| Message ID | 1374921728-9007-1-git-send-email-hanjun.guo@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: > Power aware scheduling needs the cpu topology information to improve the > cpu scheduler decision making. It's not only power aware scheduling. The scheduler already uses topology and cache sharing when CONFIG_SCHED_MC and/or CONFIG_SCHED_SMT are enable. So you should also add these configs for arm64 so the scheduler can use it Vincent > > For ARM64, we can get the topology from the MPIDR register which defines the > the affinity of processors. > > This patch is mainly based on arch/arm/kernel/topology.c written by > Vincent Guittot, and replaced the topology array with per cpu variable. > > Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org> > --- > arch/arm64/Kconfig | 9 +++ > arch/arm64/include/asm/cputype.h | 11 ++++ > arch/arm64/include/asm/topology.h | 41 ++++++++++++ > arch/arm64/kernel/Makefile | 1 + > arch/arm64/kernel/smp.c | 6 ++ > arch/arm64/kernel/topology.c | 128 +++++++++++++++++++++++++++++++++++++ > 6 files changed, 196 insertions(+) > create mode 100644 arch/arm64/include/asm/topology.h > create mode 100644 arch/arm64/kernel/topology.c > > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig > index 9737e97..f0ce91b 100644 > --- a/arch/arm64/Kconfig > +++ b/arch/arm64/Kconfig > @@ -150,6 +150,15 @@ config SMP > > If you don't know what to do here, say N. > > +config ARM64_CPU_TOPOLOGY > + bool "Support cpu topology definition" > + depends on SMP && ARM64 > + default y > + help > + Support ARM64 cpu topology definition. The MPIDR register defines > + affinity between processors which is then used to describe the cpu > + topology of an ARM64 System. > + > config NR_CPUS > int "Maximum number of CPUs (2-32)" > range 2 32 > diff --git a/arch/arm64/include/asm/cputype.h b/arch/arm64/include/asm/cputype.h > index 5fe138e..68b55af 100644 > --- a/arch/arm64/include/asm/cputype.h > +++ b/arch/arm64/include/asm/cputype.h > @@ -30,6 +30,17 @@ > > #define MPIDR_HWID_BITMASK 0xff00ffffff > > +#define MPIDR_SMP_BITMASK (0x1 << 30) > +#define MPIDR_MT_BITMASK (0x1 << 24) > + > +#define MPIDR_LEVEL_BITS 8 > +#define MPIDR_LEVEL_MASK ((1 << MPIDR_LEVEL_BITS) - 1) > + > +#define MPIDR_AFFINITY_LEVEL_0(mpidr) ((mpidr) & MPIDR_LEVEL_MASK) > +#define MPIDR_AFFINITY_LEVEL_1(mpidr) ((mpidr >> 8) & MPIDR_LEVEL_MASK) > +#define MPIDR_AFFINITY_LEVEL_2(mpidr) ((mpidr >> 16) & MPIDR_LEVEL_MASK) > +#define MPIDR_AFFINITY_LEVEL_3(mpidr) ((mpidr >> 32) & MPIDR_LEVEL_MASK) > + > #define read_cpuid(reg) ({ \ > u64 __val; \ > asm("mrs %0, " reg : "=r" (__val)); \ > diff --git a/arch/arm64/include/asm/topology.h b/arch/arm64/include/asm/topology.h > new file mode 100644 > index 0000000..8631808 > --- /dev/null > +++ b/arch/arm64/include/asm/topology.h > @@ -0,0 +1,41 @@ > +#ifndef _ASM_ARM64_TOPOLOGY_H > +#define _ASM_ARM64_TOPOLOGY_H > + > +#ifdef CONFIG_ARM64_CPU_TOPOLOGY > + > +#include <linux/cpumask.h> > + > +struct cputopo_arm64 { > + int thread_id; > + int core_id; > + int socket_id; > + cpumask_t thread_sibling; > + cpumask_t core_sibling; > +}; > + > +DECLARE_PER_CPU(struct cputopo_arm64, cpu_topology); > + > +#define cpu_topo(cpu) per_cpu(cpu_topology, cpu) > + > +#define topology_physical_package_id(cpu) (cpu_topo(cpu).socket_id) > +#define topology_core_id(cpu) (cpu_topo(cpu).core_id) > +#define topology_core_cpumask(cpu) (&cpu_topo(cpu).core_sibling) > +#define topology_thread_cpumask(cpu) (&cpu_topo(cpu).thread_sibling) > + > +#define mc_capable() (cpu_topo(0).socket_id != -1) > +#define smt_capable() (cpu_topo(0).thread_id != -1) > + > +void init_cpu_topology(void); > +void store_cpu_topology(unsigned int cpuid); > +const struct cpumask *cpu_coregroup_mask(int cpu); > + > +#else > + > +static inline void init_cpu_topology(void) { } > +static inline void store_cpu_topology(unsigned int cpuid) { } > + > +#endif > + > +#include <asm-generic/topology.h> > + > +#endif /* _ASM_ARM64_TOPOLOGY_H */ > diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile > index 7b4b564..a47c359 100644 > --- a/arch/arm64/kernel/Makefile > +++ b/arch/arm64/kernel/Makefile > @@ -18,6 +18,7 @@ arm64-obj-$(CONFIG_SMP) += smp.o smp_spin_table.o smp_psci.o > arm64-obj-$(CONFIG_HW_PERF_EVENTS) += perf_event.o > arm64-obj-$(CONFIG_HAVE_HW_BREAKPOINT)+= hw_breakpoint.o > arm64-obj-$(CONFIG_EARLY_PRINTK) += early_printk.o > +arm64-obj-$(CONFIG_ARM64_CPU_TOPOLOGY) += topology.o > > obj-y += $(arm64-obj-y) vdso/ > obj-m += $(arm64-obj-m) > diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c > index fee5cce..197b1da 100644 > --- a/arch/arm64/kernel/smp.c > +++ b/arch/arm64/kernel/smp.c > @@ -39,6 +39,7 @@ > #include <asm/atomic.h> > #include <asm/cacheflush.h> > #include <asm/cputype.h> > +#include <asm/topology.h> > #include <asm/mmu_context.h> > #include <asm/pgtable.h> > #include <asm/pgalloc.h> > @@ -215,6 +216,8 @@ asmlinkage void secondary_start_kernel(void) > local_irq_enable(); > local_fiq_enable(); > > + store_cpu_topology(cpu); > + > /* > * OK, it's off to the idle thread for us > */ > @@ -387,6 +390,9 @@ void __init smp_prepare_cpus(unsigned int max_cpus) > int cpu, err; > unsigned int ncores = num_possible_cpus(); > > + init_cpu_topology(); > + store_cpu_topology(smp_processor_id()); > + > /* > * are we trying to boot more cores than exist? > */ > diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c > new file mode 100644 > index 0000000..1eb0435 > --- /dev/null > +++ b/arch/arm64/kernel/topology.c > @@ -0,0 +1,128 @@ > +/* > + * arch/arm64/kernel/topology.c > + * > + * Copyright (C) 2013 Linaro Limited. > + * Written by: Hanjun Guo > + * > + * based on arch/arm/kernel/topology.c > + * > + * This file is subject to the terms and conditions of the GNU General Public > + * License. See the file "COPYING" in the main directory of this archive > + * for more details. > + */ > + > +#include <linux/cpu.h> > +#include <linux/cpumask.h> > +#include <linux/init.h> > +#include <linux/percpu.h> > +#include <linux/sched.h> > + > +#include <asm/cputype.h> > +#include <asm/topology.h> > + > +DEFINE_PER_CPU(struct cputopo_arm64, cpu_topology); > + > +const struct cpumask *cpu_coregroup_mask(int cpu) > +{ > + return &cpu_topo(cpu).core_sibling; > +} > + > +void update_siblings_masks(unsigned int cpuid) > +{ > + struct cputopo_arm64 *topo, *cpuid_topo = &cpu_topo(cpuid); > + int cpu; > + > + /* update core and thread sibling masks */ > + for_each_possible_cpu(cpu) { > + topo = &cpu_topo(cpu); > + > + if (cpuid_topo->socket_id != topo->socket_id) > + continue; > + > + cpumask_set_cpu(cpuid, &topo->core_sibling); > + if (cpu != cpuid) > + cpumask_set_cpu(cpu, &cpuid_topo->core_sibling); > + > + if (cpuid_topo->core_id != topo->core_id) > + continue; > + > + cpumask_set_cpu(cpuid, &topo->thread_sibling); > + if (cpu != cpuid) > + cpumask_set_cpu(cpu, &cpuid_topo->thread_sibling); > + } > + smp_wmb(); > +} > + > +/* > + * store_cpu_topology is called at boot when only one cpu is running > + * and with the mutex cpu_hotplug.lock locked, when several cpus have booted, > + * which prevents simultaneous write access to cpu_topology array > + */ > +void store_cpu_topology(unsigned int cpuid) > +{ > + struct cputopo_arm64 *cpuid_topo = &cpu_topo(cpuid); > + u64 mpidr; > + > + /* If the cpu topology has been already set, just return */ > + if (cpuid_topo->core_id != -1) > + return; > + > + mpidr = read_cpuid_mpidr(); > + > + /* create cpu topology mapping */ > + if (!(mpidr & MPIDR_SMP_BITMASK)) { > + /* > + * This is a multiprocessor system > + * multiprocessor format & multiprocessor mode field are set > + */ > + > + if (mpidr & MPIDR_MT_BITMASK) { > + /* core performance interdependency */ > + cpuid_topo->thread_id = MPIDR_AFFINITY_LEVEL_0(mpidr); > + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL_1(mpidr); > + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL_2(mpidr); > + } else { > + /* largely independent cores */ > + cpuid_topo->thread_id = -1; > + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL_0(mpidr); > + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL_1(mpidr); > + } > + } else { > + /* > + * This is an uniprocessor system > + * we are in multiprocessor format but uniprocessor system > + * or in the old uniprocessor format > + */ > + cpuid_topo->thread_id = -1; > + cpuid_topo->core_id = 0; > + cpuid_topo->socket_id = -1; > + } > + > + update_siblings_masks(cpuid); > + > + printk(KERN_INFO "CPU%u: thread %d, cpu %d, socket %d, mpidr 0x%llx\n", > + cpuid, cpu_topo(cpuid).thread_id, > + cpu_topo(cpuid).core_id, > + cpu_topo(cpuid).socket_id, mpidr); > +} > + > +/* > + * init_cpu_topology is called at boot when only one cpu is running > + * which prevent simultaneous write access to cpu_topology array > + */ > +void __init init_cpu_topology(void) > +{ > + unsigned int cpu; > + > + /* init core mask */ > + for_each_possible_cpu(cpu) { > + struct cputopo_arm64 *topo = &cpu_topo(cpu); > + > + topo->thread_id = -1; > + topo->core_id = -1; > + topo->socket_id = -1; > + cpumask_clear(&topo->core_sibling); > + cpumask_clear(&topo->thread_sibling); > + } > + smp_wmb(); > +} > -- > 1.7.9.5 >
On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: > On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: > > Power aware scheduling needs the cpu topology information to improve the > > cpu scheduler decision making. > > It's not only power aware scheduling. The scheduler already uses > topology and cache sharing when CONFIG_SCHED_MC and/or > CONFIG_SCHED_SMT are enable. So you should also add these configs for > arm64 so the scheduler can use it ... except that the architecture doesn't define what the AFF fields in MPIDR really represent. Using them to make key scheduling decisions relating to cache proximity seems pretty risky to me, especially given the track record we've seen already on AArch32 silicon. It's a convenient register if it contains the data we want it to contain, but we need to force ourselves to come to terms with reality here and simply use it as an identifier for a CPU. Can't we just use the device-tree to represent this topological data for arm64? Lorenzo has been working on bindings in this area. Will
On 29/07/13 10:46, Vincent Guittot wrote: > On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: >> Power aware scheduling needs the cpu topology information to improve the >> cpu scheduler decision making. > > It's not only power aware scheduling. The scheduler already uses > topology and cache sharing when CONFIG_SCHED_MC and/or > CONFIG_SCHED_SMT are enable. So you should also add these configs for > arm64 so the scheduler can use it > Just for my knowledge, I thought power aware using SCHED_MC/SMT was removed. I see commit 8e7fbcbc22c12414bcc9dfdd683637f58fb32759 "sched: Remove stale power aware scheduling remnants and dysfunctional knobs" I may be missing something here. Regards, Sudeep
On 29 July 2013 12:15, Sudeep KarkadaNagesha <Sudeep.KarkadaNagesha@arm.com> wrote: > On 29/07/13 10:46, Vincent Guittot wrote: >> On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: >>> Power aware scheduling needs the cpu topology information to improve the >>> cpu scheduler decision making. >> >> It's not only power aware scheduling. The scheduler already uses >> topology and cache sharing when CONFIG_SCHED_MC and/or >> CONFIG_SCHED_SMT are enable. So you should also add these configs for >> arm64 so the scheduler can use it >> > Just for my knowledge, I thought power aware using SCHED_MC/SMT was > removed. I see commit 8e7fbcbc22c12414bcc9dfdd683637f58fb32759 "sched: > Remove stale power aware scheduling remnants and dysfunctional knobs" > I may be missing something here. It's a common mistake to mixed SCHED_MC and powersaving balance with SCHED_MC. Only the powersaving policy has been removed but the SCHED_MC and SCHED_SMT are always in the scheduler and gives perf improvement on arm 32bits > > Regards, > Sudeep >
On 29 July 2013 11:54, Will Deacon <will.deacon@arm.com> wrote: > On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: >> On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: >> > Power aware scheduling needs the cpu topology information to improve the >> > cpu scheduler decision making. >> >> It's not only power aware scheduling. The scheduler already uses >> topology and cache sharing when CONFIG_SCHED_MC and/or >> CONFIG_SCHED_SMT are enable. So you should also add these configs for >> arm64 so the scheduler can use it > > ... except that the architecture doesn't define what the AFF fields in MPIDR > really represent. Using them to make key scheduling decisions relating to Do you mean that it's not define for arm64 ARM? AFAIK, there are good explanation in the arm32 ARM and it's currently used with SCHED_MC and SCHED_SMT > cache proximity seems pretty risky to me, especially given the track record > we've seen already on AArch32 silicon. It's a convenient register if it > contains the data we want it to contain, but we need to force ourselves to > come to terms with reality here and simply use it as an identifier for a > CPU. > > Can't we just use the device-tree to represent this topological data for > arm64? Lorenzo has been working on bindings in this area. I agree that we should probably use DT if we can't rely in MPIDR for arm64 Vincent > > Will > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/
On Mon, Jul 29, 2013 at 10:54:01AM +0100, Will Deacon wrote: > On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: > > On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: > > > Power aware scheduling needs the cpu topology information to improve the > > > cpu scheduler decision making. > > > > It's not only power aware scheduling. The scheduler already uses > > topology and cache sharing when CONFIG_SCHED_MC and/or > > CONFIG_SCHED_SMT are enable. So you should also add these configs for > > arm64 so the scheduler can use it > > ... except that the architecture doesn't define what the AFF fields in MPIDR > really represent. Using them to make key scheduling decisions relating to In fact, the ARM Architecture doesn't place any requirements on MPIDRs to force the aff fields to exist _at all_. It's just a recommendation. Instead, you have a 24 or 32-bit number which is unique per CPU, and which is _probably_ assigned in a way resembling the aff fields. > cache proximity seems pretty risky to me, especially given the track record > we've seen already on AArch32 silicon. It's a convenient register if it > contains the data we want it to contain, but we need to force ourselves to > come to terms with reality here and simply use it as an identifier for a > CPU. +1 Also, we should align arm and arm64. The problem is basically exactly the same, and the solution needs to be the same. struct cputopo_arm is already being abused -- for example, TC2 describes the A15 and A7 clusters on a single die as having different "socket_id" values, even though this is obviously nonsense. But there's no other way to describe that system today. > Can't we just use the device-tree to represent this topological data for > arm64? Lorenzo has been working on bindings in this area. This may become more important as we start to see things like asymmetric topologies appearing (different numbers of nodes and different interdependence characteristics in adjacent branches of the topology etc.) Cheers ---Dave
On Mon, Jul 29, 2013 at 02:36:30PM +0100, Dave Martin wrote: > On Mon, Jul 29, 2013 at 10:54:01AM +0100, Will Deacon wrote: > > On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: > > > On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: > > > > Power aware scheduling needs the cpu topology information to improve the > > > > cpu scheduler decision making. > > > > > > It's not only power aware scheduling. The scheduler already uses > > > topology and cache sharing when CONFIG_SCHED_MC and/or > > > CONFIG_SCHED_SMT are enable. So you should also add these configs for > > > arm64 so the scheduler can use it > > > > ... except that the architecture doesn't define what the AFF fields in MPIDR > > really represent. Using them to make key scheduling decisions relating to > > In fact, the ARM Architecture doesn't place any requirements on MPIDRs to > force the aff fields to exist _at all_. It's just a recommendation. > Instead, you have a 24 or 32-bit number which is unique per CPU, and which > is _probably_ assigned in a way resembling the aff fields. > > > cache proximity seems pretty risky to me, especially given the track record > > we've seen already on AArch32 silicon. It's a convenient register if it > > contains the data we want it to contain, but we need to force ourselves to > > come to terms with reality here and simply use it as an identifier for a > > CPU. > > +1 > > Also, we should align arm and arm64. The problem is basically exactly > the same, and the solution needs to be the same. struct cputopo_arm is > already being abused -- for example, TC2 describes the A15 and A7 > clusters on a single die as having different "socket_id" values, even > though this is obviously nonsense. But there's no other way to describe > that system today. > > > Can't we just use the device-tree to represent this topological data for > > arm64? Lorenzo has been working on bindings in this area. > > This may become more important as we start to see things like asymmetric > topologies appearing (different numbers of nodes and different > interdependence characteristics in adjacent branches of the topology > etc.) Will and Dave summed up the existing issues with MPIDR definition related to the topology description. FYI, a link to the current topology bindings posted on DT-discuss and LAKML: https://lists.ozlabs.org/pipermail/devicetree-discuss/2013-April/031725.html I am waiting for the dust to settle on the DT bindings review discussions to repost them and get them finalized. Lorenzo
On 2013-7-29 17:46, Vincent Guittot wrote: > On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: >> Power aware scheduling needs the cpu topology information to improve the >> cpu scheduler decision making. > > It's not only power aware scheduling. The scheduler already uses > topology and cache sharing when CONFIG_SCHED_MC and/or > CONFIG_SCHED_SMT are enable. So you should also add these configs for > arm64 so the scheduler can use it Yes, you are right, thanks for the advice. > > Vincent > >> >> For ARM64, we can get the topology from the MPIDR register which defines the >> the affinity of processors. >> >> This patch is mainly based on arch/arm/kernel/topology.c written by >> Vincent Guittot, and replaced the topology array with per cpu variable. >> >> Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org> >> --- >> arch/arm64/Kconfig | 9 +++ >> arch/arm64/include/asm/cputype.h | 11 ++++ >> arch/arm64/include/asm/topology.h | 41 ++++++++++++ >> arch/arm64/kernel/Makefile | 1 + >> arch/arm64/kernel/smp.c | 6 ++ >> arch/arm64/kernel/topology.c | 128 +++++++++++++++++++++++++++++++++++++ >> 6 files changed, 196 insertions(+) >> create mode 100644 arch/arm64/include/asm/topology.h >> create mode 100644 arch/arm64/kernel/topology.c >> >> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig >> index 9737e97..f0ce91b 100644 >> --- a/arch/arm64/Kconfig >> +++ b/arch/arm64/Kconfig >> @@ -150,6 +150,15 @@ config SMP >> >> If you don't know what to do here, say N. >> >> +config ARM64_CPU_TOPOLOGY >> + bool "Support cpu topology definition" >> + depends on SMP && ARM64 >> + default y >> + help >> + Support ARM64 cpu topology definition. The MPIDR register defines >> + affinity between processors which is then used to describe the cpu >> + topology of an ARM64 System. >> + >> config NR_CPUS >> int "Maximum number of CPUs (2-32)" >> range 2 32 >> diff --git a/arch/arm64/include/asm/cputype.h b/arch/arm64/include/asm/cputype.h >> index 5fe138e..68b55af 100644 >> --- a/arch/arm64/include/asm/cputype.h >> +++ b/arch/arm64/include/asm/cputype.h >> @@ -30,6 +30,17 @@ >> >> #define MPIDR_HWID_BITMASK 0xff00ffffff >> >> +#define MPIDR_SMP_BITMASK (0x1 << 30) >> +#define MPIDR_MT_BITMASK (0x1 << 24) >> + >> +#define MPIDR_LEVEL_BITS 8 >> +#define MPIDR_LEVEL_MASK ((1 << MPIDR_LEVEL_BITS) - 1) >> + >> +#define MPIDR_AFFINITY_LEVEL_0(mpidr) ((mpidr) & MPIDR_LEVEL_MASK) >> +#define MPIDR_AFFINITY_LEVEL_1(mpidr) ((mpidr >> 8) & MPIDR_LEVEL_MASK) >> +#define MPIDR_AFFINITY_LEVEL_2(mpidr) ((mpidr >> 16) & MPIDR_LEVEL_MASK) >> +#define MPIDR_AFFINITY_LEVEL_3(mpidr) ((mpidr >> 32) & MPIDR_LEVEL_MASK) >> + >> #define read_cpuid(reg) ({ \ >> u64 __val; \ >> asm("mrs %0, " reg : "=r" (__val)); \ >> diff --git a/arch/arm64/include/asm/topology.h b/arch/arm64/include/asm/topology.h >> new file mode 100644 >> index 0000000..8631808 >> --- /dev/null >> +++ b/arch/arm64/include/asm/topology.h >> @@ -0,0 +1,41 @@ >> +#ifndef _ASM_ARM64_TOPOLOGY_H >> +#define _ASM_ARM64_TOPOLOGY_H >> + >> +#ifdef CONFIG_ARM64_CPU_TOPOLOGY >> + >> +#include <linux/cpumask.h> >> + >> +struct cputopo_arm64 { >> + int thread_id; >> + int core_id; >> + int socket_id; >> + cpumask_t thread_sibling; >> + cpumask_t core_sibling; >> +}; >> + >> +DECLARE_PER_CPU(struct cputopo_arm64, cpu_topology); >> + >> +#define cpu_topo(cpu) per_cpu(cpu_topology, cpu) >> + >> +#define topology_physical_package_id(cpu) (cpu_topo(cpu).socket_id) >> +#define topology_core_id(cpu) (cpu_topo(cpu).core_id) >> +#define topology_core_cpumask(cpu) (&cpu_topo(cpu).core_sibling) >> +#define topology_thread_cpumask(cpu) (&cpu_topo(cpu).thread_sibling) >> + >> +#define mc_capable() (cpu_topo(0).socket_id != -1) >> +#define smt_capable() (cpu_topo(0).thread_id != -1) >> + >> +void init_cpu_topology(void); >> +void store_cpu_topology(unsigned int cpuid); >> +const struct cpumask *cpu_coregroup_mask(int cpu); >> + >> +#else >> + >> +static inline void init_cpu_topology(void) { } >> +static inline void store_cpu_topology(unsigned int cpuid) { } >> + >> +#endif >> + >> +#include <asm-generic/topology.h> >> + >> +#endif /* _ASM_ARM64_TOPOLOGY_H */ >> diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile >> index 7b4b564..a47c359 100644 >> --- a/arch/arm64/kernel/Makefile >> +++ b/arch/arm64/kernel/Makefile >> @@ -18,6 +18,7 @@ arm64-obj-$(CONFIG_SMP) += smp.o smp_spin_table.o smp_psci.o >> arm64-obj-$(CONFIG_HW_PERF_EVENTS) += perf_event.o >> arm64-obj-$(CONFIG_HAVE_HW_BREAKPOINT)+= hw_breakpoint.o >> arm64-obj-$(CONFIG_EARLY_PRINTK) += early_printk.o >> +arm64-obj-$(CONFIG_ARM64_CPU_TOPOLOGY) += topology.o >> >> obj-y += $(arm64-obj-y) vdso/ >> obj-m += $(arm64-obj-m) >> diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c >> index fee5cce..197b1da 100644 >> --- a/arch/arm64/kernel/smp.c >> +++ b/arch/arm64/kernel/smp.c >> @@ -39,6 +39,7 @@ >> #include <asm/atomic.h> >> #include <asm/cacheflush.h> >> #include <asm/cputype.h> >> +#include <asm/topology.h> >> #include <asm/mmu_context.h> >> #include <asm/pgtable.h> >> #include <asm/pgalloc.h> >> @@ -215,6 +216,8 @@ asmlinkage void secondary_start_kernel(void) >> local_irq_enable(); >> local_fiq_enable(); >> >> + store_cpu_topology(cpu); >> + >> /* >> * OK, it's off to the idle thread for us >> */ >> @@ -387,6 +390,9 @@ void __init smp_prepare_cpus(unsigned int max_cpus) >> int cpu, err; >> unsigned int ncores = num_possible_cpus(); >> >> + init_cpu_topology(); >> + store_cpu_topology(smp_processor_id()); >> + >> /* >> * are we trying to boot more cores than exist? >> */ >> diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c >> new file mode 100644 >> index 0000000..1eb0435 >> --- /dev/null >> +++ b/arch/arm64/kernel/topology.c >> @@ -0,0 +1,128 @@ >> +/* >> + * arch/arm64/kernel/topology.c >> + * >> + * Copyright (C) 2013 Linaro Limited. >> + * Written by: Hanjun Guo >> + * >> + * based on arch/arm/kernel/topology.c >> + * >> + * This file is subject to the terms and conditions of the GNU General Public >> + * License. See the file "COPYING" in the main directory of this archive >> + * for more details. >> + */ >> + >> +#include <linux/cpu.h> >> +#include <linux/cpumask.h> >> +#include <linux/init.h> >> +#include <linux/percpu.h> >> +#include <linux/sched.h> >> + >> +#include <asm/cputype.h> >> +#include <asm/topology.h> >> + >> +DEFINE_PER_CPU(struct cputopo_arm64, cpu_topology); >> + >> +const struct cpumask *cpu_coregroup_mask(int cpu) >> +{ >> + return &cpu_topo(cpu).core_sibling; >> +} >> + >> +void update_siblings_masks(unsigned int cpuid) >> +{ >> + struct cputopo_arm64 *topo, *cpuid_topo = &cpu_topo(cpuid); >> + int cpu; >> + >> + /* update core and thread sibling masks */ >> + for_each_possible_cpu(cpu) { >> + topo = &cpu_topo(cpu); >> + >> + if (cpuid_topo->socket_id != topo->socket_id) >> + continue; >> + >> + cpumask_set_cpu(cpuid, &topo->core_sibling); >> + if (cpu != cpuid) >> + cpumask_set_cpu(cpu, &cpuid_topo->core_sibling); >> + >> + if (cpuid_topo->core_id != topo->core_id) >> + continue; >> + >> + cpumask_set_cpu(cpuid, &topo->thread_sibling); >> + if (cpu != cpuid) >> + cpumask_set_cpu(cpu, &cpuid_topo->thread_sibling); >> + } >> + smp_wmb(); >> +} >> + >> +/* >> + * store_cpu_topology is called at boot when only one cpu is running >> + * and with the mutex cpu_hotplug.lock locked, when several cpus have booted, >> + * which prevents simultaneous write access to cpu_topology array >> + */ >> +void store_cpu_topology(unsigned int cpuid) >> +{ >> + struct cputopo_arm64 *cpuid_topo = &cpu_topo(cpuid); >> + u64 mpidr; >> + >> + /* If the cpu topology has been already set, just return */ >> + if (cpuid_topo->core_id != -1) >> + return; >> + >> + mpidr = read_cpuid_mpidr(); >> + >> + /* create cpu topology mapping */ >> + if (!(mpidr & MPIDR_SMP_BITMASK)) { >> + /* >> + * This is a multiprocessor system >> + * multiprocessor format & multiprocessor mode field are set >> + */ >> + >> + if (mpidr & MPIDR_MT_BITMASK) { >> + /* core performance interdependency */ >> + cpuid_topo->thread_id = MPIDR_AFFINITY_LEVEL_0(mpidr); >> + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL_1(mpidr); >> + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL_2(mpidr); >> + } else { >> + /* largely independent cores */ >> + cpuid_topo->thread_id = -1; >> + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL_0(mpidr); >> + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL_1(mpidr); >> + } >> + } else { >> + /* >> + * This is an uniprocessor system >> + * we are in multiprocessor format but uniprocessor system >> + * or in the old uniprocessor format >> + */ >> + cpuid_topo->thread_id = -1; >> + cpuid_topo->core_id = 0; >> + cpuid_topo->socket_id = -1; >> + } >> + >> + update_siblings_masks(cpuid); >> + >> + printk(KERN_INFO "CPU%u: thread %d, cpu %d, socket %d, mpidr 0x%llx\n", >> + cpuid, cpu_topo(cpuid).thread_id, >> + cpu_topo(cpuid).core_id, >> + cpu_topo(cpuid).socket_id, mpidr); >> +} >> + >> +/* >> + * init_cpu_topology is called at boot when only one cpu is running >> + * which prevent simultaneous write access to cpu_topology array >> + */ >> +void __init init_cpu_topology(void) >> +{ >> + unsigned int cpu; >> + >> + /* init core mask */ >> + for_each_possible_cpu(cpu) { >> + struct cputopo_arm64 *topo = &cpu_topo(cpu); >> + >> + topo->thread_id = -1; >> + topo->core_id = -1; >> + topo->socket_id = -1; >> + cpumask_clear(&topo->core_sibling); >> + cpumask_clear(&topo->thread_sibling); >> + } >> + smp_wmb(); >> +} >> -- >> 1.7.9.5 >>
On 2013-7-29 21:36, Dave Martin wrote: > On Mon, Jul 29, 2013 at 10:54:01AM +0100, Will Deacon wrote: >> On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: >>> On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: >>>> Power aware scheduling needs the cpu topology information to improve the >>>> cpu scheduler decision making. >>> >>> It's not only power aware scheduling. The scheduler already uses >>> topology and cache sharing when CONFIG_SCHED_MC and/or >>> CONFIG_SCHED_SMT are enable. So you should also add these configs for >>> arm64 so the scheduler can use it >> >> ... except that the architecture doesn't define what the AFF fields in MPIDR >> really represent. Using them to make key scheduling decisions relating to > > In fact, the ARM Architecture doesn't place any requirements on MPIDRs to > force the aff fields to exist _at all_. It's just a recommendation. > Instead, you have a 24 or 32-bit number which is unique per CPU, and which > is _probably_ assigned in a way resembling the aff fields. > >> cache proximity seems pretty risky to me, especially given the track record >> we've seen already on AArch32 silicon. It's a convenient register if it >> contains the data we want it to contain, but we need to force ourselves to >> come to terms with reality here and simply use it as an identifier for a >> CPU. > > +1 > > Also, we should align arm and arm64. The problem is basically exactly > the same, and the solution needs to be the same. struct cputopo_arm is > already being abused -- for example, TC2 describes the A15 and A7 > clusters on a single die as having different "socket_id" values, even > though this is obviously nonsense. But there's no other way to describe > that system today. > >> Can't we just use the device-tree to represent this topological data for >> arm64? Lorenzo has been working on bindings in this area. > > This may become more important as we start to see things like asymmetric > topologies appearing (different numbers of nodes and different > interdependence characteristics in adjacent branches of the topology > etc.) Agreed. I would like to mention that the ACPI Static Resource Affinity Table (SRAT) stores topology information for all the processors and memory, describing the physical locations of the processors and memory in the system. ACPI will be another available solution for this. Thanks Hanjun
On Mon, Jul 29, 2013 at 10:54:01AM +0100, Will Deacon wrote: > On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: > > On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: > > > Power aware scheduling needs the cpu topology information to improve the > > > cpu scheduler decision making. > > > > It's not only power aware scheduling. The scheduler already uses > > topology and cache sharing when CONFIG_SCHED_MC and/or > > CONFIG_SCHED_SMT are enable. So you should also add these configs for > > arm64 so the scheduler can use it > > ... except that the architecture doesn't define what the AFF fields in MPIDR > really represent. Using them to make key scheduling decisions relating to > cache proximity seems pretty risky to me, especially given the track record > we've seen already on AArch32 silicon. It's a convenient register if it > contains the data we want it to contain, but we need to force ourselves to > come to terms with reality here and simply use it as an identifier for a > CPU. > > Can't we just use the device-tree to represent this topological data for > arm64? Lorenzo has been working on bindings in this area. Catching up on email after holiday - I agree with Will here, we should use DT for representing the topology (or ACPI) and not rely on the MPIDR value.
On 2013-8-14 19:27, Catalin Marinas wrote: > On Mon, Jul 29, 2013 at 10:54:01AM +0100, Will Deacon wrote: >> On Mon, Jul 29, 2013 at 10:46:06AM +0100, Vincent Guittot wrote: >>> On 27 July 2013 12:42, Hanjun Guo <hanjun.guo@linaro.org> wrote: >>>> Power aware scheduling needs the cpu topology information to improve the >>>> cpu scheduler decision making. >>> >>> It's not only power aware scheduling. The scheduler already uses >>> topology and cache sharing when CONFIG_SCHED_MC and/or >>> CONFIG_SCHED_SMT are enable. So you should also add these configs for >>> arm64 so the scheduler can use it >> >> ... except that the architecture doesn't define what the AFF fields in MPIDR >> really represent. Using them to make key scheduling decisions relating to >> cache proximity seems pretty risky to me, especially given the track record >> we've seen already on AArch32 silicon. It's a convenient register if it >> contains the data we want it to contain, but we need to force ourselves to >> come to terms with reality here and simply use it as an identifier for a >> CPU. >> >> Can't we just use the device-tree to represent this topological data for >> arm64? Lorenzo has been working on bindings in this area. > > Catching up on email after holiday - I agree with Will here, we should > use DT for representing the topology (or ACPI) and not rely on the MPIDR > value. > Ok, I'm working on the ACPI part now, Thanks for your comments. Regards Hanjun
diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig index 9737e97..f0ce91b 100644 --- a/arch/arm64/Kconfig +++ b/arch/arm64/Kconfig @@ -150,6 +150,15 @@ config SMP If you don't know what to do here, say N. +config ARM64_CPU_TOPOLOGY + bool "Support cpu topology definition" + depends on SMP && ARM64 + default y + help + Support ARM64 cpu topology definition. The MPIDR register defines + affinity between processors which is then used to describe the cpu + topology of an ARM64 System. + config NR_CPUS int "Maximum number of CPUs (2-32)" range 2 32 diff --git a/arch/arm64/include/asm/cputype.h b/arch/arm64/include/asm/cputype.h index 5fe138e..68b55af 100644 --- a/arch/arm64/include/asm/cputype.h +++ b/arch/arm64/include/asm/cputype.h @@ -30,6 +30,17 @@ #define MPIDR_HWID_BITMASK 0xff00ffffff +#define MPIDR_SMP_BITMASK (0x1 << 30) +#define MPIDR_MT_BITMASK (0x1 << 24) + +#define MPIDR_LEVEL_BITS 8 +#define MPIDR_LEVEL_MASK ((1 << MPIDR_LEVEL_BITS) - 1) + +#define MPIDR_AFFINITY_LEVEL_0(mpidr) ((mpidr) & MPIDR_LEVEL_MASK) +#define MPIDR_AFFINITY_LEVEL_1(mpidr) ((mpidr >> 8) & MPIDR_LEVEL_MASK) +#define MPIDR_AFFINITY_LEVEL_2(mpidr) ((mpidr >> 16) & MPIDR_LEVEL_MASK) +#define MPIDR_AFFINITY_LEVEL_3(mpidr) ((mpidr >> 32) & MPIDR_LEVEL_MASK) + #define read_cpuid(reg) ({ \ u64 __val; \ asm("mrs %0, " reg : "=r" (__val)); \ diff --git a/arch/arm64/include/asm/topology.h b/arch/arm64/include/asm/topology.h new file mode 100644 index 0000000..8631808 --- /dev/null +++ b/arch/arm64/include/asm/topology.h @@ -0,0 +1,41 @@ +#ifndef _ASM_ARM64_TOPOLOGY_H +#define _ASM_ARM64_TOPOLOGY_H + +#ifdef CONFIG_ARM64_CPU_TOPOLOGY + +#include <linux/cpumask.h> + +struct cputopo_arm64 { + int thread_id; + int core_id; + int socket_id; + cpumask_t thread_sibling; + cpumask_t core_sibling; +}; + +DECLARE_PER_CPU(struct cputopo_arm64, cpu_topology); + +#define cpu_topo(cpu) per_cpu(cpu_topology, cpu) + +#define topology_physical_package_id(cpu) (cpu_topo(cpu).socket_id) +#define topology_core_id(cpu) (cpu_topo(cpu).core_id) +#define topology_core_cpumask(cpu) (&cpu_topo(cpu).core_sibling) +#define topology_thread_cpumask(cpu) (&cpu_topo(cpu).thread_sibling) + +#define mc_capable() (cpu_topo(0).socket_id != -1) +#define smt_capable() (cpu_topo(0).thread_id != -1) + +void init_cpu_topology(void); +void store_cpu_topology(unsigned int cpuid); +const struct cpumask *cpu_coregroup_mask(int cpu); + +#else + +static inline void init_cpu_topology(void) { } +static inline void store_cpu_topology(unsigned int cpuid) { } + +#endif + +#include <asm-generic/topology.h> + +#endif /* _ASM_ARM64_TOPOLOGY_H */ diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile index 7b4b564..a47c359 100644 --- a/arch/arm64/kernel/Makefile +++ b/arch/arm64/kernel/Makefile @@ -18,6 +18,7 @@ arm64-obj-$(CONFIG_SMP) += smp.o smp_spin_table.o smp_psci.o arm64-obj-$(CONFIG_HW_PERF_EVENTS) += perf_event.o arm64-obj-$(CONFIG_HAVE_HW_BREAKPOINT)+= hw_breakpoint.o arm64-obj-$(CONFIG_EARLY_PRINTK) += early_printk.o +arm64-obj-$(CONFIG_ARM64_CPU_TOPOLOGY) += topology.o obj-y += $(arm64-obj-y) vdso/ obj-m += $(arm64-obj-m) diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c index fee5cce..197b1da 100644 --- a/arch/arm64/kernel/smp.c +++ b/arch/arm64/kernel/smp.c @@ -39,6 +39,7 @@ #include <asm/atomic.h> #include <asm/cacheflush.h> #include <asm/cputype.h> +#include <asm/topology.h> #include <asm/mmu_context.h> #include <asm/pgtable.h> #include <asm/pgalloc.h> @@ -215,6 +216,8 @@ asmlinkage void secondary_start_kernel(void) local_irq_enable(); local_fiq_enable(); + store_cpu_topology(cpu); + /* * OK, it's off to the idle thread for us */ @@ -387,6 +390,9 @@ void __init smp_prepare_cpus(unsigned int max_cpus) int cpu, err; unsigned int ncores = num_possible_cpus(); + init_cpu_topology(); + store_cpu_topology(smp_processor_id()); + /* * are we trying to boot more cores than exist? */ diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c new file mode 100644 index 0000000..1eb0435 --- /dev/null +++ b/arch/arm64/kernel/topology.c @@ -0,0 +1,128 @@ +/* + * arch/arm64/kernel/topology.c + * + * Copyright (C) 2013 Linaro Limited. + * Written by: Hanjun Guo + * + * based on arch/arm/kernel/topology.c + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + */ + +#include <linux/cpu.h> +#include <linux/cpumask.h> +#include <linux/init.h> +#include <linux/percpu.h> +#include <linux/sched.h> + +#include <asm/cputype.h> +#include <asm/topology.h> + +DEFINE_PER_CPU(struct cputopo_arm64, cpu_topology); + +const struct cpumask *cpu_coregroup_mask(int cpu) +{ + return &cpu_topo(cpu).core_sibling; +} + +void update_siblings_masks(unsigned int cpuid) +{ + struct cputopo_arm64 *topo, *cpuid_topo = &cpu_topo(cpuid); + int cpu; + + /* update core and thread sibling masks */ + for_each_possible_cpu(cpu) { + topo = &cpu_topo(cpu); + + if (cpuid_topo->socket_id != topo->socket_id) + continue; + + cpumask_set_cpu(cpuid, &topo->core_sibling); + if (cpu != cpuid) + cpumask_set_cpu(cpu, &cpuid_topo->core_sibling); + + if (cpuid_topo->core_id != topo->core_id) + continue; + + cpumask_set_cpu(cpuid, &topo->thread_sibling); + if (cpu != cpuid) + cpumask_set_cpu(cpu, &cpuid_topo->thread_sibling); + } + smp_wmb(); +} + +/* + * store_cpu_topology is called at boot when only one cpu is running + * and with the mutex cpu_hotplug.lock locked, when several cpus have booted, + * which prevents simultaneous write access to cpu_topology array + */ +void store_cpu_topology(unsigned int cpuid) +{ + struct cputopo_arm64 *cpuid_topo = &cpu_topo(cpuid); + u64 mpidr; + + /* If the cpu topology has been already set, just return */ + if (cpuid_topo->core_id != -1) + return; + + mpidr = read_cpuid_mpidr(); + + /* create cpu topology mapping */ + if (!(mpidr & MPIDR_SMP_BITMASK)) { + /* + * This is a multiprocessor system + * multiprocessor format & multiprocessor mode field are set + */ + + if (mpidr & MPIDR_MT_BITMASK) { + /* core performance interdependency */ + cpuid_topo->thread_id = MPIDR_AFFINITY_LEVEL_0(mpidr); + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL_1(mpidr); + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL_2(mpidr); + } else { + /* largely independent cores */ + cpuid_topo->thread_id = -1; + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL_0(mpidr); + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL_1(mpidr); + } + } else { + /* + * This is an uniprocessor system + * we are in multiprocessor format but uniprocessor system + * or in the old uniprocessor format + */ + cpuid_topo->thread_id = -1; + cpuid_topo->core_id = 0; + cpuid_topo->socket_id = -1; + } + + update_siblings_masks(cpuid); + + printk(KERN_INFO "CPU%u: thread %d, cpu %d, socket %d, mpidr 0x%llx\n", + cpuid, cpu_topo(cpuid).thread_id, + cpu_topo(cpuid).core_id, + cpu_topo(cpuid).socket_id, mpidr); +} + +/* + * init_cpu_topology is called at boot when only one cpu is running + * which prevent simultaneous write access to cpu_topology array + */ +void __init init_cpu_topology(void) +{ + unsigned int cpu; + + /* init core mask */ + for_each_possible_cpu(cpu) { + struct cputopo_arm64 *topo = &cpu_topo(cpu); + + topo->thread_id = -1; + topo->core_id = -1; + topo->socket_id = -1; + cpumask_clear(&topo->core_sibling); + cpumask_clear(&topo->thread_sibling); + } + smp_wmb(); +}
Power aware scheduling needs the cpu topology information to improve the cpu scheduler decision making. For ARM64, we can get the topology from the MPIDR register which defines the the affinity of processors. This patch is mainly based on arch/arm/kernel/topology.c written by Vincent Guittot, and replaced the topology array with per cpu variable. Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org> --- arch/arm64/Kconfig | 9 +++ arch/arm64/include/asm/cputype.h | 11 ++++ arch/arm64/include/asm/topology.h | 41 ++++++++++++ arch/arm64/kernel/Makefile | 1 + arch/arm64/kernel/smp.c | 6 ++ arch/arm64/kernel/topology.c | 128 +++++++++++++++++++++++++++++++++++++ 6 files changed, 196 insertions(+) create mode 100644 arch/arm64/include/asm/topology.h create mode 100644 arch/arm64/kernel/topology.c