Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Sender: "axelrasmussen via sendgmr" <axelrasmussen@ajr0.svl.corp.google.com>
Date:   Thu,  5 Nov 2020 13:17:38 -0800
Message-Id: <20201105211739.568279-1-axelrasmussen@google.com>
Mime-Version: 1.0
Subject: [PATCH v6 0/1] mmap_lock: add tracepoints around lock acquisition
From:   Axel Rasmussen <axelrasmussen@google.com>
To:     Steven Rostedt <rostedt@goodmis.org>,
        Ingo Molnar <mingo@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Michel Lespinasse <walken@google.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        Daniel Jordan <daniel.m.jordan@oracle.com>,
        Jann Horn <jannh@google.com>,
        Chinwen Chang <chinwen.chang@mediatek.com>,
        Davidlohr Bueso <dbueso@suse.de>,
        David Rientjes <rientjes@google.com>,
        Laurent Dufour <ldufour@linux.ibm.com>
Cc:     Yafang Shao <laoar.shao@gmail.com>, linux-kernel@vger.kernel.org,
        linux-mm@kvack.org, Axel Rasmussen <axelrasmussen@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

This patchset adds tracepoints around mmap_lock acquisition. This is useful so
we can measure the latency of lock acquisition, in order to detect contention.

This version is based on v5.10-rc2.

Changes since v5:

- Michel pointed out that rwsem_release in mmap_read_trylock_non_owner doesn't
  actually release the lock, it just releases lockdep's ownership tracking. So,
  it's incorrect to call __mmap_lock_trace_released there, so the call has been
  removed.

Changes since v4:

- Redesigned buffer allocation to deal with the fact that a trace event might be
  interrupted by e.g. an IRQ, for which a per-cpu buffer is insufficient. Now we
  allocate one buffer per CPU * one buffer per context we might be called in
  (currently 4: normal, irq, softirq, NMI). We have three trace events which can
  potentially all be enabled, and all of which need a buffer; to avoid further
  multiplying the number of buffers by 3, they share the same set of buffers,
  which requires a spinlock + counter setup so we only allocate the buffers
  once, and then free them only when *all* of the trace events are _unreg()-ed.

Changes since v3:

- Switched EXPORT_SYMBOL to EXPORT_TRACEPOINT_SYMBOL, removed comment.

- Removed redundant trace_..._enabled() check.

- Defined the three TRACE_EVENTs separately, instead of sharing an event class.
  The tradeoff is 524 more bytes in .text, but the start_locking and released
  events no longer have a vestigial "success" field, so they're simpler +
  faster.

Changes since v2:

- Refactored tracing helper functions so the helpers are simper, but the locking
  functinos are slightly more verbose. Overall, this decreased the delta to
  mmap_lock.h slightly.

- Fixed a typo in a comment. :)

Changes since v1:

- Functions renamed to reserve the "trace_" prefix for actual tracepoints.

- We no longer measure the duration directly. Instead, users are expected to
  construct a synthetic event which computes the interval between "start
  locking" and "acquire returned".

- The new helper for checking if tracepoints are enabled in a header is used to
  avoid un-inlining any of the lock wrappers. This yields ~zero overhead if the
  tracepoints aren't enabled, and therefore obviates the need for a Kconfig for
  this change.

[1] https://lore.kernel.org/patchwork/patch/1316922/
[2] https://lore.kernel.org/patchwork/patch/1311996/

Axel Rasmussen (1):
  mmap_lock: add tracepoints around lock acquisition

 include/linux/mmap_lock.h        |  94 +++++++++++++++-
 include/trace/events/mmap_lock.h | 107 ++++++++++++++++++
 mm/Makefile                      |   2 +-
 mm/mmap_lock.c                   | 187 +++++++++++++++++++++++++++++++
 4 files changed, 384 insertions(+), 6 deletions(-)
 create mode 100644 include/trace/events/mmap_lock.h
 create mode 100644 mm/mmap_lock.c

--
2.29.1.341.ge80a0c044ae-goog