Re: Regression in perf bench numa convergence stats

[Ingo Molnar <mingo@kernel.org>]

* Srikar Dronamraju <srikar@linux.vnet.ibm.com> wrote:

> perf bench numa mem with -c / -m options on v4.1 and latest tip arent
> showing correct convergence statistics. I ran git bisect between v4.0 and
> v4.1. I have included the patch that fixed the problem for me.

> From 88199ad8a3d6495080eaa016b87a612bc742b1c4 Mon Sep 17 00:00:00 2001
> From: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
> Date: Wed, 24 Jun 2015 16:23:22 +0530
> Subject: [PATCH] perf tools:Fix perf_bench to show proper convergence
> 
> With commit: e1e455f (perf tools: Work around lack of sched_getcpu in
> glibc < 2.6), perf_bench numa mem with -c or -m option is not able to
> correctly calculate convergence. With the above commit, sched_getcpu
> always seems to return -1. The intention of commit e1e455f was to add a
> sched_getcpu in glibc < 2.6. Hence keep the sched_getcpu definition
> under an ifdef.
> 
> This regression happened occurred between v4.0 and v4.1
> 
> Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
> ---
>  tools/perf/util/cloexec.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/tools/perf/util/cloexec.c b/tools/perf/util/cloexec.c
> index 85b5238..2babdda 100644
> --- a/tools/perf/util/cloexec.c
> +++ b/tools/perf/util/cloexec.c
> @@ -7,11 +7,15 @@
>  
>  static unsigned long flag = PERF_FLAG_FD_CLOEXEC;
>  
> +#ifdef __GLIBC_PREREQ
> +#if !__GLIBC_PREREQ(2, 6)
>  int __weak sched_getcpu(void)
>  {
>  	errno = ENOSYS;
>  	return -1;
>  }
> +#endif
> +#endif
>  

Thanks Srikar!

Acked-by: Ingo Molnar <mingo@kernel.org>

	Ingo