From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
To: Christoph Lameter <cl@linux.com>
Cc: akpm@linux-foundation.org, Pekka Enberg <penberg@cs.helsinki.fi>,
linux-kernel@vger.kernel.org,
Eric Dumazet <eric.dumazet@gmail.com>, Tejun Heo <tj@kernel.org>
Subject: Re: [thisops uV2 04/10] x86: Support for this_cpu_add,sub,dec,inc_return
Date: Mon, 29 Nov 2010 14:22:38 -0500 [thread overview]
Message-ID: <20101129192238.GD25610@Krystal> (raw)
In-Reply-To: <alpine.DEB.2.00.1011291244410.4836@router.home>
* Christoph Lameter (cl@linux.com) wrote:
> On Mon, 29 Nov 2010, Mathieu Desnoyers wrote:
>
> > > > > +#define percpu_add_return_op(var, val) \
> > > > > +({ \
> > > > > + typedef typeof(var) pao_T__; \
> > > > > + typeof(var) pfo_ret__ = val; \
> > > > > + if (0) { \
> > > > > + pao_T__ pao_tmp__; \
> > > > > + pao_tmp__ = (val); \
> > > > > + (void)pao_tmp__; \
> > > > > + } \
> > > >
> > > > OK, I'm dumb: why is the above needed ?
> > >
> > > Ensure that the compiler agrees that *var and val are compatible. Taken
> > > over from percpu_add_op().
> >
> > Isn't that the purpose of __builtin_types_compatible_p(t1, t2) ?
>
> We also have a __same_type() macro in linux/compiler.h... But if I use
> that the kernel build fails. I guess the check is too strict.
Ah, ok, so you don't mind if var and val have different types. You just want
the compiler to be able to cast one into the other without complaining.
Isn't it already checked by "typeof(var) pfo_ret__ = val;" ?
Which semantic do you expect when using different types for var and val ?
The assembly all acts on the following pfo_ret__ :
typeof(var) pfo_ret__ = val;
But at the end, you return :
pfo_ret__ + (val);
So if pfo_ret has a type smaller than val (e.g. unsigned short vs unsigned int),
unless I'm mistakened, the type returned by percpu_add_return_op will be an
unsigned int (the larger of the two), but the atomic operation is performed on
an unsigned short. This might cause confusion for the caller. Casting
pfo_ret__ + (typeof(var)) (val);
might be appropriate.
Thanks,
Mathieu
>
> ---
> arch/x86/include/asm/percpu.h | 6 +-----
> 1 file changed, 1 insertion(+), 5 deletions(-)
>
> Index: linux-2.6/arch/x86/include/asm/percpu.h
> ===================================================================
> --- linux-2.6.orig/arch/x86/include/asm/percpu.h 2010-11-29 12:46:50.000000000 -0600
> +++ linux-2.6/arch/x86/include/asm/percpu.h 2010-11-29 12:52:25.000000000 -0600
> @@ -127,11 +127,7 @@ do { \
> typedef typeof(var) pao_T__; \
> const int pao_ID__ = (__builtin_constant_p(val) && \
> ((val) == 1 || (val) == -1)) ? (val) : 0; \
> - if (0) { \
> - pao_T__ pao_tmp__; \
> - pao_tmp__ = (val); \
> - (void)pao_tmp__; \
> - } \
> + BUILD_BUG_ON(!__same_type(var, val)); \
> switch (sizeof(var)) { \
> case 1: \
> if (pao_ID__ == 1) \
>
>
--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
next prev parent reply other threads:[~2010-11-29 19:22 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-11-26 21:09 [thisops uV2 00/10] Upgrade of this_cpu_ops V2 Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 01/10] percpucounter: Optimize __percpu_counter_add a bit through the use of this_cpu() options Christoph Lameter
2010-11-27 14:42 ` Mathieu Desnoyers
2010-11-26 21:09 ` [thisops uV2 02/10] vmstat: Optimize zone counter modifications through the use of this cpu operations Christoph Lameter
2010-11-27 8:00 ` Pekka Enberg
2010-11-27 14:49 ` Mathieu Desnoyers
2010-11-29 16:16 ` Christoph Lameter
2010-11-29 17:13 ` Christoph Lameter
2010-11-29 19:28 ` Mathieu Desnoyers
2010-11-29 20:07 ` Christoph Lameter
2010-11-29 20:59 ` Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 03/10] percpu: Generic support for this_cpu_add,sub,dec,inc_return Christoph Lameter
2010-11-27 14:58 ` Mathieu Desnoyers
2010-11-29 16:08 ` Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 04/10] x86: Support " Christoph Lameter
2010-11-27 8:06 ` Pekka Enberg
2010-11-29 16:03 ` Christoph Lameter
2010-11-27 15:00 ` Mathieu Desnoyers
2010-11-29 16:31 ` Christoph Lameter
2010-11-29 18:33 ` Mathieu Desnoyers
2010-11-29 18:54 ` Christoph Lameter
2010-11-29 19:22 ` Mathieu Desnoyers [this message]
2010-11-29 20:09 ` Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 05/10] x86: Use this_cpu_inc_return for nmi counter Christoph Lameter
2010-11-27 8:07 ` Pekka Enberg
2010-11-27 15:00 ` Mathieu Desnoyers
2010-11-26 21:09 ` [thisops uV2 06/10] vmstat: Use this_cpu_inc_return for vm statistics Christoph Lameter
2010-11-27 8:09 ` Pekka Enberg
2010-11-29 16:04 ` Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 07/10] highmem: Use this_cpu_xx_return() operations Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 08/10] percpu: generic this_cpu_cmpxchg() and this_cpu_cmpxchg_double support Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 09/10] x86: this_cpu_cmpxchg and this_cpu_cmpxchg_double operations Christoph Lameter
2010-11-27 6:30 ` Eric Dumazet
2010-11-27 15:20 ` Mathieu Desnoyers
2010-11-29 16:11 ` Christoph Lameter
2010-11-26 21:09 ` [thisops uV2 10/10] slub: Lockless fastpaths Christoph Lameter
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20101129192238.GD25610@Krystal \
--to=mathieu.desnoyers@efficios.com \
--cc=akpm@linux-foundation.org \
--cc=cl@linux.com \
--cc=eric.dumazet@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=penberg@cs.helsinki.fi \
--cc=tj@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox