/*
 * 
 *
 * SMP support for ppc.
 *
 * Written by Cort Dougan (cort@cs.nmt.edu) borrowing a great
 * deal of code from the sparc and intel versions.
 *
 * Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
 *
 * PowerPC-64 Support added by Dave Engebretsen, Peter Bergner, and
 * Mike Corrigan {engebret|bergner|mikec}@us.ibm.com
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#include <linux/config.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/smp.h>
#include <linux/smp_lock.h>
#include <linux/interrupt.h>
#include <linux/kernel_stat.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/cache.h>
#include <linux/err.h>

#include <asm/ptrace.h>
#include <asm/atomic.h>
#include <asm/irq.h>
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/hardirq.h>
#include <asm/softirq.h>
#include <asm/io.h>
#include <asm/prom.h>
#include <asm/smp.h>
#include <asm/naca.h>
#include <asm/paca.h>
#include <asm/iSeries/LparData.h>
#include <asm/iSeries/HvCall.h>
#include <asm/iSeries/HvCallCfg.h>
#include <asm/time.h>
#include <asm/ppcdebug.h>
#include "open_pic.h"
#include <asm/machdep.h>

int smp_threads_ready = 0;
spinlock_t kernel_flag __cacheline_aligned = SPIN_LOCK_UNLOCKED;
unsigned long cache_decay_ticks;

/* initialised so it doesnt end up in bss */
unsigned long cpu_online_map = 0;
int boot_cpuid = 0;

static struct smp_ops_t *smp_ops;

volatile unsigned long cpu_callin_map[NR_CPUS];

extern unsigned char stab_array[];

extern int cpu_idle(void *unused);
void smp_call_function_interrupt(void);
void smp_message_pass(int target, int msg, unsigned long data, int wait);
static unsigned long iSeries_smp_message[NR_CPUS];

void xics_setup_cpu(void);
void xics_cause_IPI(int cpu);

/*
 * XICS only has a single IPI, so encode the messages per CPU
 */
struct xics_ipi_struct {
	        volatile unsigned long value;
} ____cacheline_aligned;

struct xics_ipi_struct xics_ipi_message[NR_CPUS] __cacheline_aligned;

#define smp_message_pass(t,m,d,w) smp_ops->message_pass((t),(m),(d),(w))

static inline void set_tb(unsigned int upper, unsigned int lower)
{
	mtspr(SPRN_TBWL, 0);
	mtspr(SPRN_TBWU, upper);
	mtspr(SPRN_TBWL, lower);
}

void iSeries_smp_message_recv( struct pt_regs * regs )
{
	int cpu = smp_processor_id();
	int msg;

	if ( num_online_cpus() < 2 )
		return;

	for ( msg = 0; msg < 4; ++msg )
		if ( test_and_clear_bit( msg, &iSeries_smp_message[cpu] ) )
			smp_message_recv( msg, regs );
}

static void smp_iSeries_message_pass(int target, int msg, unsigned long data, int wait)
{
	int i;

	for (i = 0; i < NR_CPUS; ++i) {
		if (!cpu_online(i))
			continue;

		if ((target == MSG_ALL) || 
		    (target == i) || 
		    ((target == MSG_ALL_BUT_SELF) &&
		     (i != smp_processor_id())) ) {
			set_bit(msg, &iSeries_smp_message[i]);
			HvCall_sendIPI(&(paca[i]));
		}
	}
}

#ifdef CONFIG_PPC_ISERIES
static int smp_iSeries_numProcs(void)
{
	unsigned np, i;
	struct ItLpPaca * lpPaca;

	np = 0;
        for (i=0; i < MAX_PACAS; ++i) {
                lpPaca = paca[i].xLpPacaPtr;
                if ( lpPaca->xDynProcStatus < 2 ) {
                        ++np;
                }
        }
	return np;
}
#endif

static int smp_iSeries_probe(void)
{
	unsigned i;
	unsigned np = 0;
	struct ItLpPaca *lpPaca;

	for (i=0; i < MAX_PACAS; ++i) {
		lpPaca = paca[i].xLpPacaPtr;
		if (lpPaca->xDynProcStatus < 2) {
			paca[i].active = 1;
			++np;
		}
	}

	return np;
}

static void smp_iSeries_kick_cpu(int nr)
{
	struct ItLpPaca * lpPaca;
	/* Verify we have a Paca for processor nr */
	if ( ( nr <= 0 ) ||
	     ( nr >= MAX_PACAS ) )
		return;
	/* Verify that our partition has a processor nr */
	lpPaca = paca[nr].xLpPacaPtr;
	if ( lpPaca->xDynProcStatus >= 2 )
		return;

	/* The information for processor bringup must
	 * be written out to main store before we release
	 * the processor.
	 */
	mb();

	/* The processor is currently spinning, waiting
	 * for the xProcStart field to become non-zero
	 * After we set xProcStart, the processor will
	 * continue on to secondary_start in iSeries_head.S
	 */
	paca[nr].xProcStart = 1;
}

static void smp_iSeries_setup_cpu(int nr)
{
}

/* This is called very early. */
void smp_init_iSeries(void)
{
	smp_ops = &ppc_md.smp_ops;
	smp_ops->message_pass = smp_iSeries_message_pass;
	smp_ops->probe        = smp_iSeries_probe;
	smp_ops->kick_cpu     = smp_iSeries_kick_cpu;
	smp_ops->setup_cpu    = smp_iSeries_setup_cpu;
#ifdef CONFIG_PPC_ISERIES
#warning fix for iseries
	naca->processorCount	= smp_iSeries_numProcs();
#endif
}


static void
smp_openpic_message_pass(int target, int msg, unsigned long data, int wait)
{
	/* make sure we're sending something that translates to an IPI */
	if ( msg > 0x3 ){
		printk("SMP %d: smp_message_pass: unknown msg %d\n",
		       smp_processor_id(), msg);
		return;
	}
	switch ( target )
	{
	case MSG_ALL:
		openpic_cause_IPI(msg, 0xffffffff);
		break;
	case MSG_ALL_BUT_SELF:
		openpic_cause_IPI(msg,
				  0xffffffff & ~(1 << smp_processor_id()));
		break;
	default:
		openpic_cause_IPI(msg, 1<<target);
		break;
	}
}

static int smp_chrp_probe(void)
{
	int i;
	int nr_cpus = 0;

	for (i = 0; i < NR_CPUS; i++) {
		if (cpu_possible(i))
			nr_cpus++;
	}

	if (nr_cpus > 1)
		openpic_request_IPIs();

	return nr_cpus;
}

static void
smp_kick_cpu(int nr)
{
	/* Verify we have a Paca for processor nr */
	if ( ( nr <= 0 ) ||
	     ( nr >= MAX_PACAS ) )
		return;

	/* The information for processor bringup must
	 * be written out to main store before we release
	 * the processor.
	 */
	mb();

	/* The processor is currently spinning, waiting
	 * for the xProcStart field to become non-zero
	 * After we set xProcStart, the processor will
	 * continue on to secondary_start
	 */
	paca[nr].xProcStart = 1;
}

static void smp_space_timers(unsigned int max_cpus)
{
	int i;
	unsigned long offset = tb_ticks_per_jiffy / max_cpus;
	unsigned long previous_tb = paca[boot_cpuid].next_jiffy_update_tb;

	for (i = 0; i < NR_CPUS; i++) {
		if (cpu_possible(i) && i != boot_cpuid) {
			paca[i].next_jiffy_update_tb =
				previous_tb + offset;
			previous_tb = paca[i].next_jiffy_update_tb;
		}
	}
}

static void __devinit pSeries_setup_cpu(int cpu)
{
	if (OpenPIC_Addr) {
		do_openpic_setup_cpu();
	} else {
		if (cpu != boot_cpuid)
			xics_setup_cpu();
	}
}

static void
smp_xics_message_pass(int target, int msg, unsigned long data, int wait)
{
	int i;

	for (i = 0; i < NR_CPUS; ++i) {
		if (!cpu_online(i))
			continue;

		if (target == MSG_ALL || target == i
		    || (target == MSG_ALL_BUT_SELF
			&& i != smp_processor_id())) {
			set_bit(msg, &xics_ipi_message[i].value);
			mb();
			xics_cause_IPI(i);
		}
	}
}

static int smp_xics_probe(void)
{
	int i;
	int nr_cpus = 0;

	for (i = 0; i < NR_CPUS; i++) {
		if (cpu_possible(i))
			nr_cpus++;
	}

	return nr_cpus;
}

static spinlock_t timebase_lock = SPIN_LOCK_UNLOCKED;
static unsigned long timebase = 0;

static void __devinit pSeries_give_timebase(void)
{
	spin_lock(&timebase_lock);
	rtas_call(rtas_token("freeze-time-base"), 0, 1, NULL);
	timebase = get_tb();
	spin_unlock(&timebase_lock);

	while (timebase)
		barrier();
	rtas_call(rtas_token("thaw-time-base"), 0, 1, NULL);
}

static void __devinit pSeries_take_timebase(void)
{
	while (!timebase)
		barrier();
	spin_lock(&timebase_lock);
	set_tb(timebase, timebase >> 32);
	timebase = 0;
	spin_unlock(&timebase_lock);
}

/* This is called very early */
void __init smp_init_pSeries(void)
{
	smp_ops = &ppc_md.smp_ops;

	if (naca->interrupt_controller == IC_OPEN_PIC) {
		smp_ops->message_pass	= smp_openpic_message_pass;
		smp_ops->probe		= smp_chrp_probe;
	} else {
		smp_ops->message_pass	= smp_xics_message_pass;
		smp_ops->probe		= smp_xics_probe;
	}

	if (naca->platform == PLATFORM_PSERIES) {
		smp_ops->give_timebase = pSeries_give_timebase;
		smp_ops->take_timebase = pSeries_take_timebase;
	}

	smp_ops->kick_cpu = smp_kick_cpu;
	smp_ops->setup_cpu = pSeries_setup_cpu;
}

void smp_local_timer_interrupt(struct pt_regs * regs)
{
	if (!--(get_paca()->prof_counter)) {
		update_process_times(user_mode(regs));
		(get_paca()->prof_counter)=get_paca()->prof_multiplier;
	}
}

void smp_message_recv(int msg, struct pt_regs *regs)
{
	switch( msg ) {
	case PPC_MSG_CALL_FUNCTION:
		smp_call_function_interrupt();
		break;
	case PPC_MSG_RESCHEDULE: 
		/* XXX Do we have to do this? */
		set_need_resched();
		break;
#if 0
	case PPC_MSG_MIGRATE_TASK:
		/* spare */
		break;
#endif
#ifdef CONFIG_XMON
	case PPC_MSG_XMON_BREAK:
		xmon(regs);
		break;
#endif /* CONFIG_XMON */
	default:
		printk("SMP %d: smp_message_recv(): unknown msg %d\n",
		       smp_processor_id(), msg);
		break;
	}
}

void smp_send_reschedule(int cpu)
{
	smp_message_pass(cpu, PPC_MSG_RESCHEDULE, 0, 0);
}

/*
 * this function sends a reschedule IPI to all (other) CPUs.
 * This should only be used if some 'global' task became runnable,
 * such as a RT task, that must be handled now. The first CPU
 * that manages to grab the task will run it.
 */
void smp_send_reschedule_all(void)
{
	smp_message_pass(MSG_ALL_BUT_SELF, PPC_MSG_RESCHEDULE, 0, 0);
}

#ifdef CONFIG_XMON
void smp_send_xmon_break(int cpu)
{
	smp_message_pass(cpu, PPC_MSG_XMON_BREAK, 0, 0);
}
#endif /* CONFIG_XMON */

static void stop_this_cpu(void *dummy)
{
	local_irq_disable();
	while (1)
		;
}

void smp_send_stop(void)
{
	smp_call_function(stop_this_cpu, NULL, 1, 0);
}

/*
 * Structure and data for smp_call_function(). This is designed to minimise
 * static memory requirements. It also looks cleaner.
 * Stolen from the i386 version.
 */
static spinlock_t call_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;

static struct call_data_struct {
	void (*func) (void *info);
	void *info;
	atomic_t started;
	atomic_t finished;
	int wait;
} *call_data;

/*
 * This function sends a 'generic call function' IPI to all other CPUs
 * in the system.
 *
 * [SUMMARY] Run a function on all other CPUs.
 * <func> The function to run. This must be fast and non-blocking.
 * <info> An arbitrary pointer to pass to the function.
 * <nonatomic> currently unused.
 * <wait> If true, wait (atomically) until function has completed on other CPUs.
 * [RETURNS] 0 on success, else a negative status code. Does not return until
 * remote CPUs are nearly ready to execute <<func>> or are or have executed.
 *
 * You must not call this function with disabled interrupts or from a
 * hardware interrupt handler or from a bottom half handler.
 */
int smp_call_function (void (*func) (void *info), void *info, int nonatomic,
		       int wait)
{ 
	struct call_data_struct data;
	int ret = -1, cpus = num_online_cpus()-1;
	int timeout;

	if (!cpus)
		return 0;

	data.func = func;
	data.info = info;
	atomic_set(&data.started, 0);
	data.wait = wait;
	if (wait)
		atomic_set(&data.finished, 0);

	spin_lock(&call_lock);
	call_data = &data;
	/* Send a message to all other CPUs and wait for them to respond */
	smp_message_pass(MSG_ALL_BUT_SELF, PPC_MSG_CALL_FUNCTION, 0, 0);

	/* Wait for response */
	timeout = 8000000;
	while (atomic_read(&data.started) != cpus) {
		HMT_low();
		if (--timeout == 0) {
			printk("smp_call_function on cpu %d: other cpus not responding (%d)\n",
			       smp_processor_id(), atomic_read(&data.started));
#ifdef CONFIG_XMON
                        xmon(0);
#endif
#ifdef CONFIG_PPC_ISERIES
			HvCall_terminateMachineSrc();
#endif
			goto out;
		}
		barrier();
		udelay(1);
	}

	if (wait) {
		timeout = 1000000;
		while (atomic_read(&data.finished) != cpus) {
			HMT_low();
			if (--timeout == 0) {
				printk("smp_call_function on cpu %d: other cpus not finishing (%d/%d)\n",
				       smp_processor_id(), atomic_read(&data.finished), atomic_read(&data.started));
#ifdef CONFIG_PPC_ISERIES
				HvCall_terminateMachineSrc();
#endif
				goto out;
			}
			barrier();
			udelay(1);
		}
	}
	ret = 0;

 out:
	HMT_medium();
	spin_unlock(&call_lock);
	return ret;
}

void smp_call_function_interrupt(void)
{
	void (*func) (void *info) = call_data->func;
	void *info = call_data->info;
	int wait = call_data->wait;

	/*
	 * Notify initiating CPU that I've grabbed the data and am
	 * about to execute the function
	 */
	atomic_inc(&call_data->started);
	/*
	 * At this point the info structure may be out of scope unless wait==1
	 */
	(*func)(info);
	if (wait)
		atomic_inc(&call_data->finished);
}

extern unsigned long decr_overclock;
extern struct gettimeofday_struct do_gtod;

struct thread_info *current_set[NR_CPUS];

static void __devinit smp_store_cpu_info(int id)
{
	paca[id].pvr = _get_PVR();
}

void __init smp_prepare_cpus(unsigned int max_cpus)
{
	int i;

	/* Fixup boot cpu */
	smp_store_cpu_info(smp_processor_id());
	cpu_callin_map[smp_processor_id()] = 1;

	for (i = 0; i < NR_CPUS; i++) {
		paca[i].prof_counter = 1;
		paca[i].prof_multiplier = 1;
		if (i != boot_cpuid) {
			void *tmp;
		        /*
			 * the boot cpu segment table is statically 
			 * initialized to real address 0x5000.  The
			 * Other processor's tables are created and
			 * initialized here.
			 */
			tmp = &stab_array[PAGE_SIZE * (i-1)];
			memset(tmp, 0, PAGE_SIZE); 
			paca[i].xStab_data.virt = (unsigned long)tmp;
			paca[i].xStab_data.real = (unsigned long)__v2a(tmp);
			paca[i].default_decr = tb_ticks_per_jiffy /
				decr_overclock;
		}
	}

	/*
	 * XXX very rough. 
	 */
	cache_decay_ticks = HZ/100;

#ifndef CONFIG_PPC_ISERIES
	paca[boot_cpuid].next_jiffy_update_tb = tb_last_stamp = get_tb();

	/*
	 * Should update do_gtod.stamp_xsec.
	 * For now we leave it which means the time can be some
	 * number of msecs off until someone does a settimeofday()
	 */
	do_gtod.tb_orig_stamp = tb_last_stamp;
#endif

	max_cpus = smp_ops->probe();
	smp_space_timers(max_cpus);
}

int __cpu_up(unsigned int cpu)
{
	struct pt_regs regs;
	struct task_struct *p;
	int c;

	/* create a process for the processor */
	/* only regs.msr is actually used, and 0 is OK for it */
	memset(&regs, 0, sizeof(struct pt_regs));
	p = do_fork(CLONE_VM|CLONE_IDLETASK, 0, &regs, 0);
	if (IS_ERR(p))
		panic("failed fork for CPU %u: %li", cpu, PTR_ERR(p));

	init_idle(p, cpu);
	unhash_process(p);

	paca[cpu].xCurrent = (u64)p;
	current_set[cpu] = p->thread_info;

	/* wake up cpus */
	smp_ops->kick_cpu(cpu);

	/*
	 * wait to see if the cpu made a callin (is actually up).
	 * use this value that I found through experimentation.
	 * -- Cort
	 */
	for (c = 5000; c && !cpu_callin_map[cpu]; c--)
		udelay(100);

	if (!cpu_callin_map[cpu]) {
		printk("Processor %u is stuck.\n", cpu);
		return -ENOENT;
	}

	printk("Processor %u found.\n", cpu);

	if (smp_ops->give_timebase)
		smp_ops->give_timebase();
	set_bit(cpu, &cpu_online_map);
	return 0;
}

/* Activate a secondary processor. */
int __devinit start_secondary(void *unused)
{
	unsigned int cpu = smp_processor_id();

	atomic_inc(&init_mm.mm_count);
	current->active_mm = &init_mm;

	smp_store_cpu_info(cpu);
	set_dec(paca[cpu].default_decr);
	cpu_callin_map[cpu] = 1;

	smp_ops->setup_cpu(cpu);
	if (smp_ops->take_timebase)
		smp_ops->take_timebase();

	/* XXX required? */
	local_irq_enable();

	return cpu_idle(NULL);
}

int setup_profiling_timer(unsigned int multiplier)
{
	return 0;
}

void smp_cpus_done(unsigned int max_cpus)
{
	smp_ops->setup_cpu(boot_cpuid);

	/* XXX fix this, xics currently relies on it - Anton */
	smp_threads_ready = 1;
}