The AMU feature can be enabled on a subset of the cores in a system.
Because of that, it prints a message for each core as it is detected.
This becomes tedious when there are hundreds of cores. Instead, for
CPU features which can be enabled on a subset of the present cores,
lets wait until update_cpu_capabilities() and print the subset of cores
the feature was enabled on.
Signed-off-by: Jeremy Linton <[email protected]>
---
arch/arm64/include/asm/cpufeature.h | 2 ++
arch/arm64/kernel/cpufeature.c | 16 +++++++++++++---
2 files changed, 15 insertions(+), 3 deletions(-)
diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
index 5bba39376055..19b4d001d845 100644
--- a/arch/arm64/include/asm/cpufeature.h
+++ b/arch/arm64/include/asm/cpufeature.h
@@ -23,6 +23,7 @@
#include <linux/bug.h>
#include <linux/jump_label.h>
#include <linux/kernel.h>
+#include <linux/cpumask.h>
/*
* CPU feature register tracking
@@ -380,6 +381,7 @@ struct arm64_cpu_capabilities {
* method is robust against being called multiple times.
*/
const struct arm64_cpu_capabilities *match_list;
+ const struct cpumask *cpus;
};
static inline int cpucap_default_scope(const struct arm64_cpu_capabilities *cap)
diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index 444a73c2e638..18711e35924c 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -1944,8 +1944,6 @@ int get_cpu_with_amu_feat(void)
static void cpu_amu_enable(struct arm64_cpu_capabilities const *cap)
{
if (has_cpuid_feature(cap, SCOPE_LOCAL_CPU)) {
- pr_info("detected CPU%d: Activity Monitors Unit (AMU)\n",
- smp_processor_id());
cpumask_set_cpu(smp_processor_id(), &amu_cpus);
/* 0 reference values signal broken/disabled counters */
@@ -2411,10 +2409,12 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
* message to be shown until at least one CPU is detected to
* support the feature.
*/
+ .desc = "Activity Monitors Unit (AMU)",
.capability = ARM64_HAS_AMU_EXTN,
.type = ARM64_CPUCAP_WEAK_LOCAL_CPU_FEATURE,
.matches = has_amu,
.cpu_enable = cpu_amu_enable,
+ .cpus = &amu_cpus,
ARM64_CPUID_FIELDS(ID_AA64PFR0_EL1, AMU, IMP)
},
#endif /* CONFIG_ARM64_AMU_EXTN */
@@ -2981,7 +2981,7 @@ static void update_cpu_capabilities(u16 scope_mask)
!caps->matches(caps, cpucap_default_scope(caps)))
continue;
- if (caps->desc)
+ if (caps->desc && !caps->cpus)
pr_info("detected: %s\n", caps->desc);
__set_bit(caps->capability, system_cpucaps);
@@ -3330,6 +3330,7 @@ unsigned long cpu_get_elf_hwcap2(void)
static void __init setup_system_capabilities(void)
{
+ int i;
/*
* We have finalised the system-wide safe feature
* registers, finalise the capabilities that depend
@@ -3338,6 +3339,15 @@ static void __init setup_system_capabilities(void)
*/
update_cpu_capabilities(SCOPE_SYSTEM);
enable_cpu_capabilities(SCOPE_ALL & ~SCOPE_BOOT_CPU);
+
+ for (i = 0; i < ARM64_NCAPS; i++) {
+ const struct arm64_cpu_capabilities *caps = cpucap_ptrs[i];
+
+ if (caps && caps->cpus && caps->desc &&
+ cpumask_any(caps->cpus) < nr_cpu_ids)
+ pr_info("detected: %s on CPU%*pbl\n",
+ caps->desc, cpumask_pr_args(caps->cpus));
+ }
}
void __init setup_cpu_features(void)
--
2.41.0
On Thu, Oct 12, 2023 at 08:30:16PM -0500, Jeremy Linton wrote:
> The AMU feature can be enabled on a subset of the cores in a system.
> Because of that, it prints a message for each core as it is detected.
> This becomes tedious when there are hundreds of cores. Instead, for
> CPU features which can be enabled on a subset of the present cores,
> lets wait until update_cpu_capabilities() and print the subset of cores
> the feature was enabled on.
>
> Signed-off-by: Jeremy Linton <[email protected]>
> ---
> arch/arm64/include/asm/cpufeature.h | 2 ++
> arch/arm64/kernel/cpufeature.c | 16 +++++++++++++---
> 2 files changed, 15 insertions(+), 3 deletions(-)
That's pretty neat. Are there any other caps we could switch over to this
new mechanism as well? We have a weird bodge, for example, in has_hw_dbm()
to print a "detected: ..." message and I wonder whether following the
example of the AMU would be better?
Will
On Fri, Oct 13, 2023 at 09:29:48AM +0100, Will Deacon wrote:
> That's pretty neat. Are there any other caps we could switch over to this
> new mechanism as well? We have a weird bodge, for example, in has_hw_dbm()
> to print a "detected: ..." message and I wonder whether following the
> example of the AMU would be better?
There's 32 bit EL0, though that's a bit different...
Hi,
On Thursday 12 Oct 2023 at 20:30:16 (-0500), Jeremy Linton wrote:
> The AMU feature can be enabled on a subset of the cores in a system.
> Because of that, it prints a message for each core as it is detected.
> This becomes tedious when there are hundreds of cores. Instead, for
> CPU features which can be enabled on a subset of the present cores,
> lets wait until update_cpu_capabilities() and print the subset of cores
> the feature was enabled on.
>
> Signed-off-by: Jeremy Linton <[email protected]>
> ---
> arch/arm64/include/asm/cpufeature.h | 2 ++
> arch/arm64/kernel/cpufeature.c | 16 +++++++++++++---
> 2 files changed, 15 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
> index 5bba39376055..19b4d001d845 100644
> --- a/arch/arm64/include/asm/cpufeature.h
> +++ b/arch/arm64/include/asm/cpufeature.h
> @@ -23,6 +23,7 @@
> #include <linux/bug.h>
> #include <linux/jump_label.h>
> #include <linux/kernel.h>
> +#include <linux/cpumask.h>
>
> /*
> * CPU feature register tracking
> @@ -380,6 +381,7 @@ struct arm64_cpu_capabilities {
> * method is robust against being called multiple times.
> */
> const struct arm64_cpu_capabilities *match_list;
> + const struct cpumask *cpus;
> };
>
> static inline int cpucap_default_scope(const struct arm64_cpu_capabilities *cap)
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index 444a73c2e638..18711e35924c 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -1944,8 +1944,6 @@ int get_cpu_with_amu_feat(void)
> static void cpu_amu_enable(struct arm64_cpu_capabilities const *cap)
> {
> if (has_cpuid_feature(cap, SCOPE_LOCAL_CPU)) {
> - pr_info("detected CPU%d: Activity Monitors Unit (AMU)\n",
> - smp_processor_id());
> cpumask_set_cpu(smp_processor_id(), &amu_cpus);
>
> /* 0 reference values signal broken/disabled counters */
> @@ -2411,10 +2409,12 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
> * message to be shown until at least one CPU is detected to
> * support the feature.
> */
> + .desc = "Activity Monitors Unit (AMU)",
> .capability = ARM64_HAS_AMU_EXTN,
> .type = ARM64_CPUCAP_WEAK_LOCAL_CPU_FEATURE,
> .matches = has_amu,
> .cpu_enable = cpu_amu_enable,
> + .cpus = &amu_cpus,
> ARM64_CPUID_FIELDS(ID_AA64PFR0_EL1, AMU, IMP)
> },
> #endif /* CONFIG_ARM64_AMU_EXTN */
> @@ -2981,7 +2981,7 @@ static void update_cpu_capabilities(u16 scope_mask)
> !caps->matches(caps, cpucap_default_scope(caps)))
> continue;
>
> - if (caps->desc)
> + if (caps->desc && !caps->cpus)
> pr_info("detected: %s\n", caps->desc);
>
> __set_bit(caps->capability, system_cpucaps);
> @@ -3330,6 +3330,7 @@ unsigned long cpu_get_elf_hwcap2(void)
>
> static void __init setup_system_capabilities(void)
> {
> + int i;
> /*
> * We have finalised the system-wide safe feature
> * registers, finalise the capabilities that depend
> @@ -3338,6 +3339,15 @@ static void __init setup_system_capabilities(void)
> */
> update_cpu_capabilities(SCOPE_SYSTEM);
> enable_cpu_capabilities(SCOPE_ALL & ~SCOPE_BOOT_CPU);
> +
> + for (i = 0; i < ARM64_NCAPS; i++) {
> + const struct arm64_cpu_capabilities *caps = cpucap_ptrs[i];
> +
> + if (caps && caps->cpus && caps->desc &&
> + cpumask_any(caps->cpus) < nr_cpu_ids)
> + pr_info("detected: %s on CPU%*pbl\n",
> + caps->desc, cpumask_pr_args(caps->cpus));
> + }
> }
>
> void __init setup_cpu_features(void)
> --
> 2.41.0
>
It looks very nice to me! I gave it a run on an RD-N2 FVP so FWIW:
Reviewed-by: Ionela Voinescu <[email protected]>
Tested-by: Ionela Voinescu <[email protected]>
The same could be used for DBM, but I'm not sure if creating a mask of
supporting CPUs is worth it, for this purpose only.
Hope it helps,
Ionela.
Jeremy Linton <[email protected]> writes:
> The AMU feature can be enabled on a subset of the cores in a system.
> Because of that, it prints a message for each core as it is detected.
> This becomes tedious when there are hundreds of cores. Instead, for
> CPU features which can be enabled on a subset of the present cores,
> lets wait until update_cpu_capabilities() and print the subset of cores
> the feature was enabled on.
>
> Signed-off-by: Jeremy Linton <[email protected]>
> ---
> arch/arm64/include/asm/cpufeature.h | 2 ++
> arch/arm64/kernel/cpufeature.c | 16 +++++++++++++---
> 2 files changed, 15 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
> index 5bba39376055..19b4d001d845 100644
> --- a/arch/arm64/include/asm/cpufeature.h
> +++ b/arch/arm64/include/asm/cpufeature.h
> @@ -23,6 +23,7 @@
> #include <linux/bug.h>
> #include <linux/jump_label.h>
> #include <linux/kernel.h>
> +#include <linux/cpumask.h>
>
> /*
> * CPU feature register tracking
> @@ -380,6 +381,7 @@ struct arm64_cpu_capabilities {
> * method is robust against being called multiple times.
> */
> const struct arm64_cpu_capabilities *match_list;
> + const struct cpumask *cpus;
> };
>
> static inline int cpucap_default_scope(const struct arm64_cpu_capabilities *cap)
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index 444a73c2e638..18711e35924c 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -1944,8 +1944,6 @@ int get_cpu_with_amu_feat(void)
> static void cpu_amu_enable(struct arm64_cpu_capabilities const *cap)
> {
> if (has_cpuid_feature(cap, SCOPE_LOCAL_CPU)) {
> - pr_info("detected CPU%d: Activity Monitors Unit (AMU)\n",
> - smp_processor_id());
> cpumask_set_cpu(smp_processor_id(), &amu_cpus);
>
> /* 0 reference values signal broken/disabled counters */
> @@ -2411,10 +2409,12 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
> * message to be shown until at least one CPU is detected to
> * support the feature.
> */
> + .desc = "Activity Monitors Unit (AMU)",
> .capability = ARM64_HAS_AMU_EXTN,
> .type = ARM64_CPUCAP_WEAK_LOCAL_CPU_FEATURE,
> .matches = has_amu,
> .cpu_enable = cpu_amu_enable,
> + .cpus = &amu_cpus,
> ARM64_CPUID_FIELDS(ID_AA64PFR0_EL1, AMU, IMP)
> },
> #endif /* CONFIG_ARM64_AMU_EXTN */
> @@ -2981,7 +2981,7 @@ static void update_cpu_capabilities(u16 scope_mask)
> !caps->matches(caps, cpucap_default_scope(caps)))
> continue;
>
> - if (caps->desc)
> + if (caps->desc && !caps->cpus)
> pr_info("detected: %s\n", caps->desc);
>
> __set_bit(caps->capability, system_cpucaps);
> @@ -3330,6 +3330,7 @@ unsigned long cpu_get_elf_hwcap2(void)
>
> static void __init setup_system_capabilities(void)
> {
> + int i;
> /*
> * We have finalised the system-wide safe feature
> * registers, finalise the capabilities that depend
> @@ -3338,6 +3339,15 @@ static void __init setup_system_capabilities(void)
> */
> update_cpu_capabilities(SCOPE_SYSTEM);
> enable_cpu_capabilities(SCOPE_ALL & ~SCOPE_BOOT_CPU);
> +
> + for (i = 0; i < ARM64_NCAPS; i++) {
> + const struct arm64_cpu_capabilities *caps = cpucap_ptrs[i];
> +
> + if (caps && caps->cpus && caps->desc &&
> + cpumask_any(caps->cpus) < nr_cpu_ids)
> + pr_info("detected: %s on CPU%*pbl\n",
> + caps->desc, cpumask_pr_args(caps->cpus));
> + }
> }
>
> void __init setup_cpu_features(void)
Thanks, Jeremy for putting this patch together. The duplicated prints
bothered me too every time I stared at the boot logs.
Fwiw,,
Reviewed-by: Punit Agrawal <[email protected]>
Tested-by: Punit Agrawal <[email protected]>
The patch was verified on an N2 based platform.