smp.c

linux-2.6.15.6

/*
 * parameter area for the set/clear control bit callbacks
 */
typedef struct
{
	__u16 start_ctl;
	__u16 end_ctl;
	unsigned long orvals[16];
	unsigned long andvals[16];
} ec_creg_mask_parms;

/*
 * callback for setting/clearing control bits
 */
void smp_ctl_bit_callback(void *info) {
	ec_creg_mask_parms *pp;
	unsigned long cregs[16];
	int i;
	
	pp = (ec_creg_mask_parms *) info;
	__ctl_store(cregs[pp->start_ctl], pp->start_ctl, pp->end_ctl);
	for (i = pp->start_ctl; i <= pp->end_ctl; i++)
		cregs[i] = (cregs[i] & pp->andvals[i]) | pp->orvals[i];
	__ctl_load(cregs[pp->start_ctl], pp->start_ctl, pp->end_ctl);
}

/*
 * Set a bit in a control register of all cpus
 */
void smp_ctl_set_bit(int cr, int bit) {
        ec_creg_mask_parms parms;

	parms.start_ctl = cr;
	parms.end_ctl = cr;
	parms.orvals[cr] = 1 << bit;
	parms.andvals[cr] = -1L;
	preempt_disable();
	smp_call_function(smp_ctl_bit_callback, &parms, 0, 1);
        __ctl_set_bit(cr, bit);
	preempt_enable();
}

/*
 * Clear a bit in a control register of all cpus
 */
void smp_ctl_clear_bit(int cr, int bit) {
        ec_creg_mask_parms parms;

	parms.start_ctl = cr;
	parms.end_ctl = cr;
	parms.orvals[cr] = 0;
	parms.andvals[cr] = ~(1L << bit);
	preempt_disable();
	smp_call_function(smp_ctl_bit_callback, &parms, 0, 1);
        __ctl_clear_bit(cr, bit);
	preempt_enable();
}

/*
 * Lets check how many CPUs we have.
 */

void
__init smp_check_cpus(unsigned int max_cpus)
{
	int cpu, num_cpus;
	__u16 boot_cpu_addr;

	/*
	 * cpu 0 is the boot cpu. See smp_prepare_boot_cpu.
	 */

	boot_cpu_addr = S390_lowcore.cpu_data.cpu_addr;
	current_thread_info()->cpu = 0;
	num_cpus = 1;
	for (cpu = 0; cpu <= 65535 && num_cpus < max_cpus; cpu++) {
		if ((__u16) cpu == boot_cpu_addr)
			continue;
		__cpu_logical_map[num_cpus] = (__u16) cpu;
		if (signal_processor(num_cpus, sigp_sense) ==
		    sigp_not_operational)
			continue;
		cpu_set(num_cpus, cpu_present_map);
		num_cpus++;
	}

	for (cpu = 1; cpu < max_cpus; cpu++)
		cpu_set(cpu, cpu_possible_map);

	printk("Detected %d CPU's\n",(int) num_cpus);
	printk("Boot cpu address %2X\n", boot_cpu_addr);
}

/*
 *      Activate a secondary processor.
 */
extern void init_cpu_timer(void);
extern void init_cpu_vtimer(void);
extern int pfault_init(void);
extern void pfault_fini(void);

int __devinit start_secondary(void *cpuvoid)
{
        /* Setup the cpu */
        cpu_init();
	preempt_disable();
        /* init per CPU timer */
        init_cpu_timer();
#ifdef CONFIG_VIRT_TIMER
        init_cpu_vtimer();
#endif
#ifdef CONFIG_PFAULT
	/* Enable pfault pseudo page faults on this cpu. */
	if (MACHINE_IS_VM)
		pfault_init();
#endif
	/* 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;
}

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;
}

/* Reserving and releasing of CPUs */

static DEFINE_SPINLOCK(smp_reserve_lock);
static int smp_cpu_reserved[NR_CPUS];

int
smp_get_cpu(cpumask_t cpu_mask)
{
	unsigned long flags;
	int cpu;

	spin_lock_irqsave(&smp_reserve_lock, flags);
	/* Try to find an already reserved cpu. */
	for_each_cpu_mask(cpu, cpu_mask) {
		if (smp_cpu_reserved[cpu] != 0) {
			smp_cpu_reserved[cpu]++;
			/* Found one. */
			goto out;
		}
	}
	/* Reserve a new cpu from cpu_mask. */
	for_each_cpu_mask(cpu, cpu_mask) {
		if (cpu_online(cpu)) {
			smp_cpu_reserved[cpu]++;
			goto out;
		}
	}
	cpu = -ENODEV;
out:
	spin_unlock_irqrestore(&smp_reserve_lock, flags);
	return cpu;
}

void
smp_put_cpu(int cpu)
{
	unsigned long flags;

	spin_lock_irqsave(&smp_reserve_lock, flags);
	smp_cpu_reserved[cpu]--;
	spin_unlock_irqrestore(&smp_reserve_lock, flags);
}

static inline 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)
		idle->thread_info + (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();
	signal_processor(cpu,sigp_restart);

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

int
__cpu_disable(void)
{
	unsigned long flags;
	ec_creg_mask_parms cr_parms;
	int cpu = smp_processor_id();

	spin_lock_irqsave(&smp_reserve_lock, flags);
	if (smp_cpu_reserved[cpu] != 0) {
		spin_unlock_irqrestore(&smp_reserve_lock, flags);
		return -EBUSY;
	}
	cpu_clear(cpu, cpu_online_map);

#ifdef CONFIG_PFAULT
	/* Disable pfault pseudo page faults on this cpu. */
	if (MACHINE_IS_VM)
		pfault_fini();
#endif

	/* disable all external interrupts */

	cr_parms.start_ctl = 0;
	cr_parms.end_ctl = 0;
	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);
	smp_ctl_bit_callback(&cr_parms);

	/* disable all I/O interrupts */

	cr_parms.start_ctl = 6;
	cr_parms.end_ctl = 6;
	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);
	smp_ctl_bit_callback(&cr_parms);

	/* disable most machine checks */

	cr_parms.start_ctl = 14;
	cr_parms.end_ctl = 14;
	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);

	spin_unlock_irqrestore(&smp_reserve_lock, flags);
	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(;;);
}

/*
 *	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");
        smp_check_cpus(max_cpus);
        memset(lowcore_ptr,0,sizeof(lowcore_ptr));  
        /*
         *  Initialize prefix pages and stacks for all possible cpus
         */
	print_cpu_info(&S390_lowcore.cpu_data);

        for(i = 0; i < NR_CPUS; i++) {
		if (!cpu_possible(i))
			continue;
		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] == NULL || stack == 0ULL)
			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 == 0ULL)
			panic("smp_boot_cpus failed to allocate memory\n");
		lowcore_ptr[i]->panic_stack = stack + (PAGE_SIZE);
#ifndef __s390x__
		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 == 0)
				panic("smp_boot_cpus failed to "
				      "allocate memory\n");
		}
#endif
	}
#ifndef __s390x__
	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_cpu(cpu)
		if (cpu != smp_processor_id())
			smp_create_idle(cpu);
}

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

	cpu_set(0, cpu_online_map);
	cpu_set(0, cpu_present_map);
	cpu_set(0, cpu_possible_map);
	S390_lowcore.percpu_offset = __per_cpu_offset[0];
	current_set[0] = current;
}

void 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 int __init topology_init(void)
{
	int cpu;
	int ret;

	for_each_cpu(cpu) {
		ret = register_cpu(&per_cpu(cpu_devices, cpu), cpu, NULL);
		if (ret)
			printk(KERN_WARNING "topology_init: register_cpu %d "
			       "failed (%d)\n", cpu, ret);
	}
	return 0;
}

subsys_initcall(topology_init);

EXPORT_SYMBOL(cpu_possible_map);
EXPORT_SYMBOL(lowcore_ptr);
EXPORT_SYMBOL(smp_ctl_set_bit);
EXPORT_SYMBOL(smp_ctl_clear_bit);
EXPORT_SYMBOL(smp_call_function);
EXPORT_SYMBOL(smp_get_cpu);
EXPORT_SYMBOL(smp_put_cpu);