smp.c

linux 内核源代码

	boot_cpu_addr = S390_lowcore.cpu_data.cpu_addr;
	current_thread_info()->cpu = 0;
	num_cpus = 1;
	for (cpu = 0; cpu <= 65535; cpu++) {
		if ((__u16) cpu == boot_cpu_addr)
			continue;
		__cpu_logical_map[1] = (__u16) cpu;
		if (signal_processor(1, sigp_sense) == sigp_not_operational)
			continue;
		smp_get_save_area(num_cpus, cpu);
		num_cpus++;
	}
	printk("Detected %d CPU's\n", (int) num_cpus);
	printk("Boot cpu address %2X\n", boot_cpu_addr);
	return num_cpus;
}

/*
 *	Activate a secondary processor.
 */
int __cpuinit start_secondary(void *cpuvoid)
{
	/* Setup the cpu */
	cpu_init();
	preempt_disable();
	/* Enable TOD clock interrupts on the secondary cpu. */
	init_cpu_timer();
#ifdef CONFIG_VIRT_TIMER
	/* Enable cpu timer interrupts on the secondary cpu. */
	init_cpu_vtimer();
#endif
	/* Enable pfault pseudo page faults on this cpu. */
	pfault_init();

	/* Mark this cpu as online */
	cpu_set(smp_processor_id(), cpu_online_map);
	/* Switch on interrupts */
	local_irq_enable();
	/* Print info about this processor */
	print_cpu_info(&S390_lowcore.cpu_data);
	/* cpu_idle will call schedule for us */
	cpu_idle();
	return 0;
}

DEFINE_PER_CPU(struct s390_idle_data, s390_idle);

static void __init smp_create_idle(unsigned int cpu)
{
	struct task_struct *p;

	/*
	 *  don't care about the psw and regs settings since we'll never
	 *  reschedule the forked task.
	 */
	p = fork_idle(cpu);
	if (IS_ERR(p))
		panic("failed fork for CPU %u: %li", cpu, PTR_ERR(p));
	current_set[cpu] = p;
	spin_lock_init(&(&per_cpu(s390_idle, cpu))->lock);
}

static int cpu_stopped(int cpu)
{
	__u32 status;

	/* Check for stopped state */
	if (signal_processor_ps(&status, 0, cpu, sigp_sense) ==
	    sigp_status_stored) {
		if (status & 0x40)
			return 1;
	}
	return 0;
}

/* Upping and downing of CPUs */

int __cpu_up(unsigned int cpu)
{
	struct task_struct *idle;
	struct _lowcore *cpu_lowcore;
	struct stack_frame *sf;
	sigp_ccode ccode;
	int curr_cpu;

	for (curr_cpu = 0; curr_cpu <= 65535; curr_cpu++) {
		__cpu_logical_map[cpu] = (__u16) curr_cpu;
		if (cpu_stopped(cpu))
			break;
	}

	if (!cpu_stopped(cpu))
		return -ENODEV;

	ccode = signal_processor_p((__u32)(unsigned long)(lowcore_ptr[cpu]),
				   cpu, sigp_set_prefix);
	if (ccode) {
		printk("sigp_set_prefix failed for cpu %d "
		       "with condition code %d\n",
		       (int) cpu, (int) ccode);
		return -EIO;
	}

	idle = current_set[cpu];
	cpu_lowcore = lowcore_ptr[cpu];
	cpu_lowcore->kernel_stack = (unsigned long)
		task_stack_page(idle) + THREAD_SIZE;
	sf = (struct stack_frame *) (cpu_lowcore->kernel_stack
				     - sizeof(struct pt_regs)
				     - sizeof(struct stack_frame));
	memset(sf, 0, sizeof(struct stack_frame));
	sf->gprs[9] = (unsigned long) sf;
	cpu_lowcore->save_area[15] = (unsigned long) sf;
	__ctl_store(cpu_lowcore->cregs_save_area[0], 0, 15);
	asm volatile(
		"	stam	0,15,0(%0)"
		: : "a" (&cpu_lowcore->access_regs_save_area) : "memory");
	cpu_lowcore->percpu_offset = __per_cpu_offset[cpu];
	cpu_lowcore->current_task = (unsigned long) idle;
	cpu_lowcore->cpu_data.cpu_nr = cpu;
	eieio();

	while (signal_processor(cpu, sigp_restart) == sigp_busy)
		udelay(10);

	while (!cpu_online(cpu))
		cpu_relax();
	return 0;
}

static unsigned int __initdata additional_cpus;
static unsigned int __initdata possible_cpus;

void __init smp_setup_cpu_possible_map(void)
{
	unsigned int phy_cpus, pos_cpus, cpu;

	phy_cpus = smp_count_cpus();
	pos_cpus = min(phy_cpus + additional_cpus, (unsigned int) NR_CPUS);

	if (possible_cpus)
		pos_cpus = min(possible_cpus, (unsigned int) NR_CPUS);

	for (cpu = 0; cpu < pos_cpus; cpu++)
		cpu_set(cpu, cpu_possible_map);

	phy_cpus = min(phy_cpus, pos_cpus);

	for (cpu = 0; cpu < phy_cpus; cpu++)
		cpu_set(cpu, cpu_present_map);
}

#ifdef CONFIG_HOTPLUG_CPU

static int __init setup_additional_cpus(char *s)
{
	additional_cpus = simple_strtoul(s, NULL, 0);
	return 0;
}
early_param("additional_cpus", setup_additional_cpus);

static int __init setup_possible_cpus(char *s)
{
	possible_cpus = simple_strtoul(s, NULL, 0);
	return 0;
}
early_param("possible_cpus", setup_possible_cpus);

int __cpu_disable(void)
{
	struct ec_creg_mask_parms cr_parms;
	int cpu = smp_processor_id();

	cpu_clear(cpu, cpu_online_map);

	/* Disable pfault pseudo page faults on this cpu. */
	pfault_fini();

	memset(&cr_parms.orvals, 0, sizeof(cr_parms.orvals));
	memset(&cr_parms.andvals, 0xff, sizeof(cr_parms.andvals));

	/* disable all external interrupts */
	cr_parms.orvals[0] = 0;
	cr_parms.andvals[0] = ~(1 << 15 | 1 << 14 | 1 << 13 | 1 << 12 |
				1 << 11 | 1 << 10 | 1 <<  6 | 1 <<  4);
	/* disable all I/O interrupts */
	cr_parms.orvals[6] = 0;
	cr_parms.andvals[6] = ~(1 << 31 | 1 << 30 | 1 << 29 | 1 << 28 |
				1 << 27 | 1 << 26 | 1 << 25 | 1 << 24);
	/* disable most machine checks */
	cr_parms.orvals[14] = 0;
	cr_parms.andvals[14] = ~(1 << 28 | 1 << 27 | 1 << 26 |
				 1 << 25 | 1 << 24);

	smp_ctl_bit_callback(&cr_parms);

	return 0;
}

void __cpu_die(unsigned int cpu)
{
	/* Wait until target cpu is down */
	while (!smp_cpu_not_running(cpu))
		cpu_relax();
	printk("Processor %d spun down\n", cpu);
}

void cpu_die(void)
{
	idle_task_exit();
	signal_processor(smp_processor_id(), sigp_stop);
	BUG();
	for (;;);
}

#endif /* CONFIG_HOTPLUG_CPU */

/*
 *	Cycle through the processors and setup structures.
 */

void __init smp_prepare_cpus(unsigned int max_cpus)
{
	unsigned long stack;
	unsigned int cpu;
	int i;

	/* request the 0x1201 emergency signal external interrupt */
	if (register_external_interrupt(0x1201, do_ext_call_interrupt) != 0)
		panic("Couldn't request external interrupt 0x1201");
	memset(lowcore_ptr, 0, sizeof(lowcore_ptr));
	/*
	 *  Initialize prefix pages and stacks for all possible cpus
	 */
	print_cpu_info(&S390_lowcore.cpu_data);

	for_each_possible_cpu(i) {
		lowcore_ptr[i] = (struct _lowcore *)
			__get_free_pages(GFP_KERNEL | GFP_DMA,
					 sizeof(void*) == 8 ? 1 : 0);
		stack = __get_free_pages(GFP_KERNEL, ASYNC_ORDER);
		if (!lowcore_ptr[i] || !stack)
			panic("smp_boot_cpus failed to allocate memory\n");

		*(lowcore_ptr[i]) = S390_lowcore;
		lowcore_ptr[i]->async_stack = stack + ASYNC_SIZE;
		stack = __get_free_pages(GFP_KERNEL, 0);
		if (!stack)
			panic("smp_boot_cpus failed to allocate memory\n");
		lowcore_ptr[i]->panic_stack = stack + PAGE_SIZE;
#ifndef CONFIG_64BIT
		if (MACHINE_HAS_IEEE) {
			lowcore_ptr[i]->extended_save_area_addr =
				(__u32) __get_free_pages(GFP_KERNEL, 0);
			if (!lowcore_ptr[i]->extended_save_area_addr)
				panic("smp_boot_cpus failed to "
				      "allocate memory\n");
		}
#endif
	}
#ifndef CONFIG_64BIT
	if (MACHINE_HAS_IEEE)
		ctl_set_bit(14, 29); /* enable extended save area */
#endif
	set_prefix((u32)(unsigned long) lowcore_ptr[smp_processor_id()]);

	for_each_possible_cpu(cpu)
		if (cpu != smp_processor_id())
			smp_create_idle(cpu);
}

void __init smp_prepare_boot_cpu(void)
{
	BUG_ON(smp_processor_id() != 0);

	cpu_set(0, cpu_online_map);
	S390_lowcore.percpu_offset = __per_cpu_offset[0];
	current_set[0] = current;
	spin_lock_init(&(&__get_cpu_var(s390_idle))->lock);
}

void __init smp_cpus_done(unsigned int max_cpus)
{
	cpu_present_map = cpu_possible_map;
}

/*
 * the frequency of the profiling timer can be changed
 * by writing a multiplier value into /proc/profile.
 *
 * usually you want to run this on all CPUs ;)
 */
int setup_profiling_timer(unsigned int multiplier)
{
	return 0;
}

static DEFINE_PER_CPU(struct cpu, cpu_devices);

static ssize_t show_capability(struct sys_device *dev, char *buf)
{
	unsigned int capability;
	int rc;

	rc = get_cpu_capability(&capability);
	if (rc)
		return rc;
	return sprintf(buf, "%u\n", capability);
}
static SYSDEV_ATTR(capability, 0444, show_capability, NULL);

static ssize_t show_idle_count(struct sys_device *dev, char *buf)
{
	struct s390_idle_data *idle;
	unsigned long long idle_count;

	idle = &per_cpu(s390_idle, dev->id);
	spin_lock_irq(&idle->lock);
	idle_count = idle->idle_count;
	spin_unlock_irq(&idle->lock);
	return sprintf(buf, "%llu\n", idle_count);
}
static SYSDEV_ATTR(idle_count, 0444, show_idle_count, NULL);

static ssize_t show_idle_time(struct sys_device *dev, char *buf)
{
	struct s390_idle_data *idle;
	unsigned long long new_time;

	idle = &per_cpu(s390_idle, dev->id);
	spin_lock_irq(&idle->lock);
	if (idle->in_idle) {
		new_time = get_clock();
		idle->idle_time += new_time - idle->idle_enter;
		idle->idle_enter = new_time;
	}
	new_time = idle->idle_time;
	spin_unlock_irq(&idle->lock);
	return sprintf(buf, "%llu\n", new_time >> 12);
}
static SYSDEV_ATTR(idle_time_us, 0444, show_idle_time, NULL);

static struct attribute *cpu_attrs[] = {
	&attr_capability.attr,
	&attr_idle_count.attr,
	&attr_idle_time_us.attr,
	NULL,
};

static struct attribute_group cpu_attr_group = {
	.attrs = cpu_attrs,
};

static int __cpuinit smp_cpu_notify(struct notifier_block *self,
				    unsigned long action, void *hcpu)
{
	unsigned int cpu = (unsigned int)(long)hcpu;
	struct cpu *c = &per_cpu(cpu_devices, cpu);
	struct sys_device *s = &c->sysdev;
	struct s390_idle_data *idle;

	switch (action) {
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
		idle = &per_cpu(s390_idle, cpu);
		spin_lock_irq(&idle->lock);
		idle->idle_enter = 0;
		idle->idle_time = 0;
		idle->idle_count = 0;
		spin_unlock_irq(&idle->lock);
		if (sysfs_create_group(&s->kobj, &cpu_attr_group))
			return NOTIFY_BAD;
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		sysfs_remove_group(&s->kobj, &cpu_attr_group);
		break;
	}
	return NOTIFY_OK;
}

static struct notifier_block __cpuinitdata smp_cpu_nb = {
	.notifier_call = smp_cpu_notify,
};

static int __init topology_init(void)
{
	int cpu;
	int rc;

	register_cpu_notifier(&smp_cpu_nb);

	for_each_possible_cpu(cpu) {
		struct cpu *c = &per_cpu(cpu_devices, cpu);
		struct sys_device *s = &c->sysdev;

		c->hotpluggable = 1;
		register_cpu(c, cpu);
		if (!cpu_online(cpu))
			continue;
		s = &c->sysdev;
		rc = sysfs_create_group(&s->kobj, &cpu_attr_group);
		if (rc)
			return rc;
	}
	return 0;
}
subsys_initcall(topology_init);