All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v4] arm64: Enable perf events based hard lockup detector
@ 2020-10-07  8:51 ` Sumit Garg
  0 siblings, 0 replies; 10+ messages in thread
From: Sumit Garg @ 2020-10-07  8:51 UTC (permalink / raw)
  To: will, alexandru.elisei
  Cc: linux-arm-kernel, catalin.marinas, mark.rutland, peterz, mingo,
	acme, alexander.shishkin, jolsa, namhyung, tglx, swboyd,
	julien.thierry.kdev, dianders, daniel.thompson, linux-kernel,
	Sumit Garg

With the recent feature added to enable perf events to use pseudo NMIs
as interrupts on platforms which support GICv3 or later, its now been
possible to enable hard lockup detector (or NMI watchdog) on arm64
platforms. So enable corresponding support.

One thing to note here is that normally lockup detector is initialized
just after the early initcalls but PMU on arm64 comes up much later as
device_initcall(). So we need to re-initialize lockup detection once
PMU has been initialized.

Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
---

Changes in v4:
- Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
  hook to know if PMU IRQ has been requested as an NMI.
- Add check for return value prior to initializing hard-lockup detector.

[1] https://lkml.org/lkml/2020/9/24/458

Changes in v3:
- Rebased to latest pmu NMI patch-set [1].
- Addressed misc. comments from Stephen.

[1] https://lkml.org/lkml/2020/8/19/671

Changes since RFC:
- Rebased on top of Alex's WIP-pmu-nmi branch.
- Add comment for safe max. CPU frequency.
- Misc. cleanup.

 arch/arm64/Kconfig             |  2 ++
 arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
 drivers/perf/arm_pmu.c         |  5 +++++
 include/linux/perf/arm_pmu.h   |  2 ++
 4 files changed, 48 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 6d23283..b5c2594 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -167,6 +167,8 @@ config ARM64
 	select HAVE_NMI
 	select HAVE_PATA_PLATFORM
 	select HAVE_PERF_EVENTS
+	select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
+	select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
 	select HAVE_PERF_REGS
 	select HAVE_PERF_USER_STACK_DUMP
 	select HAVE_REGS_AND_STACK_ACCESS_API
diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
index ef206fb..6ad5120 100644
--- a/arch/arm64/kernel/perf_event.c
+++ b/arch/arm64/kernel/perf_event.c
@@ -23,6 +23,8 @@
 #include <linux/platform_device.h>
 #include <linux/sched_clock.h>
 #include <linux/smp.h>
+#include <linux/nmi.h>
+#include <linux/cpufreq.h>
 
 /* ARMv8 Cortex-A53 specific event types. */
 #define ARMV8_A53_PERFCTR_PREF_LINEFILL				0xC2
@@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
 
 static int __init armv8_pmu_driver_init(void)
 {
+	int ret;
+
 	if (acpi_disabled)
-		return platform_driver_register(&armv8_pmu_driver);
+		ret = platform_driver_register(&armv8_pmu_driver);
 	else
-		return arm_pmu_acpi_probe(armv8_pmuv3_init);
+		ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
+
+	/*
+	 * Try to re-initialize lockup detector after PMU init in
+	 * case PMU events are triggered via NMIs.
+	 */
+	if (ret == 0 && arm_pmu_irq_is_nmi())
+		lockup_detector_init();
+
+	return ret;
 }
 device_initcall(armv8_pmu_driver_init)
 
@@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
 	userpg->cap_user_time_zero = 1;
 	userpg->cap_user_time_short = 1;
 }
+
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
+/*
+ * Safe maximum CPU frequency in case a particular platform doesn't implement
+ * cpufreq driver. Although, architecture doesn't put any restrictions on
+ * maximum frequency but 5 GHz seems to be safe maximum given the available
+ * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
+ * hand, we can't make it much higher as it would lead to a large hard-lockup
+ * detection timeout on parts which are running slower (eg. 1GHz on
+ * Developerbox) and doesn't possess a cpufreq driver.
+ */
+#define SAFE_MAX_CPU_FREQ	5000000000UL // 5 GHz
+u64 hw_nmi_get_sample_period(int watchdog_thresh)
+{
+	unsigned int cpu = smp_processor_id();
+	unsigned long max_cpu_freq;
+
+	max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
+	if (!max_cpu_freq)
+		max_cpu_freq = SAFE_MAX_CPU_FREQ;
+
+	return (u64)max_cpu_freq * watchdog_thresh;
+}
+#endif
diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
index cb2f55f..794a37d 100644
--- a/drivers/perf/arm_pmu.c
+++ b/drivers/perf/arm_pmu.c
@@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
 	return per_cpu(hw_events->irq, cpu);
 }
 
+bool arm_pmu_irq_is_nmi(void)
+{
+	return has_nmi;
+}
+
 /*
  * PMU hardware loses all context when a CPU goes offline.
  * When a CPU is hotplugged back in, since some hardware registers are
diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
index 5b616dd..5765069 100644
--- a/include/linux/perf/arm_pmu.h
+++ b/include/linux/perf/arm_pmu.h
@@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
 static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
 #endif
 
+bool arm_pmu_irq_is_nmi(void);
+
 /* Internal functions only for core arm_pmu code */
 struct arm_pmu *armpmu_alloc(void);
 struct arm_pmu *armpmu_alloc_atomic(void);
-- 
2.7.4


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH v4] arm64: Enable perf events based hard lockup detector
@ 2020-10-07  8:51 ` Sumit Garg
  0 siblings, 0 replies; 10+ messages in thread
From: Sumit Garg @ 2020-10-07  8:51 UTC (permalink / raw)
  To: will, alexandru.elisei
  Cc: mark.rutland, Sumit Garg, daniel.thompson, linux-kernel, peterz,
	catalin.marinas, dianders, acme, swboyd, alexander.shishkin,
	mingo, julien.thierry.kdev, namhyung, tglx, jolsa,
	linux-arm-kernel

With the recent feature added to enable perf events to use pseudo NMIs
as interrupts on platforms which support GICv3 or later, its now been
possible to enable hard lockup detector (or NMI watchdog) on arm64
platforms. So enable corresponding support.

One thing to note here is that normally lockup detector is initialized
just after the early initcalls but PMU on arm64 comes up much later as
device_initcall(). So we need to re-initialize lockup detection once
PMU has been initialized.

Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
---

Changes in v4:
- Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
  hook to know if PMU IRQ has been requested as an NMI.
- Add check for return value prior to initializing hard-lockup detector.

[1] https://lkml.org/lkml/2020/9/24/458

Changes in v3:
- Rebased to latest pmu NMI patch-set [1].
- Addressed misc. comments from Stephen.

[1] https://lkml.org/lkml/2020/8/19/671

Changes since RFC:
- Rebased on top of Alex's WIP-pmu-nmi branch.
- Add comment for safe max. CPU frequency.
- Misc. cleanup.

 arch/arm64/Kconfig             |  2 ++
 arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
 drivers/perf/arm_pmu.c         |  5 +++++
 include/linux/perf/arm_pmu.h   |  2 ++
 4 files changed, 48 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 6d23283..b5c2594 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -167,6 +167,8 @@ config ARM64
 	select HAVE_NMI
 	select HAVE_PATA_PLATFORM
 	select HAVE_PERF_EVENTS
+	select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
+	select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
 	select HAVE_PERF_REGS
 	select HAVE_PERF_USER_STACK_DUMP
 	select HAVE_REGS_AND_STACK_ACCESS_API
diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
index ef206fb..6ad5120 100644
--- a/arch/arm64/kernel/perf_event.c
+++ b/arch/arm64/kernel/perf_event.c
@@ -23,6 +23,8 @@
 #include <linux/platform_device.h>
 #include <linux/sched_clock.h>
 #include <linux/smp.h>
+#include <linux/nmi.h>
+#include <linux/cpufreq.h>
 
 /* ARMv8 Cortex-A53 specific event types. */
 #define ARMV8_A53_PERFCTR_PREF_LINEFILL				0xC2
@@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
 
 static int __init armv8_pmu_driver_init(void)
 {
+	int ret;
+
 	if (acpi_disabled)
-		return platform_driver_register(&armv8_pmu_driver);
+		ret = platform_driver_register(&armv8_pmu_driver);
 	else
-		return arm_pmu_acpi_probe(armv8_pmuv3_init);
+		ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
+
+	/*
+	 * Try to re-initialize lockup detector after PMU init in
+	 * case PMU events are triggered via NMIs.
+	 */
+	if (ret == 0 && arm_pmu_irq_is_nmi())
+		lockup_detector_init();
+
+	return ret;
 }
 device_initcall(armv8_pmu_driver_init)
 
@@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
 	userpg->cap_user_time_zero = 1;
 	userpg->cap_user_time_short = 1;
 }
+
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
+/*
+ * Safe maximum CPU frequency in case a particular platform doesn't implement
+ * cpufreq driver. Although, architecture doesn't put any restrictions on
+ * maximum frequency but 5 GHz seems to be safe maximum given the available
+ * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
+ * hand, we can't make it much higher as it would lead to a large hard-lockup
+ * detection timeout on parts which are running slower (eg. 1GHz on
+ * Developerbox) and doesn't possess a cpufreq driver.
+ */
+#define SAFE_MAX_CPU_FREQ	5000000000UL // 5 GHz
+u64 hw_nmi_get_sample_period(int watchdog_thresh)
+{
+	unsigned int cpu = smp_processor_id();
+	unsigned long max_cpu_freq;
+
+	max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
+	if (!max_cpu_freq)
+		max_cpu_freq = SAFE_MAX_CPU_FREQ;
+
+	return (u64)max_cpu_freq * watchdog_thresh;
+}
+#endif
diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
index cb2f55f..794a37d 100644
--- a/drivers/perf/arm_pmu.c
+++ b/drivers/perf/arm_pmu.c
@@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
 	return per_cpu(hw_events->irq, cpu);
 }
 
+bool arm_pmu_irq_is_nmi(void)
+{
+	return has_nmi;
+}
+
 /*
  * PMU hardware loses all context when a CPU goes offline.
  * When a CPU is hotplugged back in, since some hardware registers are
diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
index 5b616dd..5765069 100644
--- a/include/linux/perf/arm_pmu.h
+++ b/include/linux/perf/arm_pmu.h
@@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
 static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
 #endif
 
+bool arm_pmu_irq_is_nmi(void);
+
 /* Internal functions only for core arm_pmu code */
 struct arm_pmu *armpmu_alloc(void);
 struct arm_pmu *armpmu_alloc_atomic(void);
-- 
2.7.4


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
  2020-10-07  8:51 ` Sumit Garg
@ 2020-10-09 10:23   ` Sumit Garg
  -1 siblings, 0 replies; 10+ messages in thread
From: Sumit Garg @ 2020-10-09 10:23 UTC (permalink / raw)
  To: Will Deacon, Alexandru Elisei
  Cc: linux-arm-kernel, Catalin Marinas, Mark Rutland, Peter Zijlstra,
	mingo, acme, alexander.shishkin, jolsa, namhyung,
	Thomas Gleixner, Stephen Boyd, julien.thierry.kdev,
	Douglas Anderson, Daniel Thompson, Linux Kernel Mailing List

Hi Will, Alex,

On Wed, 7 Oct 2020 at 14:22, Sumit Garg <sumit.garg@linaro.org> wrote:
>
> With the recent feature added to enable perf events to use pseudo NMIs
> as interrupts on platforms which support GICv3 or later, its now been
> possible to enable hard lockup detector (or NMI watchdog) on arm64
> platforms. So enable corresponding support.
>
> One thing to note here is that normally lockup detector is initialized
> just after the early initcalls but PMU on arm64 comes up much later as
> device_initcall(). So we need to re-initialize lockup detection once
> PMU has been initialized.
>
> Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> ---
>

Are there any further comments on this patch? If we don't have any,
would it be possible to queue this patch for 5.10?

-Sumit

> Changes in v4:
> - Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
>   hook to know if PMU IRQ has been requested as an NMI.
> - Add check for return value prior to initializing hard-lockup detector.
>
> [1] https://lkml.org/lkml/2020/9/24/458
>
> Changes in v3:
> - Rebased to latest pmu NMI patch-set [1].
> - Addressed misc. comments from Stephen.
>
> [1] https://lkml.org/lkml/2020/8/19/671
>
> Changes since RFC:
> - Rebased on top of Alex's WIP-pmu-nmi branch.
> - Add comment for safe max. CPU frequency.
> - Misc. cleanup.
>
>  arch/arm64/Kconfig             |  2 ++
>  arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
>  drivers/perf/arm_pmu.c         |  5 +++++
>  include/linux/perf/arm_pmu.h   |  2 ++
>  4 files changed, 48 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 6d23283..b5c2594 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -167,6 +167,8 @@ config ARM64
>         select HAVE_NMI
>         select HAVE_PATA_PLATFORM
>         select HAVE_PERF_EVENTS
> +       select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
> +       select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
>         select HAVE_PERF_REGS
>         select HAVE_PERF_USER_STACK_DUMP
>         select HAVE_REGS_AND_STACK_ACCESS_API
> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
> index ef206fb..6ad5120 100644
> --- a/arch/arm64/kernel/perf_event.c
> +++ b/arch/arm64/kernel/perf_event.c
> @@ -23,6 +23,8 @@
>  #include <linux/platform_device.h>
>  #include <linux/sched_clock.h>
>  #include <linux/smp.h>
> +#include <linux/nmi.h>
> +#include <linux/cpufreq.h>
>
>  /* ARMv8 Cortex-A53 specific event types. */
>  #define ARMV8_A53_PERFCTR_PREF_LINEFILL                                0xC2
> @@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
>
>  static int __init armv8_pmu_driver_init(void)
>  {
> +       int ret;
> +
>         if (acpi_disabled)
> -               return platform_driver_register(&armv8_pmu_driver);
> +               ret = platform_driver_register(&armv8_pmu_driver);
>         else
> -               return arm_pmu_acpi_probe(armv8_pmuv3_init);
> +               ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
> +
> +       /*
> +        * Try to re-initialize lockup detector after PMU init in
> +        * case PMU events are triggered via NMIs.
> +        */
> +       if (ret == 0 && arm_pmu_irq_is_nmi())
> +               lockup_detector_init();
> +
> +       return ret;
>  }
>  device_initcall(armv8_pmu_driver_init)
>
> @@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
>         userpg->cap_user_time_zero = 1;
>         userpg->cap_user_time_short = 1;
>  }
> +
> +#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
> +/*
> + * Safe maximum CPU frequency in case a particular platform doesn't implement
> + * cpufreq driver. Although, architecture doesn't put any restrictions on
> + * maximum frequency but 5 GHz seems to be safe maximum given the available
> + * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
> + * hand, we can't make it much higher as it would lead to a large hard-lockup
> + * detection timeout on parts which are running slower (eg. 1GHz on
> + * Developerbox) and doesn't possess a cpufreq driver.
> + */
> +#define SAFE_MAX_CPU_FREQ      5000000000UL // 5 GHz
> +u64 hw_nmi_get_sample_period(int watchdog_thresh)
> +{
> +       unsigned int cpu = smp_processor_id();
> +       unsigned long max_cpu_freq;
> +
> +       max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
> +       if (!max_cpu_freq)
> +               max_cpu_freq = SAFE_MAX_CPU_FREQ;
> +
> +       return (u64)max_cpu_freq * watchdog_thresh;
> +}
> +#endif
> diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> index cb2f55f..794a37d 100644
> --- a/drivers/perf/arm_pmu.c
> +++ b/drivers/perf/arm_pmu.c
> @@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
>         return per_cpu(hw_events->irq, cpu);
>  }
>
> +bool arm_pmu_irq_is_nmi(void)
> +{
> +       return has_nmi;
> +}
> +
>  /*
>   * PMU hardware loses all context when a CPU goes offline.
>   * When a CPU is hotplugged back in, since some hardware registers are
> diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
> index 5b616dd..5765069 100644
> --- a/include/linux/perf/arm_pmu.h
> +++ b/include/linux/perf/arm_pmu.h
> @@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
>  static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
>  #endif
>
> +bool arm_pmu_irq_is_nmi(void);
> +
>  /* Internal functions only for core arm_pmu code */
>  struct arm_pmu *armpmu_alloc(void);
>  struct arm_pmu *armpmu_alloc_atomic(void);
> --
> 2.7.4
>

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
@ 2020-10-09 10:23   ` Sumit Garg
  0 siblings, 0 replies; 10+ messages in thread
From: Sumit Garg @ 2020-10-09 10:23 UTC (permalink / raw)
  To: Will Deacon, Alexandru Elisei
  Cc: Mark Rutland, Daniel Thompson, Linux Kernel Mailing List,
	Peter Zijlstra, Catalin Marinas, Douglas Anderson, acme,
	Stephen Boyd, alexander.shishkin, mingo, julien.thierry.kdev,
	namhyung, Thomas Gleixner, jolsa, linux-arm-kernel

Hi Will, Alex,

On Wed, 7 Oct 2020 at 14:22, Sumit Garg <sumit.garg@linaro.org> wrote:
>
> With the recent feature added to enable perf events to use pseudo NMIs
> as interrupts on platforms which support GICv3 or later, its now been
> possible to enable hard lockup detector (or NMI watchdog) on arm64
> platforms. So enable corresponding support.
>
> One thing to note here is that normally lockup detector is initialized
> just after the early initcalls but PMU on arm64 comes up much later as
> device_initcall(). So we need to re-initialize lockup detection once
> PMU has been initialized.
>
> Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> ---
>

Are there any further comments on this patch? If we don't have any,
would it be possible to queue this patch for 5.10?

-Sumit

> Changes in v4:
> - Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
>   hook to know if PMU IRQ has been requested as an NMI.
> - Add check for return value prior to initializing hard-lockup detector.
>
> [1] https://lkml.org/lkml/2020/9/24/458
>
> Changes in v3:
> - Rebased to latest pmu NMI patch-set [1].
> - Addressed misc. comments from Stephen.
>
> [1] https://lkml.org/lkml/2020/8/19/671
>
> Changes since RFC:
> - Rebased on top of Alex's WIP-pmu-nmi branch.
> - Add comment for safe max. CPU frequency.
> - Misc. cleanup.
>
>  arch/arm64/Kconfig             |  2 ++
>  arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
>  drivers/perf/arm_pmu.c         |  5 +++++
>  include/linux/perf/arm_pmu.h   |  2 ++
>  4 files changed, 48 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 6d23283..b5c2594 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -167,6 +167,8 @@ config ARM64
>         select HAVE_NMI
>         select HAVE_PATA_PLATFORM
>         select HAVE_PERF_EVENTS
> +       select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
> +       select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
>         select HAVE_PERF_REGS
>         select HAVE_PERF_USER_STACK_DUMP
>         select HAVE_REGS_AND_STACK_ACCESS_API
> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
> index ef206fb..6ad5120 100644
> --- a/arch/arm64/kernel/perf_event.c
> +++ b/arch/arm64/kernel/perf_event.c
> @@ -23,6 +23,8 @@
>  #include <linux/platform_device.h>
>  #include <linux/sched_clock.h>
>  #include <linux/smp.h>
> +#include <linux/nmi.h>
> +#include <linux/cpufreq.h>
>
>  /* ARMv8 Cortex-A53 specific event types. */
>  #define ARMV8_A53_PERFCTR_PREF_LINEFILL                                0xC2
> @@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
>
>  static int __init armv8_pmu_driver_init(void)
>  {
> +       int ret;
> +
>         if (acpi_disabled)
> -               return platform_driver_register(&armv8_pmu_driver);
> +               ret = platform_driver_register(&armv8_pmu_driver);
>         else
> -               return arm_pmu_acpi_probe(armv8_pmuv3_init);
> +               ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
> +
> +       /*
> +        * Try to re-initialize lockup detector after PMU init in
> +        * case PMU events are triggered via NMIs.
> +        */
> +       if (ret == 0 && arm_pmu_irq_is_nmi())
> +               lockup_detector_init();
> +
> +       return ret;
>  }
>  device_initcall(armv8_pmu_driver_init)
>
> @@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
>         userpg->cap_user_time_zero = 1;
>         userpg->cap_user_time_short = 1;
>  }
> +
> +#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
> +/*
> + * Safe maximum CPU frequency in case a particular platform doesn't implement
> + * cpufreq driver. Although, architecture doesn't put any restrictions on
> + * maximum frequency but 5 GHz seems to be safe maximum given the available
> + * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
> + * hand, we can't make it much higher as it would lead to a large hard-lockup
> + * detection timeout on parts which are running slower (eg. 1GHz on
> + * Developerbox) and doesn't possess a cpufreq driver.
> + */
> +#define SAFE_MAX_CPU_FREQ      5000000000UL // 5 GHz
> +u64 hw_nmi_get_sample_period(int watchdog_thresh)
> +{
> +       unsigned int cpu = smp_processor_id();
> +       unsigned long max_cpu_freq;
> +
> +       max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
> +       if (!max_cpu_freq)
> +               max_cpu_freq = SAFE_MAX_CPU_FREQ;
> +
> +       return (u64)max_cpu_freq * watchdog_thresh;
> +}
> +#endif
> diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> index cb2f55f..794a37d 100644
> --- a/drivers/perf/arm_pmu.c
> +++ b/drivers/perf/arm_pmu.c
> @@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
>         return per_cpu(hw_events->irq, cpu);
>  }
>
> +bool arm_pmu_irq_is_nmi(void)
> +{
> +       return has_nmi;
> +}
> +
>  /*
>   * PMU hardware loses all context when a CPU goes offline.
>   * When a CPU is hotplugged back in, since some hardware registers are
> diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
> index 5b616dd..5765069 100644
> --- a/include/linux/perf/arm_pmu.h
> +++ b/include/linux/perf/arm_pmu.h
> @@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
>  static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
>  #endif
>
> +bool arm_pmu_irq_is_nmi(void);
> +
>  /* Internal functions only for core arm_pmu code */
>  struct arm_pmu *armpmu_alloc(void);
>  struct arm_pmu *armpmu_alloc_atomic(void);
> --
> 2.7.4
>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
  2020-10-07  8:51 ` Sumit Garg
@ 2020-10-09 10:26   ` Alexandru Elisei
  -1 siblings, 0 replies; 10+ messages in thread
From: Alexandru Elisei @ 2020-10-09 10:26 UTC (permalink / raw)
  To: Sumit Garg, will
  Cc: linux-arm-kernel, catalin.marinas, mark.rutland, peterz, mingo,
	acme, alexander.shishkin, jolsa, namhyung, tglx, swboyd,
	julien.thierry.kdev, dianders, daniel.thompson, linux-kernel

Hi,

On 10/7/20 9:51 AM, Sumit Garg wrote:
> With the recent feature added to enable perf events to use pseudo NMIs
> as interrupts on platforms which support GICv3 or later, its now been
> possible to enable hard lockup detector (or NMI watchdog) on arm64
> platforms. So enable corresponding support.
>
> One thing to note here is that normally lockup detector is initialized
> just after the early initcalls but PMU on arm64 comes up much later as
> device_initcall(). So we need to re-initialize lockup detection once
> PMU has been initialized.

Has another look, the PMU bits looks alright. Just to be on the safe side, I did a
few quick boot tests on an espressobin with a kernel with CONFIG_ARM64_PSEUDO_NMI,
with nmis enabled and disabled from the command line, and without
CONFIG_ARM64_PSEUDO_NMI, found nothing out of the ordinary. For the PMU part:

Acked-by: Alexandru Elisei <alexandru.elisei@arm.com>

Thanks,

Alex

>
> Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> ---
>
> Changes in v4:
> - Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
>   hook to know if PMU IRQ has been requested as an NMI.
> - Add check for return value prior to initializing hard-lockup detector.
>
> [1] https://lkml.org/lkml/2020/9/24/458
>
> Changes in v3:
> - Rebased to latest pmu NMI patch-set [1].
> - Addressed misc. comments from Stephen.
>
> [1] https://lkml.org/lkml/2020/8/19/671
>
> Changes since RFC:
> - Rebased on top of Alex's WIP-pmu-nmi branch.
> - Add comment for safe max. CPU frequency.
> - Misc. cleanup.
>
>  arch/arm64/Kconfig             |  2 ++
>  arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
>  drivers/perf/arm_pmu.c         |  5 +++++
>  include/linux/perf/arm_pmu.h   |  2 ++
>  4 files changed, 48 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 6d23283..b5c2594 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -167,6 +167,8 @@ config ARM64
>  	select HAVE_NMI
>  	select HAVE_PATA_PLATFORM
>  	select HAVE_PERF_EVENTS
> +	select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
> +	select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
>  	select HAVE_PERF_REGS
>  	select HAVE_PERF_USER_STACK_DUMP
>  	select HAVE_REGS_AND_STACK_ACCESS_API
> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
> index ef206fb..6ad5120 100644
> --- a/arch/arm64/kernel/perf_event.c
> +++ b/arch/arm64/kernel/perf_event.c
> @@ -23,6 +23,8 @@
>  #include <linux/platform_device.h>
>  #include <linux/sched_clock.h>
>  #include <linux/smp.h>
> +#include <linux/nmi.h>
> +#include <linux/cpufreq.h>
>  
>  /* ARMv8 Cortex-A53 specific event types. */
>  #define ARMV8_A53_PERFCTR_PREF_LINEFILL				0xC2
> @@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
>  
>  static int __init armv8_pmu_driver_init(void)
>  {
> +	int ret;
> +
>  	if (acpi_disabled)
> -		return platform_driver_register(&armv8_pmu_driver);
> +		ret = platform_driver_register(&armv8_pmu_driver);
>  	else
> -		return arm_pmu_acpi_probe(armv8_pmuv3_init);
> +		ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
> +
> +	/*
> +	 * Try to re-initialize lockup detector after PMU init in
> +	 * case PMU events are triggered via NMIs.
> +	 */
> +	if (ret == 0 && arm_pmu_irq_is_nmi())
> +		lockup_detector_init();
> +
> +	return ret;
>  }
>  device_initcall(armv8_pmu_driver_init)
>  
> @@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
>  	userpg->cap_user_time_zero = 1;
>  	userpg->cap_user_time_short = 1;
>  }
> +
> +#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
> +/*
> + * Safe maximum CPU frequency in case a particular platform doesn't implement
> + * cpufreq driver. Although, architecture doesn't put any restrictions on
> + * maximum frequency but 5 GHz seems to be safe maximum given the available
> + * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
> + * hand, we can't make it much higher as it would lead to a large hard-lockup
> + * detection timeout on parts which are running slower (eg. 1GHz on
> + * Developerbox) and doesn't possess a cpufreq driver.
> + */
> +#define SAFE_MAX_CPU_FREQ	5000000000UL // 5 GHz
> +u64 hw_nmi_get_sample_period(int watchdog_thresh)
> +{
> +	unsigned int cpu = smp_processor_id();
> +	unsigned long max_cpu_freq;
> +
> +	max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
> +	if (!max_cpu_freq)
> +		max_cpu_freq = SAFE_MAX_CPU_FREQ;
> +
> +	return (u64)max_cpu_freq * watchdog_thresh;
> +}
> +#endif
> diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> index cb2f55f..794a37d 100644
> --- a/drivers/perf/arm_pmu.c
> +++ b/drivers/perf/arm_pmu.c
> @@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
>  	return per_cpu(hw_events->irq, cpu);
>  }
>  
> +bool arm_pmu_irq_is_nmi(void)
> +{
> +	return has_nmi;
> +}
> +
>  /*
>   * PMU hardware loses all context when a CPU goes offline.
>   * When a CPU is hotplugged back in, since some hardware registers are
> diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
> index 5b616dd..5765069 100644
> --- a/include/linux/perf/arm_pmu.h
> +++ b/include/linux/perf/arm_pmu.h
> @@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
>  static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
>  #endif
>  
> +bool arm_pmu_irq_is_nmi(void);
> +
>  /* Internal functions only for core arm_pmu code */
>  struct arm_pmu *armpmu_alloc(void);
>  struct arm_pmu *armpmu_alloc_atomic(void);

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
@ 2020-10-09 10:26   ` Alexandru Elisei
  0 siblings, 0 replies; 10+ messages in thread
From: Alexandru Elisei @ 2020-10-09 10:26 UTC (permalink / raw)
  To: Sumit Garg, will
  Cc: mark.rutland, daniel.thompson, linux-kernel, peterz,
	catalin.marinas, dianders, acme, swboyd, alexander.shishkin,
	mingo, julien.thierry.kdev, namhyung, tglx, jolsa,
	linux-arm-kernel

Hi,

On 10/7/20 9:51 AM, Sumit Garg wrote:
> With the recent feature added to enable perf events to use pseudo NMIs
> as interrupts on platforms which support GICv3 or later, its now been
> possible to enable hard lockup detector (or NMI watchdog) on arm64
> platforms. So enable corresponding support.
>
> One thing to note here is that normally lockup detector is initialized
> just after the early initcalls but PMU on arm64 comes up much later as
> device_initcall(). So we need to re-initialize lockup detection once
> PMU has been initialized.

Has another look, the PMU bits looks alright. Just to be on the safe side, I did a
few quick boot tests on an espressobin with a kernel with CONFIG_ARM64_PSEUDO_NMI,
with nmis enabled and disabled from the command line, and without
CONFIG_ARM64_PSEUDO_NMI, found nothing out of the ordinary. For the PMU part:

Acked-by: Alexandru Elisei <alexandru.elisei@arm.com>

Thanks,

Alex

>
> Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> ---
>
> Changes in v4:
> - Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
>   hook to know if PMU IRQ has been requested as an NMI.
> - Add check for return value prior to initializing hard-lockup detector.
>
> [1] https://lkml.org/lkml/2020/9/24/458
>
> Changes in v3:
> - Rebased to latest pmu NMI patch-set [1].
> - Addressed misc. comments from Stephen.
>
> [1] https://lkml.org/lkml/2020/8/19/671
>
> Changes since RFC:
> - Rebased on top of Alex's WIP-pmu-nmi branch.
> - Add comment for safe max. CPU frequency.
> - Misc. cleanup.
>
>  arch/arm64/Kconfig             |  2 ++
>  arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
>  drivers/perf/arm_pmu.c         |  5 +++++
>  include/linux/perf/arm_pmu.h   |  2 ++
>  4 files changed, 48 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 6d23283..b5c2594 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -167,6 +167,8 @@ config ARM64
>  	select HAVE_NMI
>  	select HAVE_PATA_PLATFORM
>  	select HAVE_PERF_EVENTS
> +	select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
> +	select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
>  	select HAVE_PERF_REGS
>  	select HAVE_PERF_USER_STACK_DUMP
>  	select HAVE_REGS_AND_STACK_ACCESS_API
> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
> index ef206fb..6ad5120 100644
> --- a/arch/arm64/kernel/perf_event.c
> +++ b/arch/arm64/kernel/perf_event.c
> @@ -23,6 +23,8 @@
>  #include <linux/platform_device.h>
>  #include <linux/sched_clock.h>
>  #include <linux/smp.h>
> +#include <linux/nmi.h>
> +#include <linux/cpufreq.h>
>  
>  /* ARMv8 Cortex-A53 specific event types. */
>  #define ARMV8_A53_PERFCTR_PREF_LINEFILL				0xC2
> @@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
>  
>  static int __init armv8_pmu_driver_init(void)
>  {
> +	int ret;
> +
>  	if (acpi_disabled)
> -		return platform_driver_register(&armv8_pmu_driver);
> +		ret = platform_driver_register(&armv8_pmu_driver);
>  	else
> -		return arm_pmu_acpi_probe(armv8_pmuv3_init);
> +		ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
> +
> +	/*
> +	 * Try to re-initialize lockup detector after PMU init in
> +	 * case PMU events are triggered via NMIs.
> +	 */
> +	if (ret == 0 && arm_pmu_irq_is_nmi())
> +		lockup_detector_init();
> +
> +	return ret;
>  }
>  device_initcall(armv8_pmu_driver_init)
>  
> @@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
>  	userpg->cap_user_time_zero = 1;
>  	userpg->cap_user_time_short = 1;
>  }
> +
> +#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
> +/*
> + * Safe maximum CPU frequency in case a particular platform doesn't implement
> + * cpufreq driver. Although, architecture doesn't put any restrictions on
> + * maximum frequency but 5 GHz seems to be safe maximum given the available
> + * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
> + * hand, we can't make it much higher as it would lead to a large hard-lockup
> + * detection timeout on parts which are running slower (eg. 1GHz on
> + * Developerbox) and doesn't possess a cpufreq driver.
> + */
> +#define SAFE_MAX_CPU_FREQ	5000000000UL // 5 GHz
> +u64 hw_nmi_get_sample_period(int watchdog_thresh)
> +{
> +	unsigned int cpu = smp_processor_id();
> +	unsigned long max_cpu_freq;
> +
> +	max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
> +	if (!max_cpu_freq)
> +		max_cpu_freq = SAFE_MAX_CPU_FREQ;
> +
> +	return (u64)max_cpu_freq * watchdog_thresh;
> +}
> +#endif
> diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> index cb2f55f..794a37d 100644
> --- a/drivers/perf/arm_pmu.c
> +++ b/drivers/perf/arm_pmu.c
> @@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
>  	return per_cpu(hw_events->irq, cpu);
>  }
>  
> +bool arm_pmu_irq_is_nmi(void)
> +{
> +	return has_nmi;
> +}
> +
>  /*
>   * PMU hardware loses all context when a CPU goes offline.
>   * When a CPU is hotplugged back in, since some hardware registers are
> diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
> index 5b616dd..5765069 100644
> --- a/include/linux/perf/arm_pmu.h
> +++ b/include/linux/perf/arm_pmu.h
> @@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
>  static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
>  #endif
>  
> +bool arm_pmu_irq_is_nmi(void);
> +
>  /* Internal functions only for core arm_pmu code */
>  struct arm_pmu *armpmu_alloc(void);
>  struct arm_pmu *armpmu_alloc_atomic(void);

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
  2020-10-09 10:26   ` Alexandru Elisei
@ 2020-10-29  9:22     ` Sumit Garg
  -1 siblings, 0 replies; 10+ messages in thread
From: Sumit Garg @ 2020-10-29  9:22 UTC (permalink / raw)
  To: Alexandru Elisei, Will Deacon
  Cc: linux-arm-kernel, Catalin Marinas, Mark Rutland, Peter Zijlstra,
	mingo, acme, alexander.shishkin, jolsa, namhyung,
	Thomas Gleixner, Stephen Boyd, julien.thierry.kdev,
	Douglas Anderson, Daniel Thompson, Linux Kernel Mailing List

On Fri, 9 Oct 2020 at 15:55, Alexandru Elisei <alexandru.elisei@arm.com> wrote:
>
> Hi,
>
> On 10/7/20 9:51 AM, Sumit Garg wrote:
> > With the recent feature added to enable perf events to use pseudo NMIs
> > as interrupts on platforms which support GICv3 or later, its now been
> > possible to enable hard lockup detector (or NMI watchdog) on arm64
> > platforms. So enable corresponding support.
> >
> > One thing to note here is that normally lockup detector is initialized
> > just after the early initcalls but PMU on arm64 comes up much later as
> > device_initcall(). So we need to re-initialize lockup detection once
> > PMU has been initialized.
>
> Has another look, the PMU bits looks alright. Just to be on the safe side, I did a
> few quick boot tests on an espressobin with a kernel with CONFIG_ARM64_PSEUDO_NMI,
> with nmis enabled and disabled from the command line, and without
> CONFIG_ARM64_PSEUDO_NMI, found nothing out of the ordinary. For the PMU part:
>
> Acked-by: Alexandru Elisei <alexandru.elisei@arm.com>
>

Thanks Alex.

Will,

Do you have any further comments on this patch? If no, can you please
pick up this patch?

-Sumit

> Thanks,
>
> Alex
>
> >
> > Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> > ---
> >
> > Changes in v4:
> > - Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
> >   hook to know if PMU IRQ has been requested as an NMI.
> > - Add check for return value prior to initializing hard-lockup detector.
> >
> > [1] https://lkml.org/lkml/2020/9/24/458
> >
> > Changes in v3:
> > - Rebased to latest pmu NMI patch-set [1].
> > - Addressed misc. comments from Stephen.
> >
> > [1] https://lkml.org/lkml/2020/8/19/671
> >
> > Changes since RFC:
> > - Rebased on top of Alex's WIP-pmu-nmi branch.
> > - Add comment for safe max. CPU frequency.
> > - Misc. cleanup.
> >
> >  arch/arm64/Kconfig             |  2 ++
> >  arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
> >  drivers/perf/arm_pmu.c         |  5 +++++
> >  include/linux/perf/arm_pmu.h   |  2 ++
> >  4 files changed, 48 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index 6d23283..b5c2594 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -167,6 +167,8 @@ config ARM64
> >       select HAVE_NMI
> >       select HAVE_PATA_PLATFORM
> >       select HAVE_PERF_EVENTS
> > +     select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
> > +     select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
> >       select HAVE_PERF_REGS
> >       select HAVE_PERF_USER_STACK_DUMP
> >       select HAVE_REGS_AND_STACK_ACCESS_API
> > diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
> > index ef206fb..6ad5120 100644
> > --- a/arch/arm64/kernel/perf_event.c
> > +++ b/arch/arm64/kernel/perf_event.c
> > @@ -23,6 +23,8 @@
> >  #include <linux/platform_device.h>
> >  #include <linux/sched_clock.h>
> >  #include <linux/smp.h>
> > +#include <linux/nmi.h>
> > +#include <linux/cpufreq.h>
> >
> >  /* ARMv8 Cortex-A53 specific event types. */
> >  #define ARMV8_A53_PERFCTR_PREF_LINEFILL                              0xC2
> > @@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
> >
> >  static int __init armv8_pmu_driver_init(void)
> >  {
> > +     int ret;
> > +
> >       if (acpi_disabled)
> > -             return platform_driver_register(&armv8_pmu_driver);
> > +             ret = platform_driver_register(&armv8_pmu_driver);
> >       else
> > -             return arm_pmu_acpi_probe(armv8_pmuv3_init);
> > +             ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
> > +
> > +     /*
> > +      * Try to re-initialize lockup detector after PMU init in
> > +      * case PMU events are triggered via NMIs.
> > +      */
> > +     if (ret == 0 && arm_pmu_irq_is_nmi())
> > +             lockup_detector_init();
> > +
> > +     return ret;
> >  }
> >  device_initcall(armv8_pmu_driver_init)
> >
> > @@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
> >       userpg->cap_user_time_zero = 1;
> >       userpg->cap_user_time_short = 1;
> >  }
> > +
> > +#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
> > +/*
> > + * Safe maximum CPU frequency in case a particular platform doesn't implement
> > + * cpufreq driver. Although, architecture doesn't put any restrictions on
> > + * maximum frequency but 5 GHz seems to be safe maximum given the available
> > + * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
> > + * hand, we can't make it much higher as it would lead to a large hard-lockup
> > + * detection timeout on parts which are running slower (eg. 1GHz on
> > + * Developerbox) and doesn't possess a cpufreq driver.
> > + */
> > +#define SAFE_MAX_CPU_FREQ    5000000000UL // 5 GHz
> > +u64 hw_nmi_get_sample_period(int watchdog_thresh)
> > +{
> > +     unsigned int cpu = smp_processor_id();
> > +     unsigned long max_cpu_freq;
> > +
> > +     max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
> > +     if (!max_cpu_freq)
> > +             max_cpu_freq = SAFE_MAX_CPU_FREQ;
> > +
> > +     return (u64)max_cpu_freq * watchdog_thresh;
> > +}
> > +#endif
> > diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> > index cb2f55f..794a37d 100644
> > --- a/drivers/perf/arm_pmu.c
> > +++ b/drivers/perf/arm_pmu.c
> > @@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
> >       return per_cpu(hw_events->irq, cpu);
> >  }
> >
> > +bool arm_pmu_irq_is_nmi(void)
> > +{
> > +     return has_nmi;
> > +}
> > +
> >  /*
> >   * PMU hardware loses all context when a CPU goes offline.
> >   * When a CPU is hotplugged back in, since some hardware registers are
> > diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
> > index 5b616dd..5765069 100644
> > --- a/include/linux/perf/arm_pmu.h
> > +++ b/include/linux/perf/arm_pmu.h
> > @@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
> >  static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
> >  #endif
> >
> > +bool arm_pmu_irq_is_nmi(void);
> > +
> >  /* Internal functions only for core arm_pmu code */
> >  struct arm_pmu *armpmu_alloc(void);
> >  struct arm_pmu *armpmu_alloc_atomic(void);

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
@ 2020-10-29  9:22     ` Sumit Garg
  0 siblings, 0 replies; 10+ messages in thread
From: Sumit Garg @ 2020-10-29  9:22 UTC (permalink / raw)
  To: Alexandru Elisei, Will Deacon
  Cc: Mark Rutland, Daniel Thompson, Linux Kernel Mailing List,
	Peter Zijlstra, Catalin Marinas, Douglas Anderson, acme,
	Stephen Boyd, alexander.shishkin, mingo, julien.thierry.kdev,
	namhyung, Thomas Gleixner, jolsa, linux-arm-kernel

On Fri, 9 Oct 2020 at 15:55, Alexandru Elisei <alexandru.elisei@arm.com> wrote:
>
> Hi,
>
> On 10/7/20 9:51 AM, Sumit Garg wrote:
> > With the recent feature added to enable perf events to use pseudo NMIs
> > as interrupts on platforms which support GICv3 or later, its now been
> > possible to enable hard lockup detector (or NMI watchdog) on arm64
> > platforms. So enable corresponding support.
> >
> > One thing to note here is that normally lockup detector is initialized
> > just after the early initcalls but PMU on arm64 comes up much later as
> > device_initcall(). So we need to re-initialize lockup detection once
> > PMU has been initialized.
>
> Has another look, the PMU bits looks alright. Just to be on the safe side, I did a
> few quick boot tests on an espressobin with a kernel with CONFIG_ARM64_PSEUDO_NMI,
> with nmis enabled and disabled from the command line, and without
> CONFIG_ARM64_PSEUDO_NMI, found nothing out of the ordinary. For the PMU part:
>
> Acked-by: Alexandru Elisei <alexandru.elisei@arm.com>
>

Thanks Alex.

Will,

Do you have any further comments on this patch? If no, can you please
pick up this patch?

-Sumit

> Thanks,
>
> Alex
>
> >
> > Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> > ---
> >
> > Changes in v4:
> > - Rebased to latest pmu v7 NMI patch-set [1] and in turn use "has_nmi"
> >   hook to know if PMU IRQ has been requested as an NMI.
> > - Add check for return value prior to initializing hard-lockup detector.
> >
> > [1] https://lkml.org/lkml/2020/9/24/458
> >
> > Changes in v3:
> > - Rebased to latest pmu NMI patch-set [1].
> > - Addressed misc. comments from Stephen.
> >
> > [1] https://lkml.org/lkml/2020/8/19/671
> >
> > Changes since RFC:
> > - Rebased on top of Alex's WIP-pmu-nmi branch.
> > - Add comment for safe max. CPU frequency.
> > - Misc. cleanup.
> >
> >  arch/arm64/Kconfig             |  2 ++
> >  arch/arm64/kernel/perf_event.c | 41 +++++++++++++++++++++++++++++++++++++++--
> >  drivers/perf/arm_pmu.c         |  5 +++++
> >  include/linux/perf/arm_pmu.h   |  2 ++
> >  4 files changed, 48 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index 6d23283..b5c2594 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -167,6 +167,8 @@ config ARM64
> >       select HAVE_NMI
> >       select HAVE_PATA_PLATFORM
> >       select HAVE_PERF_EVENTS
> > +     select HAVE_PERF_EVENTS_NMI if ARM64_PSEUDO_NMI
> > +     select HAVE_HARDLOCKUP_DETECTOR_PERF if PERF_EVENTS && HAVE_PERF_EVENTS_NMI
> >       select HAVE_PERF_REGS
> >       select HAVE_PERF_USER_STACK_DUMP
> >       select HAVE_REGS_AND_STACK_ACCESS_API
> > diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
> > index ef206fb..6ad5120 100644
> > --- a/arch/arm64/kernel/perf_event.c
> > +++ b/arch/arm64/kernel/perf_event.c
> > @@ -23,6 +23,8 @@
> >  #include <linux/platform_device.h>
> >  #include <linux/sched_clock.h>
> >  #include <linux/smp.h>
> > +#include <linux/nmi.h>
> > +#include <linux/cpufreq.h>
> >
> >  /* ARMv8 Cortex-A53 specific event types. */
> >  #define ARMV8_A53_PERFCTR_PREF_LINEFILL                              0xC2
> > @@ -1224,10 +1226,21 @@ static struct platform_driver armv8_pmu_driver = {
> >
> >  static int __init armv8_pmu_driver_init(void)
> >  {
> > +     int ret;
> > +
> >       if (acpi_disabled)
> > -             return platform_driver_register(&armv8_pmu_driver);
> > +             ret = platform_driver_register(&armv8_pmu_driver);
> >       else
> > -             return arm_pmu_acpi_probe(armv8_pmuv3_init);
> > +             ret = arm_pmu_acpi_probe(armv8_pmuv3_init);
> > +
> > +     /*
> > +      * Try to re-initialize lockup detector after PMU init in
> > +      * case PMU events are triggered via NMIs.
> > +      */
> > +     if (ret == 0 && arm_pmu_irq_is_nmi())
> > +             lockup_detector_init();
> > +
> > +     return ret;
> >  }
> >  device_initcall(armv8_pmu_driver_init)
> >
> > @@ -1285,3 +1298,27 @@ void arch_perf_update_userpage(struct perf_event *event,
> >       userpg->cap_user_time_zero = 1;
> >       userpg->cap_user_time_short = 1;
> >  }
> > +
> > +#ifdef CONFIG_HARDLOCKUP_DETECTOR_PERF
> > +/*
> > + * Safe maximum CPU frequency in case a particular platform doesn't implement
> > + * cpufreq driver. Although, architecture doesn't put any restrictions on
> > + * maximum frequency but 5 GHz seems to be safe maximum given the available
> > + * Arm CPUs in the market which are clocked much less than 5 GHz. On the other
> > + * hand, we can't make it much higher as it would lead to a large hard-lockup
> > + * detection timeout on parts which are running slower (eg. 1GHz on
> > + * Developerbox) and doesn't possess a cpufreq driver.
> > + */
> > +#define SAFE_MAX_CPU_FREQ    5000000000UL // 5 GHz
> > +u64 hw_nmi_get_sample_period(int watchdog_thresh)
> > +{
> > +     unsigned int cpu = smp_processor_id();
> > +     unsigned long max_cpu_freq;
> > +
> > +     max_cpu_freq = cpufreq_get_hw_max_freq(cpu) * 1000UL;
> > +     if (!max_cpu_freq)
> > +             max_cpu_freq = SAFE_MAX_CPU_FREQ;
> > +
> > +     return (u64)max_cpu_freq * watchdog_thresh;
> > +}
> > +#endif
> > diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> > index cb2f55f..794a37d 100644
> > --- a/drivers/perf/arm_pmu.c
> > +++ b/drivers/perf/arm_pmu.c
> > @@ -726,6 +726,11 @@ static int armpmu_get_cpu_irq(struct arm_pmu *pmu, int cpu)
> >       return per_cpu(hw_events->irq, cpu);
> >  }
> >
> > +bool arm_pmu_irq_is_nmi(void)
> > +{
> > +     return has_nmi;
> > +}
> > +
> >  /*
> >   * PMU hardware loses all context when a CPU goes offline.
> >   * When a CPU is hotplugged back in, since some hardware registers are
> > diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
> > index 5b616dd..5765069 100644
> > --- a/include/linux/perf/arm_pmu.h
> > +++ b/include/linux/perf/arm_pmu.h
> > @@ -160,6 +160,8 @@ int arm_pmu_acpi_probe(armpmu_init_fn init_fn);
> >  static inline int arm_pmu_acpi_probe(armpmu_init_fn init_fn) { return 0; }
> >  #endif
> >
> > +bool arm_pmu_irq_is_nmi(void);
> > +
> >  /* Internal functions only for core arm_pmu code */
> >  struct arm_pmu *armpmu_alloc(void);
> >  struct arm_pmu *armpmu_alloc_atomic(void);

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
  2020-10-07  8:51 ` Sumit Garg
@ 2020-11-25 15:44   ` Will Deacon
  -1 siblings, 0 replies; 10+ messages in thread
From: Will Deacon @ 2020-11-25 15:44 UTC (permalink / raw)
  To: Sumit Garg, alexandru.elisei
  Cc: catalin.marinas, kernel-team, Will Deacon, tglx,
	linux-arm-kernel, linux-kernel, mingo, mark.rutland, peterz,
	dianders, acme, swboyd, namhyung, alexander.shishkin, jolsa,
	julien.thierry.kdev, daniel.thompson

On Wed, 7 Oct 2020 14:21:43 +0530, Sumit Garg wrote:
> With the recent feature added to enable perf events to use pseudo NMIs
> as interrupts on platforms which support GICv3 or later, its now been
> possible to enable hard lockup detector (or NMI watchdog) on arm64
> platforms. So enable corresponding support.
> 
> One thing to note here is that normally lockup detector is initialized
> just after the early initcalls but PMU on arm64 comes up much later as
> device_initcall(). So we need to re-initialize lockup detection once
> PMU has been initialized.

Applied to will (for-next/perf), thanks!

[1/1] arm64: Enable perf events based hard lockup detector
      https://git.kernel.org/will/c/367c820ef080

Cheers,
-- 
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v4] arm64: Enable perf events based hard lockup detector
@ 2020-11-25 15:44   ` Will Deacon
  0 siblings, 0 replies; 10+ messages in thread
From: Will Deacon @ 2020-11-25 15:44 UTC (permalink / raw)
  To: Sumit Garg, alexandru.elisei
  Cc: mark.rutland, daniel.thompson, Will Deacon, peterz,
	catalin.marinas, jolsa, linux-kernel, acme, dianders,
	alexander.shishkin, mingo, julien.thierry.kdev, namhyung, tglx,
	swboyd, kernel-team, linux-arm-kernel

On Wed, 7 Oct 2020 14:21:43 +0530, Sumit Garg wrote:
> With the recent feature added to enable perf events to use pseudo NMIs
> as interrupts on platforms which support GICv3 or later, its now been
> possible to enable hard lockup detector (or NMI watchdog) on arm64
> platforms. So enable corresponding support.
> 
> One thing to note here is that normally lockup detector is initialized
> just after the early initcalls but PMU on arm64 comes up much later as
> device_initcall(). So we need to re-initialize lockup detection once
> PMU has been initialized.

Applied to will (for-next/perf), thanks!

[1/1] arm64: Enable perf events based hard lockup detector
      https://git.kernel.org/will/c/367c820ef080

Cheers,
-- 
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2020-11-25 15:45 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-10-07  8:51 [PATCH v4] arm64: Enable perf events based hard lockup detector Sumit Garg
2020-10-07  8:51 ` Sumit Garg
2020-10-09 10:23 ` Sumit Garg
2020-10-09 10:23   ` Sumit Garg
2020-10-09 10:26 ` Alexandru Elisei
2020-10-09 10:26   ` Alexandru Elisei
2020-10-29  9:22   ` Sumit Garg
2020-10-29  9:22     ` Sumit Garg
2020-11-25 15:44 ` Will Deacon
2020-11-25 15:44   ` Will Deacon

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.