linux-kernel - Re: linux-next: build warnings after merge of the access

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: <20150326144153.GE2805@arm.com>
Date:	Thu, 26 Mar 2015 14:41:54 +0000
From:	Will Deacon <will.deacon@....com>
To:	Peter Zijlstra <peterz@...radead.org>
Cc:	Stephen Rothwell <sfr@...b.auug.org.au>,
	Christian Borntraeger <borntraeger@...ibm.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Ingo Molnar <mingo@...e.hu>, "H. Peter Anvin" <hpa@...or.com>,
	"linux-next@...r.kernel.org" <linux-next@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Davidlohr Bueso <dave@...olabs.net>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Paul McKenney <paulmck@...ux.vnet.ibm.com>
Subject: Re: linux-next: build warnings after merge of the access_once tree

On Thu, Mar 26, 2015 at 02:22:20PM +0000, Peter Zijlstra wrote:
> On Thu, Mar 26, 2015 at 01:27:50PM +0000, Will Deacon wrote:
> > On Thu, Mar 26, 2015 at 10:34:42AM +0000, Peter Zijlstra wrote:
> > > On Thu, Mar 26, 2015 at 07:31:12PM +1100, Stephen Rothwell wrote:
> > > > In function '__read_once_size',
> > > >     inlined from 'lockref_get' at lib/lockref.c:50:2:
> 
> 
> > Yeah, I think it's fine because, as you point out, the cmpxchg can only
> > succeed if the 64-bit load appeared to be single-copy atomic (amongst other
> > things).
> 
> So one option to get rid of this warning is to rely on the fact that all
> CMPXCHG_LOOP users are at the beginning of !pure function calls, which
> already imply a compiler barrier and therefore it must already emit that
> load.
> 
> And as already argued, split loads aren't an issue because the cmpxchg
> will catch those for us.
> 
> So we can either just remove the READ_ONCE(), or replace it with a
> leading barrier() call just to be on the paranoid side of things.

If we remove the READ_ONCE then I think the barrier is a good idea, just in
case the LTO guys get their paws on this and we see subtle breakage.

> Any preferences?
> 
> Something like so, but with a sensible comment I suppose.
> 
> ---
>  lib/lockref.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/lockref.c b/lib/lockref.c
> index 494994bf17c8..b5ca1f65c8a3 100644
> --- a/lib/lockref.c
> +++ b/lib/lockref.c
> @@ -18,7 +18,8 @@
>  #define CMPXCHG_LOOP(CODE, SUCCESS) do {					\
>  	struct lockref old;							\
>  	BUILD_BUG_ON(sizeof(old) != 8);						\
> -	old.lock_count = READ_ONCE(lockref->lock_count);			\
> +	barrier();								\
> +	old.lock_count = lockref->lock_count;					\
>  	while (likely(arch_spin_value_unlocked(old.lock.rlock.raw_lock))) {  	\
>  		struct lockref new = old, prev = old;				\
>  		CODE								\

Is ACCESS_ONCE actually going away? It has its problems, but I think it's
what we want here and reads better than magic barrier() imo.

Will
--
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/