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=tj@kernel.org \
--cc=will@kernel.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.