Date:   Thu, 18 Jun 2020 16:55:16 +0206
From:   John Ogness <john.ogness@...utronix.de>
To:     Petr Mladek <pmladek@...e.com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Andrea Parri <parri.andrea@...il.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Paul McKenney <paulmck@...nel.org>, kexec@...ts.infradead.org,
        linux-kernel@...r.kernel.org
Subject: [PATCH v3 0/3] printk: replace ringbuffer

Hello,

Here is a v3 for the first series to rework the printk
subsystem. The v2 and history are here [0]. This first series
only replaces the existing ringbuffer implementation. No locking
is removed. No semantics/behavior of printk are changed.

Reviews on the ringbuffer are still ongoing, but I was asked to
post this new version since several changes from v2 have been
already agreed upon.

The series is based on v5.8-rc1.

The list of changes since v2:

printk.c
========

- console_unlock(): fix extended console printing [1]

printk_ringbuffer
=================

- data_push_tail(): fixed handling when another CPU already
  pushed the tail [2]

- desc_push_tail(): added a full memory barrier before the
  descriptor tail push [3]

- data_make_reusable()/data_alloc(): changed block ID
  reading/writing from READ_ONCE()/WRITE_ONCE() to regular
  assignments [4]

- data_make_reusable(): removed unnecessary data tail
  re-check [5]

- general: folded all smp_mb() and smp_wmb() calls into their
  neighboring cmpxchg_relaxed(), changing them to full cmpxchg()
  calls [6]

- desc_read(): changed descriptor content reading from
  READ_ONCE() to memcpy() since it served no purpose for the
  legal data race

- general: cleaned up memory barrier comments; in particular
  made sure that the reader and writer sides of the memory
  barrier pairs match in their descriptions

- added a new section in the memory barrier documentation (near
  the top of printk_ringbuffer.c) that lists all 10 memory
  barrier pairs and briefly describes what they are ordering

- _prb_read_valid(): changed the helper function to static

- general: changed block size argument type from "unsigned long"
  to "unsigned int" since a record's *_buf_size fields are of
  type "unsigned int"

- general: allow some lines to go beyond 80 characters

John Ogness

[0] https://lkml.kernel.org/r/20200501094010.17694-1-john.ogness@linutronix.de
[1] https://lkml.kernel.org/r/87ftcd86d2.fsf@vostro.fn.ogness.net
[2] https://lkml.kernel.org/r/87v9ktcs3q.fsf@vostro.fn.ogness.net
[3] https://lkml.kernel.org/r/87bllpyzgr.fsf@vostro.fn.ogness.net
[4] https://lkml.kernel.org/r/87tuzkuxtw.fsf@vostro.fn.ogness.net
[5] https://lkml.kernel.org/r/87pna5mjtp.fsf@vostro.fn.ogness.net

John Ogness (3):
  crash: add VMCOREINFO macro to define offset in a struct declared by
    typedef
  printk: add lockless ringbuffer
  printk: use the lockless ringbuffer

 include/linux/crash_core.h        |    3 +
 include/linux/kmsg_dump.h         |    2 -
 kernel/printk/Makefile            |    1 +
 kernel/printk/printk.c            |  944 ++++++++--------
 kernel/printk/printk_ringbuffer.c | 1674 +++++++++++++++++++++++++++++
 kernel/printk/printk_ringbuffer.h |  352 ++++++
 6 files changed, 2527 insertions(+), 449 deletions(-)
 create mode 100644 kernel/printk/printk_ringbuffer.c
 create mode 100644 kernel/printk/printk_ringbuffer.h

-- 
2.20.1

Powered by blists - more mailing lists