smpboot.c

LINUX 2.6.17.4的源码

/*
 * Activate a secondary processor.  head.S calls this.
 */
int __devinit
start_secondary (void *unused)
{
	/* Early console may use I/O ports */
	ia64_set_kr(IA64_KR_IO_BASE, __pa(ia64_iobase));
	Dprintk("start_secondary: starting CPU 0x%x\n", hard_smp_processor_id());
	efi_map_pal_code();
	cpu_init();
	preempt_disable();
	smp_callin();

	cpu_idle();
	return 0;
}

struct pt_regs * __devinit idle_regs(struct pt_regs *regs)
{
	return NULL;
}

struct create_idle {
	struct task_struct *idle;
	struct completion done;
	int cpu;
};

void
do_fork_idle(void *_c_idle)
{
	struct create_idle *c_idle = _c_idle;

	c_idle->idle = fork_idle(c_idle->cpu);
	complete(&c_idle->done);
}

static int __devinit
do_boot_cpu (int sapicid, int cpu)
{
	int timeout;
	struct create_idle c_idle = {
		.cpu	= cpu,
		.done	= COMPLETION_INITIALIZER(c_idle.done),
	};
	DECLARE_WORK(work, do_fork_idle, &c_idle);

 	c_idle.idle = get_idle_for_cpu(cpu);
 	if (c_idle.idle) {
		init_idle(c_idle.idle, cpu);
 		goto do_rest;
	}

	/*
	 * We can't use kernel_thread since we must avoid to reschedule the child.
	 */
	if (!keventd_up() || current_is_keventd())
		work.func(work.data);
	else {
		schedule_work(&work);
		wait_for_completion(&c_idle.done);
	}

	if (IS_ERR(c_idle.idle))
		panic("failed fork for CPU %d", cpu);

	set_idle_for_cpu(cpu, c_idle.idle);

do_rest:
	task_for_booting_cpu = c_idle.idle;

	Dprintk("Sending wakeup vector %lu to AP 0x%x/0x%x.\n", ap_wakeup_vector, cpu, sapicid);

	set_brendez_area(cpu);
	platform_send_ipi(cpu, ap_wakeup_vector, IA64_IPI_DM_INT, 0);

	/*
	 * Wait 10s total for the AP to start
	 */
	Dprintk("Waiting on callin_map ...");
	for (timeout = 0; timeout < 100000; timeout++) {
		if (cpu_isset(cpu, cpu_callin_map))
			break;  /* It has booted */
		udelay(100);
	}
	Dprintk("\n");

	if (!cpu_isset(cpu, cpu_callin_map)) {
		printk(KERN_ERR "Processor 0x%x/0x%x is stuck.\n", cpu, sapicid);
		ia64_cpu_to_sapicid[cpu] = -1;
		cpu_clear(cpu, cpu_online_map);  /* was set in smp_callin() */
		return -EINVAL;
	}
	return 0;
}

static int __init
decay (char *str)
{
	int ticks;
	get_option (&str, &ticks);
	return 1;
}

__setup("decay=", decay);

/*
 * Initialize the logical CPU number to SAPICID mapping
 */
void __init
smp_build_cpu_map (void)
{
	int sapicid, cpu, i;
	int boot_cpu_id = hard_smp_processor_id();

	for (cpu = 0; cpu < NR_CPUS; cpu++) {
		ia64_cpu_to_sapicid[cpu] = -1;
	}

	ia64_cpu_to_sapicid[0] = boot_cpu_id;
	cpus_clear(cpu_present_map);
	cpu_set(0, cpu_present_map);
	cpu_set(0, cpu_possible_map);
	for (cpu = 1, i = 0; i < smp_boot_data.cpu_count; i++) {
		sapicid = smp_boot_data.cpu_phys_id[i];
		if (sapicid == boot_cpu_id)
			continue;
		cpu_set(cpu, cpu_present_map);
		cpu_set(cpu, cpu_possible_map);
		ia64_cpu_to_sapicid[cpu] = sapicid;
		cpu++;
	}
}

/*
 * Cycle through the APs sending Wakeup IPIs to boot each.
 */
void __init
smp_prepare_cpus (unsigned int max_cpus)
{
	int boot_cpu_id = hard_smp_processor_id();

	/*
	 * Initialize the per-CPU profiling counter/multiplier
	 */

	smp_setup_percpu_timer();

	/*
	 * We have the boot CPU online for sure.
	 */
	cpu_set(0, cpu_online_map);
	cpu_set(0, cpu_callin_map);

	local_cpu_data->loops_per_jiffy = loops_per_jiffy;
	ia64_cpu_to_sapicid[0] = boot_cpu_id;

	printk(KERN_INFO "Boot processor id 0x%x/0x%x\n", 0, boot_cpu_id);

	current_thread_info()->cpu = 0;

	/*
	 * If SMP should be disabled, then really disable it!
	 */
	if (!max_cpus) {
		printk(KERN_INFO "SMP mode deactivated.\n");
		cpus_clear(cpu_online_map);
		cpus_clear(cpu_present_map);
		cpus_clear(cpu_possible_map);
		cpu_set(0, cpu_online_map);
		cpu_set(0, cpu_present_map);
		cpu_set(0, cpu_possible_map);
		return;
	}
}

void __devinit smp_prepare_boot_cpu(void)
{
	cpu_set(smp_processor_id(), cpu_online_map);
	cpu_set(smp_processor_id(), cpu_callin_map);
	per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
}

#ifdef CONFIG_HOTPLUG_CPU
static inline void
clear_cpu_sibling_map(int cpu)
{
	int i;

	for_each_cpu_mask(i, cpu_sibling_map[cpu])
		cpu_clear(cpu, cpu_sibling_map[i]);
	for_each_cpu_mask(i, cpu_core_map[cpu])
		cpu_clear(cpu, cpu_core_map[i]);

	cpu_sibling_map[cpu] = cpu_core_map[cpu] = CPU_MASK_NONE;
}

static void
remove_siblinginfo(int cpu)
{
	int last = 0;

	if (cpu_data(cpu)->threads_per_core == 1 &&
	    cpu_data(cpu)->cores_per_socket == 1) {
		cpu_clear(cpu, cpu_core_map[cpu]);
		cpu_clear(cpu, cpu_sibling_map[cpu]);
		return;
	}

	last = (cpus_weight(cpu_core_map[cpu]) == 1 ? 1 : 0);

	/* remove it from all sibling map's */
	clear_cpu_sibling_map(cpu);
}

extern void fixup_irqs(void);

int migrate_platform_irqs(unsigned int cpu)
{
	int new_cpei_cpu;
	irq_desc_t *desc = NULL;
	cpumask_t 	mask;
	int 		retval = 0;

	/*
	 * dont permit CPEI target to removed.
	 */
	if (cpe_vector > 0 && is_cpu_cpei_target(cpu)) {
		printk ("CPU (%d) is CPEI Target\n", cpu);
		if (can_cpei_retarget()) {
			/*
			 * Now re-target the CPEI to a different processor
			 */
			new_cpei_cpu = any_online_cpu(cpu_online_map);
			mask = cpumask_of_cpu(new_cpei_cpu);
			set_cpei_target_cpu(new_cpei_cpu);
			desc = irq_descp(ia64_cpe_irq);
			/*
			 * Switch for now, immediatly, we need to do fake intr
			 * as other interrupts, but need to study CPEI behaviour with
			 * polling before making changes.
			 */
			if (desc) {
				desc->handler->disable(ia64_cpe_irq);
				desc->handler->set_affinity(ia64_cpe_irq, mask);
				desc->handler->enable(ia64_cpe_irq);
				printk ("Re-targetting CPEI to cpu %d\n", new_cpei_cpu);
			}
		}
		if (!desc) {
			printk ("Unable to retarget CPEI, offline cpu [%d] failed\n", cpu);
			retval = -EBUSY;
		}
	}
	return retval;
}

/* must be called with cpucontrol mutex held */
int __cpu_disable(void)
{
	int cpu = smp_processor_id();

	/*
	 * dont permit boot processor for now
	 */
	if (cpu == 0 && !bsp_remove_ok) {
		printk ("Your platform does not support removal of BSP\n");
		return (-EBUSY);
	}

	cpu_clear(cpu, cpu_online_map);

	if (migrate_platform_irqs(cpu)) {
		cpu_set(cpu, cpu_online_map);
		return (-EBUSY);
	}

	remove_siblinginfo(cpu);
	cpu_clear(cpu, cpu_online_map);
	fixup_irqs();
	local_flush_tlb_all();
	cpu_clear(cpu, cpu_callin_map);
	return 0;
}

void __cpu_die(unsigned int cpu)
{
	unsigned int i;

	for (i = 0; i < 100; i++) {
		/* They ack this in play_dead by setting CPU_DEAD */
		if (per_cpu(cpu_state, cpu) == CPU_DEAD)
		{
			printk ("CPU %d is now offline\n", cpu);
			return;
		}
		msleep(100);
	}
 	printk(KERN_ERR "CPU %u didn't die...\n", cpu);
}
#else /* !CONFIG_HOTPLUG_CPU */
int __cpu_disable(void)
{
	return -ENOSYS;
}

void __cpu_die(unsigned int cpu)
{
	/* We said "no" in __cpu_disable */
	BUG();
}
#endif /* CONFIG_HOTPLUG_CPU */

void
smp_cpus_done (unsigned int dummy)
{
	int cpu;
	unsigned long bogosum = 0;

	/*
	 * Allow the user to impress friends.
	 */

	for_each_online_cpu(cpu) {
		bogosum += cpu_data(cpu)->loops_per_jiffy;
	}

	printk(KERN_INFO "Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
	       (int)num_online_cpus(), bogosum/(500000/HZ), (bogosum/(5000/HZ))%100);
}

static inline void __devinit
set_cpu_sibling_map(int cpu)
{
	int i;

	for_each_online_cpu(i) {
		if ((cpu_data(cpu)->socket_id == cpu_data(i)->socket_id)) {
			cpu_set(i, cpu_core_map[cpu]);
			cpu_set(cpu, cpu_core_map[i]);
			if (cpu_data(cpu)->core_id == cpu_data(i)->core_id) {
				cpu_set(i, cpu_sibling_map[cpu]);
				cpu_set(cpu, cpu_sibling_map[i]);
			}
		}
	}
}

int __devinit
__cpu_up (unsigned int cpu)
{
	int ret;
	int sapicid;

	sapicid = ia64_cpu_to_sapicid[cpu];
	if (sapicid == -1)
		return -EINVAL;

	/*
	 * Already booted cpu? not valid anymore since we dont
	 * do idle loop tightspin anymore.
	 */
	if (cpu_isset(cpu, cpu_callin_map))
		return -EINVAL;

	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
	/* Processor goes to start_secondary(), sets online flag */
	ret = do_boot_cpu(sapicid, cpu);
	if (ret < 0)
		return ret;

	if (cpu_data(cpu)->threads_per_core == 1 &&
	    cpu_data(cpu)->cores_per_socket == 1) {
		cpu_set(cpu, cpu_sibling_map[cpu]);
		cpu_set(cpu, cpu_core_map[cpu]);
		return 0;
	}

	set_cpu_sibling_map(cpu);

	return 0;
}

/*
 * Assume that CPU's have been discovered by some platform-dependent interface.  For
 * SoftSDV/Lion, that would be ACPI.
 *
 * Setup of the IPI irq handler is done in irq.c:init_IRQ_SMP().
 */
void __init
init_smp_config(void)
{
	struct fptr {
		unsigned long fp;
		unsigned long gp;
	} *ap_startup;
	long sal_ret;

	/* Tell SAL where to drop the AP's.  */
	ap_startup = (struct fptr *) start_ap;
	sal_ret = ia64_sal_set_vectors(SAL_VECTOR_OS_BOOT_RENDEZ,
				       ia64_tpa(ap_startup->fp), ia64_tpa(ap_startup->gp), 0, 0, 0, 0);
	if (sal_ret < 0)
		printk(KERN_ERR "SMP: Can't set SAL AP Boot Rendezvous: %s\n",
		       ia64_sal_strerror(sal_ret));
}

/*
 * identify_siblings(cpu) gets called from identify_cpu. This populates the 
 * information related to logical execution units in per_cpu_data structure.
 */
void __devinit
identify_siblings(struct cpuinfo_ia64 *c)
{
	s64 status;
	u16 pltid;
	pal_logical_to_physical_t info;

	if (smp_num_cpucores == 1 && smp_num_siblings == 1)
		return;

	if ((status = ia64_pal_logical_to_phys(-1, &info)) != PAL_STATUS_SUCCESS) {
		printk(KERN_ERR "ia64_pal_logical_to_phys failed with %ld\n",
		       status);
		return;
	}
	if ((status = ia64_sal_physical_id_info(&pltid)) != PAL_STATUS_SUCCESS) {
		printk(KERN_ERR "ia64_sal_pltid failed with %ld\n", status);
		return;
	}

	c->socket_id =  (pltid << 8) | info.overview_ppid;
	c->cores_per_socket = info.overview_cpp;
	c->threads_per_core = info.overview_tpc;
	c->num_log = info.overview_num_log;

	c->core_id = info.log1_cid;
	c->thread_id = info.log1_tid;
}