From mboxrd@z Thu Jan  1 00:00:00 1970
From: pbonzini@redhat.com (Paolo Bonzini)
Date: Sat, 20 Feb 2010 01:10:45 +0100
Subject: [ltt-dev] lock-free data structures (was Re: [PATCH 12/12]
 centralize definition of BITS_PER_LONG)
In-Reply-To: <20100219235139.GF6778@linux.vnet.ibm.com>
References: <1266260686-17588-1-git-send-email-pbonzini@redhat.com>	<1266260686-17588-13-git-send-email-pbonzini@redhat.com>	<20100218032554.GG11338@Krystal>
	<4B7CFE10.8070803@redhat.com>	<20100218134600.GA5825@Krystal>
	<4B7D5C45.9020804@redhat.com>	<20100218160934.GA17107@Krystal>
	<4B7EE590.2010807@redhat.com>
	<20100219235139.GF6778@linux.vnet.ibm.com>
Message-ID: <4B7F2885.50505@redhat.com>

On 02/20/2010 12:51 AM, Paul E. McKenney wrote:
> Doesn't the use of RCU prevent the ABA scenario, and doesn't that make
> DCAS unnecessary?

Yeah, looks like it's so!  So by wrapping enqueue/dequeue with RCU you 
could strip the generation counts (giving the algorithm that is usually 
taught for garbage-collected environments only)... nice, I guess DCAS is 
not so much needed in the context of liburcu.

I'll try to write an example.

Paolo