linux-kernel - Re: [PATCH v2] futex: replace bare barrier() with a READ

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] [day] [month] [year] [list]

Message-ID: <20160302164821.GB1092@dvhart-mobl5.amr.corp.intel.com>
Date:	Wed, 2 Mar 2016 08:48:21 -0800
From:	Darren Hart <dvhart@...radead.org>
To:	Jianyu Zhan <nasa4836@...il.com>
Cc:	linux-kernel@...r.kernel.org, paulmck@...ux.vnet.ibm.com,
	peterz@...radead.org, tglx@...utronix.de, dave@...olabs.net,
	akpm@...ux-foundation.org, mingo@...nel.org,
	linux@...musvillemoes.dk, dvhart@...ux.intel.com,
	bigeasy@...utronix.de, schwidefsky@...ibm.com,
	heiko.carstens@...ibm.com
Subject: Re: [PATCH v2] futex:  replace bare barrier() with a READ_ONCE()

On Wed, Mar 02, 2016 at 09:08:31PM +0800, Jianyu Zhan wrote:

...

> 3) READ_ONCE() _might_ prevent more _theoretical_ "optimizations" by the compiler:
> 
>    The above retry logic is effectively the same as:
> 
> 	while (lock_ptr = READ_ONCE(q->lock_ptr)) {
>                    spin_lock(lock_ptr)


The spin_lock() is memory barrier, and therefor a general compiler barrier. The
READ_ONCE would be redundant in this case.

Unless you can demonstrate a failure mode in disassembly, or can point out how
the spin_lock barrier is insufficient that I have missed, this third point is
already covered.

-- 
Darren Hart
Intel Open Source Technology Center