| 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
| ||
|
Message-ID: <20210217002422.GA17422@arm.com>
Date: Wed, 17 Feb 2021 00:24:22 +0000
From: Ionela Voinescu <ionela.voinescu@....com>
To: Viresh Kumar <viresh.kumar@...aro.org>
Cc: Rafael Wysocki <rjw@...ysocki.net>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Vincent Guittot <vincent.guittot@...aro.org>,
linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-pm@...r.kernel.org, Sudeep Holla <sudeep.holla@....com>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Subject: Re: [PATCH V3 1/2] topology: Allow multiple entities to provide
sched_freq_tick() callback
Hey,
On Friday 05 Feb 2021 at 14:44:24 (+0530), Viresh Kumar wrote:
> On 03-02-21, 11:45, Ionela Voinescu wrote:
> > Therefore, I think system level invariance management (checks and
> > call to rebuild_sched_domains_energy()) also needs to move from arm64
> > code to arch_topology code.
>
> Here is the 3rd patch of this series then :)
>
I think it could be merged in patch 1/2 as it's part of enabling the use
of multiple sources of information for FIE. Up to you!
> From: Viresh Kumar <viresh.kumar@...aro.org>
> Date: Fri, 5 Feb 2021 13:31:53 +0530
> Subject: [PATCH] drivers: arch_topology: rebuild sched domains on invariance
> change
>
> We already do this for the arm64, move it to arch_topology.c as we
> manage all sched_freq_tick sources here now.
>
> Reported-by: Ionela Voinescu <ionela.voinescu@....com>
> Signed-off-by: Viresh Kumar <viresh.kumar@...aro.org>
> ---
> arch/arm64/kernel/topology.c | 16 ----------------
> drivers/base/arch_topology.c | 22 ++++++++++++++++++++++
> 2 files changed, 22 insertions(+), 16 deletions(-)
>
> diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
> index 1e47dfd465f8..47fca7376c93 100644
> --- a/arch/arm64/kernel/topology.c
> +++ b/arch/arm64/kernel/topology.c
> @@ -240,7 +240,6 @@ static struct scale_freq_data amu_sfd = {
>
> static void amu_fie_setup(const struct cpumask *cpus)
> {
> - bool invariant;
> int cpu;
>
> /* We are already set since the last insmod of cpufreq driver */
> @@ -257,25 +256,10 @@ static void amu_fie_setup(const struct cpumask *cpus)
>
> cpumask_or(amu_fie_cpus, amu_fie_cpus, cpus);
>
> - invariant = topology_scale_freq_invariant();
> -
> - /* We aren't fully invariant yet */
> - if (!invariant && !cpumask_equal(amu_fie_cpus, cpu_present_mask))
> - return;
> -
You still need these checks, otherwise you could end up with only part
of the CPUs setting a scale factor, when only part of the CPUs support
AMUs and there is no cpufreq support for FIE.
> topology_set_scale_freq_source(&amu_sfd, amu_fie_cpus);
>
> pr_debug("CPUs[%*pbl]: counters will be used for FIE.",
> cpumask_pr_args(cpus));
> -
> - /*
> - * Task scheduler behavior depends on frequency invariance support,
> - * either cpufreq or counter driven. If the support status changes as
> - * a result of counter initialisation and use, retrigger the build of
> - * scheduling domains to ensure the information is propagated properly.
> - */
> - if (!invariant)
> - rebuild_sched_domains_energy();
> }
>
> static int init_amu_fie_callback(struct notifier_block *nb, unsigned long val,
> diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
> index 20b511949cd8..3631877f4440 100644
> --- a/drivers/base/arch_topology.c
> +++ b/drivers/base/arch_topology.c
> @@ -23,6 +23,7 @@
>
> static DEFINE_PER_CPU(struct scale_freq_data *, sft_data);
> static struct cpumask scale_freq_counters_mask;
> +static bool scale_freq_invariant;
>
> static bool supports_scale_freq_counters(const struct cpumask *cpus)
> {
> @@ -35,6 +36,23 @@ bool topology_scale_freq_invariant(void)
> supports_scale_freq_counters(cpu_online_mask);
> }
>
> +static void update_scale_freq_invariant(bool status)
> +{
> + if (scale_freq_invariant == status)
> + return;
> +
> + /*
> + * Task scheduler behavior depends on frequency invariance support,
> + * either cpufreq or counter driven. If the support status changes as
> + * a result of counter initialisation and use, retrigger the build of
> + * scheduling domains to ensure the information is propagated properly.
> + */
> + if (topology_scale_freq_invariant() == status) {
> + scale_freq_invariant = status;
> + rebuild_sched_domains_energy();
> + }
> +}
> +
> void topology_set_scale_freq_source(struct scale_freq_data *data,
> const struct cpumask *cpus)
> {
> @@ -50,6 +68,8 @@ void topology_set_scale_freq_source(struct scale_freq_data *data,
> cpumask_set_cpu(cpu, &scale_freq_counters_mask);
> }
> }
Small(ish) optimisation at the beginning of this function:
if (cpumask_empty(&scale_freq_counters_mask))
scale_freq_invariant = topology_scale_freq_invariant();
This will save you a call to rebuild_sched_domains_energy(), which is
quite expensive, when cpufreq supports FIE and we also have counters.
After comments addressed,
Reviewed-by: Ionela Voinescu <ionela.voinescu@....com>
Tested-by: Ionela Voinescu <ionela.voinescu@....com>
..for 1/2 and this addition.
Thanks,
Ionela.
> +
> + update_scale_freq_invariant(true);
> }
> EXPORT_SYMBOL_GPL(topology_set_scale_freq_source);
>
> @@ -67,6 +87,8 @@ void topology_clear_scale_freq_source(enum scale_freq_source source,
> cpumask_clear_cpu(cpu, &scale_freq_counters_mask);
> }
> }
> +
> + update_scale_freq_invariant(false);
> }
> EXPORT_SYMBOL_GPL(topology_clear_scale_freq_source);
>
> --
> 2.25.0.rc1.19.g042ed3e048af
>
> --
> viresh
Powered by blists - more mailing lists