From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: [PATCH v2 tip/core/rcu 07/13] ipv6/ip6_tunnel: Apply
 rcu_access_pointer() to avoid sparse false positive
Date: Wed, 09 Oct 2013 16:17:55 -0700
Message-ID: <1381360675.4971.45.camel@edumazet-glaptop.roam.corp.google.com>
References: <20131009212920.GA15413@linux.vnet.ibm.com>
	 <1381354186-16285-1-git-send-email-paulmck@linux.vnet.ibm.com>
	 <1381354186-16285-7-git-send-email-paulmck@linux.vnet.ibm.com>
	 <1381354949.4971.20.camel@edumazet-glaptop.roam.corp.google.com>
	 <20131009215747.GA5790@linux.vnet.ibm.com>
	 <1381356624.4971.26.camel@edumazet-glaptop.roam.corp.google.com>
	 <20131009223652.GC5790@linux.vnet.ibm.com>
	 <1381359077.4971.37.camel@edumazet-glaptop.roam.corp.google.com>
	 <20131009225617.GH11709@jtriplet-mobl1>
Mime-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
Cc: paulmck@linux.vnet.ibm.com, linux-kernel@vger.kernel.org,
	mingo@kernel.org, laijs@cn.fujitsu.com, dipankar@in.ibm.com,
	akpm@linux-foundation.org, mathieu.desnoyers@efficios.com,
	niv@us.ibm.com, tglx@linutronix.de, peterz@infradead.org,
	rostedt@goodmis.org, dhowells@redhat.com, edumazet@google.com,
	darren@dvhart.com, fweisbec@gmail.com, sbw@mit.edu,
	"David S. Miller" <davem@davemloft.net>,
	Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>,
	James Morris <jmorris@namei.org>,
	Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>,
	Patrick McHardy <kaber@trash.net>, netdev@vger.kernel.org
To: Josh Triplett <josh@joshtriplett.org>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20131009225617.GH11709@jtriplet-mobl1>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: netdev.vger.kernel.org

On Wed, 2013-10-09 at 15:56 -0700, Josh Triplett wrote:

> I'd be really hesitant to introduce that type of direct assignment to an
> __rcu pointer without wrapping it in some appropriately named macro, or
> at the very least adding a comment.

Well, there is no special magic here, in this specific case :

- deleting an item in an rcu list

Check list_del_rcu(), and you'll notice there is no _barrier_

Adding correct barriers is good, but please do not add them when not
needed.

It makes code hard to understand.


ACCESS(*ptr) = value;

is clear and autodocumented, because it highlights the potential
problem, that is *ptr can be read without any barrier from another cpu.
So we ask the compiler to not write temporary garbage in it.

rcu_assign_pointer(*ptr, rcu_access_pointer(value)) 

is very confusing, because it hides the _real_ problem and add defensive
programming tricks.