linux-kernel - Re: [PATCH -v10 0/4] Lock-less list

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20110120055523.GA27974@Krystal>
Date:	Thu, 20 Jan 2011 00:55:23 -0500
From:	Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
To:	Huang Ying <ying.huang@...el.com>
Cc:	linux-kernel@...r.kernel.org, Andi Kleen <andi@...stfloor.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Ingo Molnar <mingo@...e.hu>,
	Chris Mason <chris.mason@...cle.com>,
	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
Subject: Re: [PATCH -v10 0/4] Lock-less list

Hi Huang,

I just found out about your lockless linked-list implementation, it looks
interesting. Paul McKenney and myself have created a similar implementation for
userspace within the Userspace RCU project, you might want to have a look (it's
LGPLv2.1).

One point about semantic: a singly-linked list for which you can either delete
the first element or all the elements should probably be called a stack ? I'm
therefore going to refer to "push/pop" in this email rather than "add/delete".

We've got two linked-list stack implementations in the userspace RCU tree:

One has wait-free push, and blocking pop. Useful if you can afford to block when
you pop from the list. There is no restriction on the number of concurrent
push/pop to/from the list. The code is at:

http://git.lttng.org/?p=userspace-rcu.git;a=blob;f=urcu/wfstack-static.h

The other has lock-free push and pop, no restriction on the number of concurrent
push/pop, but uses a clever trick that ensures that the cmpxchg loop will never
see a re-used pointer by using a RCU read-side to protect from memory reclaim.
It therefore requires that the memory used for the list must only be freed after
a RCU grace period. The code is at:

http://git.lttng.org/?p=userspace-rcu.git;a=blob;f=urcu/rculfstack-static.h

We could probably extend this to allow popping all stack entries in one go, but
I haven't given it much thought. 

We also have queue implementations (enqueue at head, dequeue at tail). The first
has wait-free enqueue/blocking dequeue and the second has lock-free
enqueue/dequeue. We use the wait-free enqueue/blocking dequeue queue to push
RCU callbacks when call_rcu() is executed from real-time threads, and we then
dequeue the callbacks to execute with a blocking thread. The lock-free
enqueue/dequeue queue also needs a reference count on the "dummy" node it keeps
in addition to use RCU to delay memory reclamation (see comments in the code).

wait-free enqueue/blocking dequeue:

http://git.lttng.org/?p=userspace-rcu.git;a=blob;f=urcu/wfqueue-static.h

lock-free enqueue/dequeue (beware, the refcounting makes the API less elegant
than the other implementations):

http://git.lttng.org/?p=userspace-rcu.git;a=blob;f=urcu/rculfqueue-static.h

Hoping this might be helpful to you,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/