All of lore.kernel.org
 help / color / mirror / Atom feed
From: John Ogness <john.ogness@linutronix.de>
To: Petr Mladek <pmladek@suse.com>
Cc: Andrea Parri <parri.andrea@gmail.com>,
	Andrea Parri <andrea.parri@amarulasolutions.com>,
	Sergey Senozhatsky <sergey.senozhatsky@gmail.com>,
	Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Brendan Higgins <brendanhiggins@google.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	linux-kernel@vger.kernel.org
Subject: Re: dataring_push() barriers Re: [RFC PATCH v4 1/9] printk-rb: add a new printk ringbuffer implementation
Date: Wed, 28 Aug 2019 15:43:50 +0200	[thread overview]
Message-ID: <87pnkpjtgp.fsf@linutronix.de> (raw)
In-Reply-To: <20190827143635.4taqjj6wjz7gdlea@pathway.suse.cz> (Petr Mladek's message of "Tue, 27 Aug 2019 16:36:35 +0200")

On 2019-08-27, Petr Mladek <pmladek@suse.com> wrote:
>>>> +/**
>>>> + * dataring_push() - Reserve a data block in the data array.
>>>> + *
>>>> + * @dr:   The data ringbuffer to reserve data in.
>>>> + *
>>>> + * @size: The size to reserve.
>>>> + *
>>>> + * @desc: A pointer to a descriptor to store the data block information.
>>>> + *
>>>> + * @id:   The ID of the descriptor to be associated.
>>>> + *        The data block will not be set with @id, but rather initialized with
>>>> + *        a value that is explicitly different than @id. This is to handle the
>>>> + *        case when newly available garbage by chance matches the descriptor
>>>> + *        ID.
>>>> + *
>>>> + * This function expects to move the head pointer forward. If this would
>>>> + * result in overtaking the data array index of the tail, the tail data block
>>>> + * will be invalidated.
>>>> + *
>>>> + * Return: A pointer to the reserved writer data, otherwise NULL.
>>>> + *
>>>> + * This will only fail if it was not possible to invalidate the tail data
>>>> + * block.
>>>> + */
>>>> +char *dataring_push(struct dataring *dr, unsigned int size,
>>>> +		    struct dr_desc *desc, unsigned long id)
>>>> +{
>>>> +	unsigned long begin_lpos;
>>>> +	unsigned long next_lpos;
>>>> +	struct dr_datablock *db;
>>>> +	bool ret;
>>>> +
>>>> +	to_db_size(&size);
>>>> +
>>>> +	do {
>>>> +		/* fA: */
>>>> +		ret = get_new_lpos(dr, size, &begin_lpos, &next_lpos);
>>>> +
>>>> +		/*
>>>> +		 * fB:
>>>> +		 *
>>>> +		 * The data ringbuffer tail may have been pushed (by this or
>>>> +		 * any other task). The updated @tail_lpos must be visible to
>>>> +		 * all observers before changes to @begin_lpos, @next_lpos, or
>>>> +		 * @head_lpos by this task are visible in order to allow other
>>>> +		 * tasks to recognize the invalidation of the data
>>>> +		 * blocks.
>>>
>>> This sounds strange. The write barrier should be done only on CPU
>>> that really modified tail_lpos. I.e. it should be in _dataring_pop()
>>> after successful dr->tail_lpos modification.
>> 
>> The problem is that there are no data dependencies between the different
>> variables. When a new datablock is being reserved, it is critical that
>> all other observers see that the tail_lpos moved forward _before_ any
>> other changes. _dataring_pop() uses an smp_rmb() to synchronize for
>> tail_lpos movement.
>
> It should be symmetric. It makes sense that _dataring_pop() uses an
> smp_rmb(). Then there should be wmb() in dataring_push().

dataring_pop() is adjusting the tail. dataring_push() is adjusting the
head. These operations are handled (ordered) separately. They do not
need be happening in lockstep. They don't need to be happening on the
same CPU.

> The wmb() should be done only by the CPU that actually did the write.
> And it should be done after the write. This is why I suggested to
> do it after cmpxchg(dr->head_lpos).

If CPU0 issues an smp_wmb() after moving the tail and (after seeing the
moved tail) CPU1 issues an smp_wmb() after updating the head, it is
still possible for CPU2 to see the head move (and possibly even overtake
the tail) before seeing the tail move.

If a CPU didn't move the tail but _will_ move the head, only a full
memory barrier will allow _all_ observers to see the tail move before
seeing the head move.

>> This CPU is about to make some changes and may have seen an updated
>> tail_lpos. An smp_wmb() is useless if this is not the CPU that
>> performed that update. The full memory barrier ensures that all other
>> observers will see what this CPU sees before any of its future
>> changes are seen.
>
> I do not understand it. Full memory barrier will not cause that all
> CPUs will see the same.

I did not write that. I wrote (emphasis added):

    The full memory barrier ensures that all other observers will see
    what _this_ CPU sees before any of _its_ future changes are seen.

> These barriers need to be symmetric.

They are. The comments for fB list the pairs (all being
smp_mb()/smp_rmb() pairings).

> Back to our situation:
>
>     + rmb() should not be needed here because get_new_lpos() provided
>       a valid lpos.
>
>     + wmb() is not needed because we have not written anything yet
>
> If there was a race with another CPU than cmpxchg(dr->head_lpos)
> would fail and we will need to repeat everything again.

It's not about racing to update the head. It's about making sure that
_all_ CPUs observe that a datablock was invalidated _before_ observing
that _this_ CPU started modifying other shared variables. And again,
this CPU might _not_ be the one that invalidated the datablock
(i.e. moved the tail).

John Ogness

  reply	other threads:[~2019-08-28 13:45 UTC|newest]

Thread overview: 134+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-07 22:26 [RFC PATCH v4 0/9] printk: new ringbuffer implementation John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 1/9] printk-rb: add a new printk " John Ogness
2019-08-20  8:15   ` numlist_pop(): " Petr Mladek
2019-08-21  5:41     ` John Ogness
2019-09-04 12:19     ` Peter Zijlstra
2019-08-20  8:22   ` assign_desc() barriers: " Petr Mladek
2019-08-20 14:14     ` Petr Mladek
2019-08-21  5:52       ` John Ogness
2019-08-22 11:53         ` Petr Mladek
2019-08-25  2:06           ` John Ogness
2019-08-26  8:21             ` John Ogness
2019-08-20  8:55   ` comments style: " Petr Mladek
2019-08-20  9:27     ` Sergey Senozhatsky
2019-08-21  5:46       ` John Ogness
2019-08-22 13:50         ` Petr Mladek
2019-08-22 17:38           ` Andrea Parri
2019-08-23 10:47             ` Petr Mladek
2019-08-23 14:27               ` Andrea Parri
2019-08-23  9:49           ` Sergey Senozhatsky
2019-08-23  5:54         ` Sergey Senozhatsky
2019-08-23 10:29           ` Petr Mladek
2019-08-21  5:42     ` John Ogness
2019-08-22 12:44       ` Petr Mladek
2019-08-20 13:50   ` dataring_push() barriers " Petr Mladek
2019-08-25  2:42     ` John Ogness
2019-08-27 14:36       ` Petr Mladek
2019-08-28 13:43         ` John Ogness [this message]
2019-08-20 15:12   ` datablock reuse races " Petr Mladek
2019-08-23  9:21   ` numlist_push() barriers " Petr Mladek
2019-08-26  8:34     ` Andrea Parri
2019-08-26  8:43       ` Andrea Parri
2019-08-26 14:10       ` Petr Mladek
2019-08-26 16:01         ` Andrea Parri
2019-08-26 22:36     ` John Ogness
2019-08-27  7:40       ` Petr Mladek
2019-08-27 14:28         ` John Ogness
2019-08-27 15:07           ` Petr Mladek
2019-08-28 10:24             ` John Ogness
2019-08-23 17:18   ` numlist API " Petr Mladek
2019-08-26 23:57     ` John Ogness
2019-08-27 13:03       ` Petr Mladek
2019-08-28  7:13         ` John Ogness
2019-08-28  8:58           ` Petr Mladek
2019-08-28 14:03             ` John Ogness
2019-08-29 11:28               ` Petr Mladek
2019-09-03  7:58         ` Sergey Senozhatsky
2019-08-30 14:48   ` dataring " Petr Mladek
2019-08-07 22:26 ` [RFC PATCH v4 2/9] printk-rb: add test module John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 3/9] printk-rb: fix missing includes/exports John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 4/9] printk-rb: initialize new descriptors as invalid John Ogness
2019-08-20  9:23   ` Petr Mladek
2019-08-20 10:16     ` Sergey Senozhatsky
2019-08-21  5:56     ` John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 5/9] printk-rb: remove extra data buffer size allocation John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 6/9] printk-rb: adjust test module ringbuffer sizes John Ogness
2019-08-19 21:29   ` [PATCH] printk-rb: fix test module macro usage John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 7/9] printk-rb: increase size of seq and size variables John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 8/9] printk-rb: new functionality to support printk John Ogness
2019-08-20  9:59   ` Sergey Senozhatsky
2019-08-21  5:47     ` John Ogness
2019-08-07 22:26 ` [RFC PATCH v4 9/9] printk: use a new ringbuffer implementation John Ogness
2019-08-08 19:07   ` Linus Torvalds
2019-08-08 22:55     ` John Ogness
2019-08-08 23:33       ` Linus Torvalds
2019-08-08 23:45         ` Steven Rostedt
2019-08-09  0:21           ` Linus Torvalds
2019-08-09  0:48             ` Steven Rostedt
2019-08-09  1:15               ` Linus Torvalds
2019-08-09 11:15                 ` Thomas Gleixner
2019-08-09 16:00                   ` Linus Torvalds
2019-08-09 20:07                     ` Thomas Gleixner
2019-08-09 20:20                       ` Linus Torvalds
2019-08-09  6:14     ` Peter Zijlstra
2019-08-09  7:08       ` John Ogness
2019-08-09 15:57       ` Linus Torvalds
2019-08-10  5:53         ` Thomas Gleixner
2019-09-10  3:19           ` Sergey Senozhatsky
2019-08-12  9:54       ` Geert Uytterhoeven
2019-08-16  5:46   ` Dave Young
2019-08-16  5:46     ` Dave Young
2019-08-16  5:54     ` Dave Young
2019-08-16  5:54       ` Dave Young
2019-08-16  9:40     ` John Ogness
2019-08-16  9:40       ` John Ogness
2019-09-04 12:35 ` [RFC PATCH v4 0/9] printk: " Peter Zijlstra
2019-09-05 13:05   ` Petr Mladek
2019-09-05 14:31     ` Peter Zijlstra
2019-09-05 15:38       ` Thomas Gleixner
2019-09-05 16:11         ` Steven Rostedt
2019-09-05 21:10           ` John Ogness
2019-09-06  9:39           ` Petr Mladek
2019-09-09 14:11           ` printk meeting at LPC Thomas Gleixner
2019-09-13 13:26             ` John Ogness
2019-09-13 14:48               ` Daniel Vetter
2019-09-15 13:47                 ` John Ogness
2019-09-16  8:44                   ` Daniel Vetter
2019-09-16  4:30               ` Tetsuo Handa
2019-09-16 10:46                 ` Petr Mladek
2019-09-16 13:43                   ` Steven Rostedt
2019-09-16 14:28                     ` John Ogness
2019-09-17  8:11                       ` Petr Mladek
2019-09-17  7:52                     ` Petr Mladek
2019-09-17 13:02                       ` Steven Rostedt
2019-09-17 13:12                         ` Greg Kroah-Hartman
2019-09-17 13:37                           ` Steven Rostedt
2019-09-17 14:08                             ` Tetsuo Handa
2019-09-17  7:51                   ` Sergey Senozhatsky
2019-09-18  1:25               ` Sergey Senozhatsky
2019-09-18  2:08                 ` Steven Rostedt
2019-09-18  2:36                   ` Sergey Senozhatsky
2019-09-18  5:19                     ` Sergey Senozhatsky
2019-09-18  7:42                       ` John Ogness
2019-09-18  8:10                         ` Sergey Senozhatsky
2019-09-18  9:05                           ` John Ogness
2019-09-18  9:11                             ` Sergey Senozhatsky
2019-09-18 16:41                             ` Petr Mladek
2019-09-18 16:48                               ` Steven Rostedt
2019-09-24 14:24                                 ` Petr Mladek
2019-09-19  8:06                         ` Daniel Vetter
2019-09-18  7:33                     ` John Ogness
2019-09-18  8:08                       ` Sergey Senozhatsky
2019-10-04 14:48               ` Tony Asleson
2019-10-07 12:01                 ` Petr Mladek
2019-09-06  9:06       ` [RFC PATCH v4 0/9] printk: new ringbuffer implementation Peter Zijlstra
2019-09-06 10:09         ` Sergey Senozhatsky
2019-09-06 10:49           ` Peter Zijlstra
2019-09-06 13:44             ` Sergey Senozhatsky
2019-09-06 12:42         ` Petr Mladek
2019-09-06 14:01           ` Peter Zijlstra
2019-09-06 14:22             ` Peter Zijlstra
2019-09-06 19:53             ` Sergey Senozhatsky
2019-09-06 22:47             ` John Ogness
2019-09-08 22:18             ` Peter Zijlstra
2019-09-10  3:22             ` Sergey Senozhatsky

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=87pnkpjtgp.fsf@linutronix.de \
    --to=john.ogness@linutronix.de \
    --cc=andrea.parri@amarulasolutions.com \
    --cc=brendanhiggins@google.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=parri.andrea@gmail.com \
    --cc=peterz@infradead.org \
    --cc=pmladek@suse.com \
    --cc=rostedt@goodmis.org \
    --cc=sergey.senozhatsky.work@gmail.com \
    --cc=sergey.senozhatsky@gmail.com \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.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.