Date: Wed, 24 Nov 2010 10:55:40 -0600 (CST)
From: Christoph Lameter <cl@linux.com>
To: Pekka Enberg <penberg@kernel.org>
cc: akpm@linux-foundation.org, Pekka Enberg <penberg@cs.helsinki.fi>,
        Ingo Molnar <mingo@elte.hu>, Peter Zijlstra <peterz@infradead.org>,
        linux-kernel@vger.kernel.org, Eric Dumazet <eric.dumazet@gmail.com>,
        Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
        Tejun Heo <tj@kernel.org>
Subject: Re: [thiscpuops upgrade 10/10] Lockless (and preemptless) fastpaths
 for slub
In-Reply-To: <AANLkTin02TYMurwmGZi2meSmVg6W4a37qmOLJLrFvMCm@mail.gmail.com>
Message-ID: <alpine.DEB.2.00.1011241051480.16025@router.home>
References: <20101123235139.908255844@linux.com> <20101123235201.758191189@linux.com> <AANLkTikVFWYS9UeCP1DJxhzYn9_rMcrovSfbQpNmXsvk@mail.gmail.com> <alpine.DEB.2.00.1011241012180.14268@router.home> <AANLkTikxrvbDJ=rDhj=gC-FJTa1Jeky1uLWjnP6VF_pX@mail.gmail.com>
 <alpine.DEB.2.00.1011241042100.16025@router.home> <AANLkTin02TYMurwmGZi2meSmVg6W4a37qmOLJLrFvMCm@mail.gmail.com>
User-Agent: Alpine 2.00 (DEB 1167 2008-08-23)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1343
Lines: 28

On Wed, 24 Nov 2010, Pekka Enberg wrote:

> On Wed, 24 Nov 2010, Pekka Enberg wrote:
> >> > The critical section begins with the retrieval of the tid and it ends with
> >> > the replacement of the tid with the newly generated one. This means that
> >> > all state data for the alloc and free operation needs to be retrieved in
> >> > that critical section. The change must be saved with the final
> >> > cmpxchg_double of the critical section.
> >>
> >> Right and we don't need a *memory barrier* here because we're
> >> accessing a per-CPU variable which means operations appear in-order.
>
> On Wed, Nov 24, 2010 at 6:45 PM, Christoph Lameter <cl@linux.com> wrote:
> > The compiler is still free to rearrange the tid fetch. A possible
> > optimization that the compiler may do is to move the tid fetch into the
> > next if statement since that is the only block in which the tid variable
> > is actually used.
>
> Yes, which is why we need a *compiler barrier* but not a *memory barrier*.

Exactly. That is the reason there is a compiler barrier there. A memory
barrier would be smp_mb() or so.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/