| Message ID | 20240502095115.177713-3-Dhananjay.Ugwekar@amd.com (mailing list archive) |
|---|---|
| State | In Next |
| Commit | 5d4d0283c0546559600dee7e9a4d87e402f3f4d9 |
| Headers | show |
| Series | Fix energy-pkg event for AMD CPUs | expand |
* Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com> wrote: > After commit ("x86/cpu/topology: Add support for the AMD 0x80000026 leaf"), > on AMD processors that support extended CPUID leaf 0x80000026, the > topology_die_cpumask() and topology_logical_die_id() macros, no longer > return the package cpumask and package id, instead they return the CCD > (Core Complex Die) mask and id respectively. This leads to the energy-pkg > event scope to be modified to CCD instead of package. > > Replacing these macros with their package counterparts fixes the > energy-pkg event for AMD CPUs. > > However due to the difference between the scope of energy-pkg event for > Intel and AMD CPUs, we have to replace these macros conditionally only for > AMD CPUs. > > On a 12 CCD 1 Package AMD Zen4 Genoa machine: > > Before: > $ cat /sys/devices/power/cpumask > 0,8,16,24,32,40,48,56,64,72,80,88. > > The expected cpumask here is supposed to be just "0", as it is a package > scope event, only one CPU will be collecting the event for all the CPUs in > the package. > > After: > $ cat /sys/devices/power/cpumask > 0 > > Signed-off-by: Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com> > Fixes: 63edbaa48a57 ("x86/cpu/topology: Add support for the AMD 0x80000026 leaf") > --- > arch/x86/events/rapl.c | 30 ++++++++++++++++++++++++++---- > 1 file changed, 26 insertions(+), 4 deletions(-) > > diff --git a/arch/x86/events/rapl.c b/arch/x86/events/rapl.c > index 46e673585560..d77bf7959a43 100644 > --- a/arch/x86/events/rapl.c > +++ b/arch/x86/events/rapl.c > @@ -102,6 +102,10 @@ static struct perf_pmu_events_attr event_attr_##v = { \ > .event_str = str, \ > }; > > +#define rapl_pmu_is_pkg_scope() \ > + (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || \ > + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) > + > struct rapl_pmu { > raw_spinlock_t lock; > int n_active; > @@ -139,9 +143,21 @@ static unsigned int rapl_cntr_mask; > static u64 rapl_timer_ms; > static struct perf_msr *rapl_msrs; > > +static inline unsigned int get_rapl_pmu_idx(int cpu) > +{ > + return rapl_pmu_is_pkg_scope() ? topology_logical_package_id(cpu) : > + topology_logical_die_id(cpu); > +} > + > +static inline cpumask_t *get_rapl_pmu_cpumask(int cpu) > +{ > + return rapl_pmu_is_pkg_scope() ? topology_core_cpumask(cpu) : > + topology_die_cpumask(cpu); > +} Note that this breaks the build with this config: https://download.01.org/0day-ci/archive/20240503/202405030828.RgFuznL9-lkp@intel.com/config I've removed the commit from perf/core for now. Thanks, Ingo
Hi Ingo, On 5/3/2024 3:24 PM, Ingo Molnar wrote: > > * Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com> wrote: > >> After commit ("x86/cpu/topology: Add support for the AMD 0x80000026 leaf"), >> on AMD processors that support extended CPUID leaf 0x80000026, the >> topology_die_cpumask() and topology_logical_die_id() macros, no longer >> return the package cpumask and package id, instead they return the CCD >> (Core Complex Die) mask and id respectively. This leads to the energy-pkg >> event scope to be modified to CCD instead of package. >> >> Replacing these macros with their package counterparts fixes the >> energy-pkg event for AMD CPUs. >> >> However due to the difference between the scope of energy-pkg event for >> Intel and AMD CPUs, we have to replace these macros conditionally only for >> AMD CPUs. >> >> On a 12 CCD 1 Package AMD Zen4 Genoa machine: >> >> Before: >> $ cat /sys/devices/power/cpumask >> 0,8,16,24,32,40,48,56,64,72,80,88. >> >> The expected cpumask here is supposed to be just "0", as it is a package >> scope event, only one CPU will be collecting the event for all the CPUs in >> the package. >> >> After: >> $ cat /sys/devices/power/cpumask >> 0 >> >> Signed-off-by: Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com> >> Fixes: 63edbaa48a57 ("x86/cpu/topology: Add support for the AMD 0x80000026 leaf") >> --- >> arch/x86/events/rapl.c | 30 ++++++++++++++++++++++++++---- >> 1 file changed, 26 insertions(+), 4 deletions(-) >> >> diff --git a/arch/x86/events/rapl.c b/arch/x86/events/rapl.c >> index 46e673585560..d77bf7959a43 100644 >> --- a/arch/x86/events/rapl.c >> +++ b/arch/x86/events/rapl.c >> @@ -102,6 +102,10 @@ static struct perf_pmu_events_attr event_attr_##v = { \ >> .event_str = str, \ >> }; >> >> +#define rapl_pmu_is_pkg_scope() \ >> + (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || \ >> + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) >> + >> struct rapl_pmu { >> raw_spinlock_t lock; >> int n_active; >> @@ -139,9 +143,21 @@ static unsigned int rapl_cntr_mask; >> static u64 rapl_timer_ms; >> static struct perf_msr *rapl_msrs; >> >> +static inline unsigned int get_rapl_pmu_idx(int cpu) >> +{ >> + return rapl_pmu_is_pkg_scope() ? topology_logical_package_id(cpu) : >> + topology_logical_die_id(cpu); >> +} >> + >> +static inline cpumask_t *get_rapl_pmu_cpumask(int cpu) >> +{ >> + return rapl_pmu_is_pkg_scope() ? topology_core_cpumask(cpu) : >> + topology_die_cpumask(cpu); >> +} > > Note that this breaks the build with this config: > > https://download.01.org/0day-ci/archive/20240503/202405030828.RgFuznL9-lkp@intel.com/config > > I've removed the commit from perf/core for now. Sure, will fix this in v2. Regards, Dhananjay > > Thanks, > > Ingo
diff --git a/arch/x86/events/rapl.c b/arch/x86/events/rapl.c index 46e673585560..d77bf7959a43 100644 --- a/arch/x86/events/rapl.c +++ b/arch/x86/events/rapl.c @@ -102,6 +102,10 @@ static struct perf_pmu_events_attr event_attr_##v = { \ .event_str = str, \ }; +#define rapl_pmu_is_pkg_scope() \ + (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || \ + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) + struct rapl_pmu { raw_spinlock_t lock; int n_active; @@ -139,9 +143,21 @@ static unsigned int rapl_cntr_mask; static u64 rapl_timer_ms; static struct perf_msr *rapl_msrs; +static inline unsigned int get_rapl_pmu_idx(int cpu) +{ + return rapl_pmu_is_pkg_scope() ? topology_logical_package_id(cpu) : + topology_logical_die_id(cpu); +} + +static inline cpumask_t *get_rapl_pmu_cpumask(int cpu) +{ + return rapl_pmu_is_pkg_scope() ? topology_core_cpumask(cpu) : + topology_die_cpumask(cpu); +} + static inline struct rapl_pmu *cpu_to_rapl_pmu(unsigned int cpu) { - unsigned int rapl_pmu_idx = topology_logical_die_id(cpu); + unsigned int rapl_pmu_idx = get_rapl_pmu_idx(cpu); /* * The unsigned check also catches the '-1' return value for non @@ -542,6 +558,7 @@ static struct perf_msr amd_rapl_msrs[] = { static int rapl_cpu_offline(unsigned int cpu) { + cpumask_t *cpumask = get_rapl_pmu_cpumask(cpu); struct rapl_pmu *pmu = cpu_to_rapl_pmu(cpu); int target; @@ -551,7 +568,7 @@ static int rapl_cpu_offline(unsigned int cpu) pmu->cpu = -1; /* Find a new cpu to collect rapl events */ - target = cpumask_any_but(topology_die_cpumask(cpu), cpu); + target = cpumask_any_but(cpumask, cpu); /* Migrate rapl events to the new target */ if (target < nr_cpu_ids) { @@ -564,6 +581,8 @@ static int rapl_cpu_offline(unsigned int cpu) static int rapl_cpu_online(unsigned int cpu) { + unsigned int rapl_pmu_idx = get_rapl_pmu_idx(cpu); + cpumask_t *cpumask = get_rapl_pmu_cpumask(cpu); struct rapl_pmu *pmu = cpu_to_rapl_pmu(cpu); int target; @@ -578,14 +597,14 @@ static int rapl_cpu_online(unsigned int cpu) pmu->timer_interval = ms_to_ktime(rapl_timer_ms); rapl_hrtimer_init(pmu); - rapl_pmus->pmus[topology_logical_die_id(cpu)] = pmu; + rapl_pmus->pmus[rapl_pmu_idx] = pmu; } /* * Check if there is an online cpu in the package which collects rapl * events already. */ - target = cpumask_any_and(&rapl_cpu_mask, topology_die_cpumask(cpu)); + target = cpumask_any_and(&rapl_cpu_mask, cpumask); if (target < nr_cpu_ids) return 0; @@ -676,6 +695,9 @@ static int __init init_rapl_pmus(void) { int nr_rapl_pmu = topology_max_packages() * topology_max_dies_per_package(); + if (rapl_pmu_is_pkg_scope()) + nr_rapl_pmu = topology_max_packages(); + rapl_pmus = kzalloc(struct_size(rapl_pmus, pmus, nr_rapl_pmu), GFP_KERNEL); if (!rapl_pmus) return -ENOMEM;
After commit ("x86/cpu/topology: Add support for the AMD 0x80000026 leaf"), on AMD processors that support extended CPUID leaf 0x80000026, the topology_die_cpumask() and topology_logical_die_id() macros, no longer return the package cpumask and package id, instead they return the CCD (Core Complex Die) mask and id respectively. This leads to the energy-pkg event scope to be modified to CCD instead of package. Replacing these macros with their package counterparts fixes the energy-pkg event for AMD CPUs. However due to the difference between the scope of energy-pkg event for Intel and AMD CPUs, we have to replace these macros conditionally only for AMD CPUs. On a 12 CCD 1 Package AMD Zen4 Genoa machine: Before: $ cat /sys/devices/power/cpumask 0,8,16,24,32,40,48,56,64,72,80,88. The expected cpumask here is supposed to be just "0", as it is a package scope event, only one CPU will be collecting the event for all the CPUs in the package. After: $ cat /sys/devices/power/cpumask 0 Signed-off-by: Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com> Fixes: 63edbaa48a57 ("x86/cpu/topology: Add support for the AMD 0x80000026 leaf") --- arch/x86/events/rapl.c | 30 ++++++++++++++++++++++++++---- 1 file changed, 26 insertions(+), 4 deletions(-)