1. Patchew
  2. linux
  3. [PATCH v4 0/4] Add support for AArch64 AMUv1-based arch_freq_get_on_cpu
  4. View patch

[PATCH v4 3/4] arm64: Update AMU-based frequency scale factor on entering idle

Beata Michalska posted 4 patches 1 year, 10 months ago
There is a newer version of this series
[PATCH v4 3/4] arm64: Update AMU-based frequency scale factor on entering idle
Posted by Beata Michalska 1 year, 10 months ago 
Now that the frequency scale factor has been activated for retrieving
current frequency on a given CPU, trigger its update upon entering
idle. This will, to an extent, allow querying last known frequency
in a non-invasive way. It will also improve the frequency scale factor
accuracy when a CPU entering idle did not receive a tick for a while.
As a consequence, for idle cores, the reported frequency will be the
last one observed before entering the idle state.

Suggested-by: Vanshidhar Konda <vanshikonda@os.amperecomputing.com>
Signed-off-by: Beata Michalska <beata.michalska@arm.com>
---
 arch/arm64/kernel/topology.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
index b03fe8617721..f204f6489f98 100644
--- a/arch/arm64/kernel/topology.c
+++ b/arch/arm64/kernel/topology.c
@@ -207,6 +207,19 @@ static struct scale_freq_data amu_sfd = {
 	.set_freq_scale = amu_scale_freq_tick,
 };
 
+void arch_cpu_idle_enter(void)
+{
+	unsigned int cpu = smp_processor_id();
+
+	if (!cpumask_test_cpu(cpu, amu_fie_cpus))
+		return;
+
+	/* Kick in AMU update but only if one has not happened already */
+	if (housekeeping_cpu(cpu, HK_TYPE_TICK) &&
+	    time_is_before_jiffies(per_cpu(cpu_amu_samples.last_update, cpu)))
+		amu_scale_freq_tick();
+}
+
 #define AMU_SAMPLE_EXP_MS	20
 
 unsigned int arch_freq_get_on_cpu(int cpu)
@@ -232,8 +245,8 @@ unsigned int arch_freq_get_on_cpu(int cpu)
 	 * this boils down to identifying an active cpu within the same freq
 	 * domain, if any.
 	 */
-	if (!housekeeping_cpu(cpu, HK_TYPE_TICK) ||
-	    time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS))) {
+	if (!housekeeping_cpu(cpu, HK_TYPE_TICK) || (!idle_cpu(cpu) &&
+	    time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS)))) {
 		struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
 		int ref_cpu = cpu;
 
-- 
2.25.1
Re: [PATCH v4 3/4] arm64: Update AMU-based frequency scale factor on entering idle
Posted by Sumit Gupta 1 year, 10 months ago 
Hi Beata,

This patch is giving abort due to 'amu_fie_cpus' mask which gets 
allocated later in 'init_amu_fie()'.

  ] smp: Bringing up secondary CPUs ...
  ] Unable to handle kernel read from unreadable memory at virtual
    address 0000000000000000
    .......
  ] Call trace:
  ]  arch_cpu_idle_enter+0x30/0xe0
  ]  do_idle+0xb8/0x2e0
  ]  cpu_startup_entry+0x3c/0x50
  ]  rest_init+0x108/0x128
  ]  start_kernel+0x7a4/0xa50
  ]  __primary_switched+0x80/0x90
  ] Code: d53cd042 b8626800 f943c821 53067c02 (f8627821)
  ] ---[ end trace 0000000000000000 ]---
  ] Kernel panic - not syncing: Oops: Fatal exception

Added cpumask_available() check before access to fix.

  +++ b/arch/arm64/kernel/topology.c
  @@ -211,9 +211,13 @@ void arch_cpu_idle_enter(void)
   {
          unsigned int cpu = smp_processor_id();

  -       if (!cpumask_test_cpu(cpu, amu_fie_cpus))
  +       if (cpumask_available(amu_fie_cpus) &&
  +           !cpumask_test_cpu(cpu, amu_fie_cpus))
                  return;

Thank you,
Sumit Gupta

On 05/04/24 19:03, Beata Michalska wrote:
> External email: Use caution opening links or attachments
> 
> 
> Now that the frequency scale factor has been activated for retrieving
> current frequency on a given CPU, trigger its update upon entering
> idle. This will, to an extent, allow querying last known frequency
> in a non-invasive way. It will also improve the frequency scale factor
> accuracy when a CPU entering idle did not receive a tick for a while.
> As a consequence, for idle cores, the reported frequency will be the
> last one observed before entering the idle state.
> 
> Suggested-by: Vanshidhar Konda <vanshikonda@os.amperecomputing.com>
> Signed-off-by: Beata Michalska <beata.michalska@arm.com>
> ---
>   arch/arm64/kernel/topology.c | 17 +++++++++++++++--
>   1 file changed, 15 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
> index b03fe8617721..f204f6489f98 100644
> --- a/arch/arm64/kernel/topology.c
> +++ b/arch/arm64/kernel/topology.c
> @@ -207,6 +207,19 @@ static struct scale_freq_data amu_sfd = {
>          .set_freq_scale = amu_scale_freq_tick,
>   };
> 
> +void arch_cpu_idle_enter(void)
> +{
> +       unsigned int cpu = smp_processor_id();
> +
> +       if (!cpumask_test_cpu(cpu, amu_fie_cpus))
> +               return;
> +
> +       /* Kick in AMU update but only if one has not happened already */
> +       if (housekeeping_cpu(cpu, HK_TYPE_TICK) &&
> +           time_is_before_jiffies(per_cpu(cpu_amu_samples.last_update, cpu)))
> +               amu_scale_freq_tick();
> +}
> +
>   #define AMU_SAMPLE_EXP_MS      20
> 
>   unsigned int arch_freq_get_on_cpu(int cpu)
> @@ -232,8 +245,8 @@ unsigned int arch_freq_get_on_cpu(int cpu)
>           * this boils down to identifying an active cpu within the same freq
>           * domain, if any.
>           */
> -       if (!housekeeping_cpu(cpu, HK_TYPE_TICK) ||
> -           time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS))) {
> +       if (!housekeeping_cpu(cpu, HK_TYPE_TICK) || (!idle_cpu(cpu) &&
> +           time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS)))) {
>                  struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
>                  int ref_cpu = cpu;
> 
> --
> 2.25.1
>
Re: [PATCH v4 3/4] arm64: Update AMU-based frequency scale factor on entering idle
Posted by Beata Michalska 1 year, 10 months ago 
On Thu, Apr 11, 2024 at 12:27:00AM +0530, Sumit Gupta wrote:

Hi Sumit,
> Hi Beata,
> 
> This patch is giving abort due to 'amu_fie_cpus' mask which gets allocated
> later in 'init_amu_fie()'.
> 
>  ] smp: Bringing up secondary CPUs ...
>  ] Unable to handle kernel read from unreadable memory at virtual
>    address 0000000000000000
>    .......
>  ] Call trace:
>  ]  arch_cpu_idle_enter+0x30/0xe0
>  ]  do_idle+0xb8/0x2e0
>  ]  cpu_startup_entry+0x3c/0x50
>  ]  rest_init+0x108/0x128
>  ]  start_kernel+0x7a4/0xa50
>  ]  __primary_switched+0x80/0x90
>  ] Code: d53cd042 b8626800 f943c821 53067c02 (f8627821)
>  ] ---[ end trace 0000000000000000 ]---
>  ] Kernel panic - not syncing: Oops: Fatal exception
> 
> Added cpumask_available() check before access to fix.
> 
>  +++ b/arch/arm64/kernel/topology.c
>  @@ -211,9 +211,13 @@ void arch_cpu_idle_enter(void)
>   {
>          unsigned int cpu = smp_processor_id();
> 
>  -       if (!cpumask_test_cpu(cpu, amu_fie_cpus))
>  +       if (cpumask_available(amu_fie_cpus) &&
>  +           !cpumask_test_cpu(cpu, amu_fie_cpus))
>                  return;
> 
That's actually a gruesome mistake on my side. Thanks for catching that one.
On that note: arch_freq_get_on_cpu should be also protected case amu fie init
fails for some reason. Will be sending an update.

Thanks again.

---
BR
Beata
> Thank you,
> Sumit Gupta
> 
> On 05/04/24 19:03, Beata Michalska wrote:
> > External email: Use caution opening links or attachments
> > 
> > 
> > Now that the frequency scale factor has been activated for retrieving
> > current frequency on a given CPU, trigger its update upon entering
> > idle. This will, to an extent, allow querying last known frequency
> > in a non-invasive way. It will also improve the frequency scale factor
> > accuracy when a CPU entering idle did not receive a tick for a while.
> > As a consequence, for idle cores, the reported frequency will be the
> > last one observed before entering the idle state.
> > 
> > Suggested-by: Vanshidhar Konda <vanshikonda@os.amperecomputing.com>
> > Signed-off-by: Beata Michalska <beata.michalska@arm.com>
> > ---
> >   arch/arm64/kernel/topology.c | 17 +++++++++++++++--
> >   1 file changed, 15 insertions(+), 2 deletions(-)
> > 
> > diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
> > index b03fe8617721..f204f6489f98 100644
> > --- a/arch/arm64/kernel/topology.c
> > +++ b/arch/arm64/kernel/topology.c
> > @@ -207,6 +207,19 @@ static struct scale_freq_data amu_sfd = {
> >          .set_freq_scale = amu_scale_freq_tick,
> >   };
> > 
> > +void arch_cpu_idle_enter(void)
> > +{
> > +       unsigned int cpu = smp_processor_id();
> > +
> > +       if (!cpumask_test_cpu(cpu, amu_fie_cpus))
> > +               return;
> > +
> > +       /* Kick in AMU update but only if one has not happened already */
> > +       if (housekeeping_cpu(cpu, HK_TYPE_TICK) &&
> > +           time_is_before_jiffies(per_cpu(cpu_amu_samples.last_update, cpu)))
> > +               amu_scale_freq_tick();
> > +}
> > +
> >   #define AMU_SAMPLE_EXP_MS      20
> > 
> >   unsigned int arch_freq_get_on_cpu(int cpu)
> > @@ -232,8 +245,8 @@ unsigned int arch_freq_get_on_cpu(int cpu)
> >           * this boils down to identifying an active cpu within the same freq
> >           * domain, if any.
> >           */
> > -       if (!housekeeping_cpu(cpu, HK_TYPE_TICK) ||
> > -           time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS))) {
> > +       if (!housekeeping_cpu(cpu, HK_TYPE_TICK) || (!idle_cpu(cpu) &&
> > +           time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS)))) {
> >                  struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
> >                  int ref_cpu = cpu;
> > 
> > --
> > 2.25.1
> >

Patchew 2.1-devel-b67e4c2

© 2016 - 2026 Red Hat, Inc.