diff options
author | Mark Rutland <mark.rutland@arm.com> | 2016-11-03 23:23:11 +0300 |
---|---|---|
committer | Catalin Marinas <catalin.marinas@arm.com> | 2016-11-11 21:25:45 +0300 |
commit | 57c82954e77fa12c1023e87210d2ede77aaa0058 (patch) | |
tree | f038c648ae9e6352b1e7ba0590c2bea37cb56fc2 | |
parent | 580efaa7ccfb8c0790dce4396434f0e5ac8d86ee (diff) | |
download | linux-57c82954e77fa12c1023e87210d2ede77aaa0058.tar.xz |
arm64: make cpu number a percpu variable
In the absence of CONFIG_THREAD_INFO_IN_TASK, core code maintains
thread_info::cpu, and low-level architecture code can access this to
build raw_smp_processor_id(). With CONFIG_THREAD_INFO_IN_TASK, core code
maintains task_struct::cpu, which for reasons of hte header soup is not
accessible to low-level arch code.
Instead, we can maintain a percpu variable containing the cpu number.
For both the old and new implementation of raw_smp_processor_id(), we
read a syreg into a GPR, add an offset, and load the result. As the
offset is now larger, it may not be folded into the load, but otherwise
the assembly shouldn't change much.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: James Morse <james.morse@arm.com>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
-rw-r--r-- | arch/arm64/include/asm/smp.h | 11 | ||||
-rw-r--r-- | arch/arm64/kernel/smp.c | 5 |
2 files changed, 15 insertions, 1 deletions
diff --git a/arch/arm64/include/asm/smp.h b/arch/arm64/include/asm/smp.h index 022644704a93..968b08de820d 100644 --- a/arch/arm64/include/asm/smp.h +++ b/arch/arm64/include/asm/smp.h @@ -29,11 +29,20 @@ #ifndef __ASSEMBLY__ +#include <asm/percpu.h> + #include <linux/threads.h> #include <linux/cpumask.h> #include <linux/thread_info.h> -#define raw_smp_processor_id() (current_thread_info()->cpu) +DECLARE_PER_CPU_READ_MOSTLY(int, cpu_number); + +/* + * We don't use this_cpu_read(cpu_number) as that has implicit writes to + * preempt_count, and associated (compiler) barriers, that we'd like to avoid + * the expense of. If we're preemptible, the value can be stale at use anyway. + */ +#define raw_smp_processor_id() (*this_cpu_ptr(&cpu_number)) struct seq_file; diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c index f64401faa091..6f42c68e457f 100644 --- a/arch/arm64/kernel/smp.c +++ b/arch/arm64/kernel/smp.c @@ -58,6 +58,9 @@ #define CREATE_TRACE_POINTS #include <trace/events/ipi.h> +DEFINE_PER_CPU_READ_MOSTLY(int, cpu_number); +EXPORT_PER_CPU_SYMBOL(cpu_number); + /* * as from 2.5, kernels no longer have an init_tasks structure * so we need some other way of telling a new secondary core @@ -719,6 +722,8 @@ void __init smp_prepare_cpus(unsigned int max_cpus) */ for_each_possible_cpu(cpu) { + per_cpu(cpu_number, cpu) = cpu; + if (cpu == smp_processor_id()) continue; |