Re: [PATCH v7 06/10] RISC-V: Add cpu_ops and modify default booting method

From: Anup Patel
Date: Mon Jan 27 2020 - 23:31:24 EST


On Tue, Jan 28, 2020 at 7:58 AM Atish Patra <atish.patra@xxxxxxx> wrote:
>
> Currently, all non-booting harts start booting after the booting hart
> updates the per-hart stack pointer. This is done in a way that, it's
> difficult to implement any other booting method without breaking the
> backward compatibility.
>
> Define a cpu_ops method that allows to introduce other booting methods
> in future. Modify the current booting method to be compatible with
> cpu_ops.
>
> Signed-off-by: Atish Patra <atish.patra@xxxxxxx>
> ---
> arch/riscv/include/asm/cpu_ops.h | 31 ++++++++++++++++
> arch/riscv/kernel/Makefile | 1 +
> arch/riscv/kernel/cpu_ops.c | 61 ++++++++++++++++++++++++++++++++
> arch/riscv/kernel/setup.c | 4 ++-
> arch/riscv/kernel/smpboot.c | 52 ++++++++++++++++-----------
> 5 files changed, 127 insertions(+), 22 deletions(-)
> create mode 100644 arch/riscv/include/asm/cpu_ops.h
> create mode 100644 arch/riscv/kernel/cpu_ops.c

This has to be more modular considering the fact that CONFIG_RISCV_SBI
can be disabled.

I have few suggestions on how to break cpu_ops.c

>
> diff --git a/arch/riscv/include/asm/cpu_ops.h b/arch/riscv/include/asm/cpu_ops.h
> new file mode 100644
> index 000000000000..27e9dfee5460
> --- /dev/null
> +++ b/arch/riscv/include/asm/cpu_ops.h
> @@ -0,0 +1,31 @@
> +/* SPDX-License-Identifier: GPL-2.0-only */
> +/*
> + * Copyright (c) 2019 Western Digital Corporation or its affiliates.
> + * Based on arch/arm64/include/asm/cpu_ops.h
> + */
> +#ifndef __ASM_CPU_OPS_H
> +#define __ASM_CPU_OPS_H
> +
> +#include <linux/init.h>
> +#include <linux/threads.h>
> +
> +/**
> + * struct cpu_operations - Callback operations for hotplugging CPUs.
> + *
> + * @name: Name of the boot protocol.
> + * @cpu_prepare: Early one-time preparation step for a cpu. If there
> + * is a mechanism for doing so, tests whether it is
> + * possible to boot the given HART.
> + * @cpu_start: Boots a cpu into the kernel.
> + */
> +struct cpu_operations {
> + const char *name;
> + int (*cpu_prepare)(unsigned int cpu);
> + int (*cpu_start)(unsigned int cpu,
> + struct task_struct *tidle);
> +};
> +
> +extern const struct cpu_operations *cpu_ops[NR_CPUS];

Add following here:
extern void *__cpu_up_stack_pointer[NR_CPUS];
extern void *__cpu_up_task_pointer[NR_CPUS];

> +int __init cpu_set_ops(int cpu);
> +
> +#endif /* ifndef __ASM_CPU_OPS_H */
> diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
> index f40205cb9a22..d77def5b4e87 100644
> --- a/arch/riscv/kernel/Makefile
> +++ b/arch/riscv/kernel/Makefile
> @@ -32,6 +32,7 @@ obj-$(CONFIG_RISCV_M_MODE) += clint.o
> obj-$(CONFIG_FPU) += fpu.o
> obj-$(CONFIG_SMP) += smpboot.o
> obj-$(CONFIG_SMP) += smp.o
> +obj-$(CONFIG_SMP) += cpu_ops.o
> obj-$(CONFIG_MODULES) += module.o
> obj-$(CONFIG_MODULE_SECTIONS) += module-sections.o
>
> diff --git a/arch/riscv/kernel/cpu_ops.c b/arch/riscv/kernel/cpu_ops.c
> new file mode 100644
> index 000000000000..099dbb6ff9f0
> --- /dev/null
> +++ b/arch/riscv/kernel/cpu_ops.c
> @@ -0,0 +1,61 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * Copyright (c) 2019 Western Digital Corporation or its affiliates.
> + *
> + */
> +
> +#include <linux/errno.h>
> +#include <linux/mm.h>
> +#include <linux/of.h>
> +#include <linux/string.h>
> +#include <linux/sched/task_stack.h>
> +#include <asm/cpu_ops.h>
> +#include <asm/sbi.h>
> +#include <asm/smp.h>
> +
> +const struct cpu_operations *cpu_ops[NR_CPUS] __ro_after_init;
> +
> +void *__cpu_up_stack_pointer[NR_CPUS];
> +void *__cpu_up_task_pointer[NR_CPUS];
> +
> +const struct cpu_operations cpu_spinwait_ops;
> +
> +static int spinwait_cpu_prepare(unsigned int cpuid)
> +{
> + if (!cpu_spinwait_ops.cpu_start) {
> + pr_err("cpu start method not defined for CPU [%d]\n", cpuid);
> + return -ENODEV;
> + }
> + return 0;
> +}
> +
> +static int spinwait_cpu_start(unsigned int cpuid, struct task_struct *tidle)
> +{
> + int hartid = cpuid_to_hartid_map(cpuid);
> +
> + /*
> + * In this protocol, all cpus boot on their own accord. _start
> + * selects the first cpu to boot the kernel and causes the remainder
> + * of the cpus to spin in a loop waiting for their stack pointer to be
> + * setup by that main cpu. Writing __cpu_up_stack_pointer signals to
> + * the spinning cpus that they can continue the boot process.
> + */
> + smp_mb();
> + WRITE_ONCE(__cpu_up_stack_pointer[hartid],
> + task_stack_page(tidle) + THREAD_SIZE);
> + WRITE_ONCE(__cpu_up_task_pointer[hartid], tidle);
> +
> + return 0;
> +}
> +
> +const struct cpu_operations cpu_spinwait_ops = {
> + .name = "spinwait",
> + .cpu_prepare = spinwait_cpu_prepare,
> + .cpu_start = spinwait_cpu_start,
> +};

Move cpu_spinwait_ops, spinwait_cpu_start, and spinwait_cpu_prepare
to arch/riscv/kernel/cpu_ops_spinwait.c

Have "extern const struct cpu_operations cpu_spinwait_ops;" here.

> +
> +int __init cpu_set_ops(int cpuid)
> +{
> + cpu_ops[cpuid] = &cpu_spinwait_ops;
> + return 0;
> +}
> diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c
> index de3e65dae83a..8208d1109ddb 100644
> --- a/arch/riscv/kernel/setup.c
> +++ b/arch/riscv/kernel/setup.c
> @@ -16,12 +16,13 @@
> #include <linux/of_platform.h>
> #include <linux/sched/task.h>
> #include <linux/swiotlb.h>
> +#include <linux/smp.h>
>
> #include <asm/clint.h>
> +#include <asm/cpu_ops.h>
> #include <asm/setup.h>
> #include <asm/sections.h>
> #include <asm/pgtable.h>
> -#include <asm/smp.h>
> #include <asm/sbi.h>
> #include <asm/tlbflush.h>
> #include <asm/thread_info.h>
> @@ -79,6 +80,7 @@ void __init setup_arch(char **cmdline_p)
> sbi_init();
>
> #ifdef CONFIG_SMP
> + cpu_set_ops(0);
> setup_smp();
> #endif
>
> diff --git a/arch/riscv/kernel/smpboot.c b/arch/riscv/kernel/smpboot.c
> index 8bc01f0ca73b..f2cf541bc895 100644
> --- a/arch/riscv/kernel/smpboot.c
> +++ b/arch/riscv/kernel/smpboot.c
> @@ -25,6 +25,7 @@
> #include <linux/sched/task_stack.h>
> #include <linux/sched/mm.h>
> #include <asm/clint.h>
> +#include <asm/cpu_ops.h>
> #include <asm/irq.h>
> #include <asm/mmu_context.h>
> #include <asm/tlbflush.h>
> @@ -34,8 +35,6 @@
>
> #include "head.h"
>
> -void *__cpu_up_stack_pointer[NR_CPUS];
> -void *__cpu_up_task_pointer[NR_CPUS];
> static DECLARE_COMPLETION(cpu_running);
>
> void __init smp_prepare_boot_cpu(void)
> @@ -46,6 +45,7 @@ void __init smp_prepare_boot_cpu(void)
> void __init smp_prepare_cpus(unsigned int max_cpus)
> {
> int cpuid;
> + int ret;
>
> /* This covers non-smp usecase mandated by "nosmp" option */
> if (max_cpus == 0)
> @@ -54,6 +54,11 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
> for_each_possible_cpu(cpuid) {
> if (cpuid == smp_processor_id())
> continue;
> + if (cpu_ops[cpuid]->cpu_prepare) {
> + ret = cpu_ops[cpuid]->cpu_prepare(cpuid);
> + if (ret)
> + continue;
> + }
> set_cpu_present(cpuid, true);
> }
> }
> @@ -92,36 +97,41 @@ void __init setup_smp(void)
> cpuid, nr_cpu_ids);
>
> for (cpuid = 1; cpuid < nr_cpu_ids; cpuid++) {
> - if (cpuid_to_hartid_map(cpuid) != INVALID_HARTID)
> + if (cpuid_to_hartid_map(cpuid) != INVALID_HARTID) {
> + if (cpu_set_ops(cpuid)) {
> + cpuid_to_hartid_map(cpuid) = INVALID_HARTID;
> + continue;
> + }
> set_cpu_possible(cpuid, true);
> + }
> }
> }
>
> +int start_secondary_cpu(int cpu, struct task_struct *tidle)
> +{
> + if (cpu_ops[cpu]->cpu_start)
> + return cpu_ops[cpu]->cpu_start(cpu, tidle);
> +
> + return -EOPNOTSUPP;
> +}
> +
> int __cpu_up(unsigned int cpu, struct task_struct *tidle)
> {
> int ret = 0;
> - int hartid = cpuid_to_hartid_map(cpu);
> tidle->thread_info.cpu = cpu;
>
> - /*
> - * On RISC-V systems, all harts boot on their own accord. Our _start
> - * selects the first hart to boot the kernel and causes the remainder
> - * of the harts to spin in a loop waiting for their stack pointer to be
> - * setup by that main hart. Writing __cpu_up_stack_pointer signals to
> - * the spinning harts that they can continue the boot process.
> - */
> - smp_mb();
> - WRITE_ONCE(__cpu_up_stack_pointer[hartid],
> - task_stack_page(tidle) + THREAD_SIZE);
> - WRITE_ONCE(__cpu_up_task_pointer[hartid], tidle);
> -
> - lockdep_assert_held(&cpu_running);
> - wait_for_completion_timeout(&cpu_running,
> + ret = start_secondary_cpu(cpu, tidle);
> + if (!ret) {
> + lockdep_assert_held(&cpu_running);
> + wait_for_completion_timeout(&cpu_running,
> msecs_to_jiffies(1000));
>
> - if (!cpu_online(cpu)) {
> - pr_crit("CPU%u: failed to come online\n", cpu);
> - ret = -EIO;
> + if (!cpu_online(cpu)) {
> + pr_crit("CPU%u: failed to come online\n", cpu);
> + ret = -EIO;
> + }
> + } else {
> + pr_crit("CPU%u: failed to start\n", cpu);
> }
>
> return ret;
> --
> 2.24.0
>

Regards,
Anup