/* $Id: irq.c,v 1.114 2002/01/11 08:45:38 davem Exp $
 * irq.c: UltraSparc IRQ handling/init/registry.
 *
 * Copyright (C) 1997  David S. Miller  (davem@caip.rutgers.edu)
 * Copyright (C) 1998  Eddie C. Dost    (ecd@skynet.be)
 * Copyright (C) 1998  Jakub Jelinek    (jj@ultra.linux.cz)
 */

#include <linux/config.h>
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/ptrace.h>
#include <linux/errno.h>
#include <linux/kernel_stat.h>
#include <linux/signal.h>
#include <linux/mm.h>
#include <linux/interrupt.h>
#include <linux/slab.h>
#include <linux/random.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>

#include <asm/ptrace.h>
#include <asm/processor.h>
#include <asm/atomic.h>
#include <asm/system.h>
#include <asm/irq.h>
#include <asm/sbus.h>
#include <asm/iommu.h>
#include <asm/upa.h>
#include <asm/oplib.h>
#include <asm/timer.h>
#include <asm/smp.h>
#include <asm/starfire.h>
#include <asm/uaccess.h>
#include <asm/cache.h>
#include <asm/cpudata.h>

#ifdef CONFIG_SMP
static void distribute_irqs(void);
#endif

/* UPA nodes send interrupt packet to UltraSparc with first data reg
 * value low 5 (7 on Starfire) bits holding the IRQ identifier being
 * delivered.  We must translate this into a non-vector IRQ so we can
 * set the softint on this cpu.
 *
 * To make processing these packets efficient and race free we use
 * an array of irq buckets below.  The interrupt vector handler in
 * entry.S feeds incoming packets into per-cpu pil-indexed lists.
 * The IVEC handler does not need to act atomically, the PIL dispatch
 * code uses CAS to get an atomic snapshot of the list and clear it
 * at the same time.
 */

struct ino_bucket ivector_table[NUM_IVECS] __attribute__ ((aligned (SMP_CACHE_BYTES)));

/* This has to be in the main kernel image, it cannot be
 * turned into per-cpu data.  The reason is that the main
 * kernel image is locked into the TLB and this structure
 * is accessed from the vectored interrupt trap handler.  If
 * access to this structure takes a TLB miss it could cause
 * the 5-level sparc v9 trap stack to overflow.
 */
struct irq_work_struct {
	unsigned int	irq_worklists[16];
};
struct irq_work_struct __irq_work[NR_CPUS];
#define irq_work(__cpu, __pil)	&(__irq_work[(__cpu)].irq_worklists[(__pil)])

#ifdef CONFIG_PCI
/* This is a table of physical addresses used to deal with IBF_DMA_SYNC.
 * It is used for PCI only to synchronize DMA transfers with IRQ delivery
 * for devices behind busses other than APB on Sabre systems.
 *
 * Currently these physical addresses are just config space accesses
 * to the command register for that device.
 */
unsigned long pci_dma_wsync;
unsigned long dma_sync_reg_table[256];
unsigned char dma_sync_reg_table_entry = 0;
#endif

/* This is based upon code in the 32-bit Sparc kernel written mostly by
 * David Redman (djhr@tadpole.co.uk).
 */
#define MAX_STATIC_ALLOC	4
static struct irqaction static_irqaction[MAX_STATIC_ALLOC];
static int static_irq_count;

/* This is exported so that fast IRQ handlers can get at it... -DaveM */
struct irqaction *irq_action[NR_IRQS+1] = {
	  NULL, NULL, NULL, NULL, NULL, NULL , NULL, NULL,
	  NULL, NULL, NULL, NULL, NULL, NULL , NULL, NULL
};

/* This only synchronizes entities which modify IRQ handler
 * state and some selected user-level spots that want to
 * read things in the table.  IRQ handler processing orders
 * its' accesses such that no locking is needed.
 */
static DEFINE_SPINLOCK(irq_action_lock);

static void register_irq_proc (unsigned int irq);

/*
 * Upper 2b of irqaction->flags holds the ino.
 * irqaction->mask holds the smp affinity information.
 */
#define put_ino_in_irqaction(action, irq) \
	action->flags &= 0xffffffffffffUL; \
	if (__bucket(irq) == &pil0_dummy_bucket) \
		action->flags |= 0xdeadUL << 48;  \
	else \
		action->flags |= __irq_ino(irq) << 48;
#define get_ino_in_irqaction(action)	(action->flags >> 48)

#define put_smpaff_in_irqaction(action, smpaff)	(action)->mask = (smpaff)
#define get_smpaff_in_irqaction(action) 	((action)->mask)

int show_interrupts(struct seq_file *p, void *v)
{
	unsigned long flags;
	int i = *(loff_t *) v;
	struct irqaction *action;
#ifdef CONFIG_SMP
	int j;
#endif

	spin_lock_irqsave(&irq_action_lock, flags);
	if (i <= NR_IRQS) {
		if (!(action = *(i + irq_action)))
			goto out_unlock;
		seq_printf(p, "%3d: ", i);
#ifndef CONFIG_SMP
		seq_printf(p, "%10u ", kstat_irqs(i));
#else
		for (j = 0; j < NR_CPUS; j++) {
			if (!cpu_online(j))
				continue;
			seq_printf(p, "%10u ",
				   kstat_cpu(j).irqs[i]);
		}
#endif
		seq_printf(p, " %s:%lx", action->name,
			   get_ino_in_irqaction(action));
		for (action = action->next; action; action = action->next) {
			seq_printf(p, ", %s:%lx", action->name,
				   get_ino_in_irqaction(action));
		}
		seq_putc(p, '\n');
	}
out_unlock:
	spin_unlock_irqrestore(&irq_action_lock, flags);

	return 0;
}

/* Now these are always passed a true fully specified sun4u INO. */
void enable_irq(unsigned int irq)
{
	struct ino_bucket *bucket = __bucket(irq);
	unsigned long imap;
	unsigned long tid;

	imap = bucket->imap;
	if (imap == 0UL)
		return;

	preempt_disable();

	if (tlb_type == cheetah || tlb_type == cheetah_plus) {
		unsigned long ver;

		__asm__ ("rdpr %%ver, %0" : "=r" (ver));
		if ((ver >> 32) == 0x003e0016) {
			/* We set it to our JBUS ID. */
			__asm__ __volatile__("ldxa [%%g0] %1, %0"
					     : "=r" (tid)
					     : "i" (ASI_JBUS_CONFIG));
			tid = ((tid & (0x1fUL<<17)) << 9);
			tid &= IMAP_TID_JBUS;
		} else {
			/* We set it to our Safari AID. */
			__asm__ __volatile__("ldxa [%%g0] %1, %0"
					     : "=r" (tid)
					     : "i" (ASI_SAFARI_CONFIG));
			tid = ((tid & (0x3ffUL<<17)) << 9);
			tid &= IMAP_AID_SAFARI;
		}
	} else if (this_is_starfire == 0) {
		/* We set it to our UPA MID. */
		__asm__ __volatile__("ldxa [%%g0] %1, %0"
				     : "=r" (tid)
				     : "i" (ASI_UPA_CONFIG));
		tid = ((tid & UPA_CONFIG_MID) << 9);
		tid &= IMAP_TID_UPA;
	} else {
		tid = (starfire_translate(imap, smp_processor_id()) << 26);
		tid &= IMAP_TID_UPA;
	}

	/* NOTE NOTE NOTE, IGN and INO are read-only, IGN is a product
	 * of this SYSIO's preconfigured IGN in the SYSIO Control
	 * Register, the hardware just mirrors that value here.
	 * However for Graphics and UPA Slave devices the full
	 * IMAP_INR field can be set by the programmer here.
	 *
	 * Things like FFB can now be handled via the new IRQ mechanism.
	 */
	upa_writel(tid | IMAP_VALID, imap);

	preempt_enable();
}

/* This now gets passed true ino's as well. */
void disable_irq(unsigned int irq)
{
	struct ino_bucket *bucket = __bucket(irq);
	unsigned long imap;

	imap = bucket->imap;
	if (imap != 0UL) {
		u32 tmp;

		/* NOTE: We do not want to futz with the IRQ clear registers
		 *       and move the state to IDLE, the SCSI code does call
		 *       disable_irq() to assure atomicity in the queue cmd
		 *       SCSI adapter driver code.  Thus we'd lose interrupts.
		 */
		tmp = upa_readl(imap);
		tmp &= ~IMAP_VALID;
		upa_writel(tmp, imap);
	}
}

/* The timer is the one "weird" interrupt which is generated by
 * the CPU %tick register and not by some normal vectored interrupt
 * source.  To handle this special case, we use this dummy INO bucket.
 */
static struct ino_bucket pil0_dummy_bucket = {
	0,	/* irq_chain */
	0,	/* pil */
	0,	/* pending */
	0,	/* flags */
	0,	/* __unused */
	NULL,	/* irq_info */
	0UL,	/* iclr */
	0UL,	/* imap */
};

unsigned int build_irq(int pil, int inofixup, unsigned long iclr, unsigned long imap)
{
	struct ino_bucket *bucket;
	int ino;

	if (pil == 0) {
		if (iclr != 0UL || imap != 0UL) {
			prom_printf("Invalid dummy bucket for PIL0 (%lx:%lx)\n",
				    iclr, imap);
			prom_halt();
		}
		return __irq(&pil0_dummy_bucket);
	}

	/* RULE: Both must be specified in all other cases. */
	if (iclr == 0UL || imap == 0UL) {
		prom_printf("Invalid build_irq %d %d %016lx %016lx\n",
			    pil, inofixup, iclr, imap);
		prom_halt();
	}
	
	ino = (upa_readl(imap) & (IMAP_IGN | IMAP_INO)) + inofixup;
	if (ino > NUM_IVECS) {
		prom_printf("Invalid INO %04x (%d:%d:%016lx:%016lx)\n",
			    ino, pil, inofixup, iclr, imap);
		prom_halt();
	}

	/* Ok, looks good, set it up.  Don't touch the irq_chain or
	 * the pending flag.
	 */
	bucket = &ivector_table[ino];
	if ((bucket->flags & IBF_ACTIVE) ||
	    (bucket->irq_info != NULL)) {
		/* This is a gross fatal error if it happens here. */
		prom_printf("IRQ: Trying to reinit INO bucket, fatal error.\n");
		prom_printf("IRQ: Request INO %04x (%d:%d:%016lx:%016lx)\n",
			    ino, pil, inofixup, iclr, imap);
		prom_printf("IRQ: Existing (%d:%016lx:%016lx)\n",
			    bucket->pil, bucket->iclr, bucket->imap);
		prom_printf("IRQ: Cannot continue, halting...\n");
		prom_halt();
	}
	bucket->imap  = imap;
	bucket->iclr  = iclr;
	bucket->pil   = pil;
	bucket->flags = 0;

	bucket->irq_info = NULL;

	return __irq(bucket);
}

static void atomic_bucket_insert(struct ino_bucket *bucket)
{
	unsigned long pstate;
	unsigned int *ent;

	__asm__ __volatile__("rdpr %%pstate, %0" : "=r" (pstate));
	__asm__ __volatile__("wrpr %0, %1, %%pstate"
			     : : "r" (pstate), "i" (PSTATE_IE));
	ent = irq_work(smp_processor_id(), bucket->pil);
	bucket->irq_chain = *ent;
	*ent = __irq(bucket);
	__asm__ __volatile__("wrpr %0, 0x0, %%pstate" : : "r" (pstate));
}

int request_irq(unsigned int irq, irqreturn_t (*handler)(int, void *, struct pt_regs *),
		unsigned long irqflags, const char *name, void *dev_id)
{
	struct irqaction *action, *tmp = NULL;
	struct ino_bucket *bucket = __bucket(irq);
	unsigned long flags;
	int pending = 0;

	if ((bucket != &pil0_dummy_bucket) &&
	    (bucket < &ivector_table[0] ||
	     bucket >= &ivector_table[NUM_IVECS])) {
		unsigned int *caller;

		__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
		printk(KERN_CRIT "request_irq: Old style IRQ registry attempt "
		       "from %p, irq %08x.\n", caller, irq);
		return -EINVAL;
	}	
	if (!handler)
	    return -EINVAL;

	if ((bucket != &pil0_dummy_bucket) && (irqflags & SA_SAMPLE_RANDOM)) {
		/*
	 	 * This function might sleep, we want to call it first,
	 	 * outside of the atomic block. In SA_STATIC_ALLOC case,
		 * random driver's kmalloc will fail, but it is safe.
		 * If already initialized, random driver will not reinit.
	 	 * Yes, this might clear the entropy pool if the wrong
	 	 * driver is attempted to be loaded, without actually
	 	 * installing a new handler, but is this really a problem,
	 	 * only the sysadmin is able to do this.
	 	 */
		rand_initialize_irq(irq);
	}

	spin_lock_irqsave(&irq_action_lock, flags);

	action = *(bucket->pil + irq_action);
	if (action) {
		if ((action->flags & SA_SHIRQ) && (irqflags & SA_SHIRQ))
			for (tmp = action; tmp->next; tmp = tmp->next)
				;
		else {
			spin_unlock_irqrestore(&irq_action_lock, flags);
			return -EBUSY;
		}
		action = NULL;		/* Or else! */
	}

	/* If this is flagged as statically allocated then we use our
	 * private struct which is never freed.
	 */
	if (irqflags & SA_STATIC_ALLOC) {
	    if (static_irq_count < MAX_STATIC_ALLOC)
		action = &static_irqaction[static_irq_count++];
	    else
		printk("Request for IRQ%d (%s) SA_STATIC_ALLOC failed "
		       "using kmalloc\n", irq, name);
	}	
	if (action == NULL)
	    action = (struct irqaction *)kmalloc(sizeof(struct irqaction),
						 GFP_ATOMIC);
	
	if (!action) { 
		spin_unlock_irqrestore(&irq_action_lock, flags);
		return -ENOMEM;
	}

	if (bucket == &pil0_dummy_bucket) {
		bucket->irq_info = action;
		bucket->flags |= IBF_ACTIVE;
	} else {
		if ((bucket->flags & IBF_ACTIVE) != 0) {
			void *orig = bucket->irq_info;
			void **vector = NULL;

			if ((bucket->flags & IBF_PCI) == 0) {
				printk("IRQ: Trying to share non-PCI bucket.\n");
				goto free_and_ebusy;
			}
			if ((bucket->flags & IBF_MULTI) == 0) {
				vector = kmalloc(sizeof(void *) * 4, GFP_ATOMIC);
				if (vector == NULL)
					goto free_and_enomem;

				/* We might have slept. */
				if ((bucket->flags & IBF_MULTI) != 0) {
					int ent;

					kfree(vector);
					vector = (void **)bucket->irq_info;
					for(ent = 0; ent < 4; ent++) {
						if (vector[ent] == NULL) {
							vector[ent] = action;
							break;
						}
					}
					if (ent == 4)
						goto free_and_ebusy;
				} else {
					vector[0] = orig;
					vector[1] = action;
					vector[2] = NULL;
					vector[3] = NULL;
					bucket->irq_info = vector;
					bucket->flags |= IBF_MULTI;
				}
			} else {
				int ent;

				vector = (void **)orig;
				for (ent = 0; ent < 4; ent++) {
					if (vector[ent] == NULL) {
						vector[ent] = action;
						break;
					}
				}
				if (ent == 4)
					goto free_and_ebusy;
			}
		} else {
			bucket->irq_info = action;
			bucket->flags |= IBF_ACTIVE;
		}
		pending = bucket->pending;
		if (pending)
			bucket->pending = 0;
	}

	action->handler = handler;
	action->flags = irqflags;
	action->name = name;
	action->next = NULL;
	action->dev_id = dev_id;
	put_ino_in_irqaction(action, irq);
	put_smpaff_in_irqaction(action, CPU_MASK_NONE);

	if (tmp)
		tmp->next = action;
	else
		*(bucket->pil + irq_action) = action;

	enable_irq(irq);

	/* We ate the IVEC already, this makes sure it does not get lost. */
	if (pending) {
		atomic_bucket_insert(bucket);
		set_softint(1 << bucket->pil);
	}
	spin_unlock_irqrestore(&irq_action_lock, flags);
	if ((bucket != &pil0_dummy_bucket) && (!(irqflags & SA_STATIC_ALLOC)))
		register_irq_proc(__irq_ino(irq));

#ifdef CONFIG_SMP
	distribute_irqs();
#endif
	return 0;

free_and_ebusy:
	kfree(action);
	spin_unlock_irqrestore(&irq_action_lock, flags);
	return -EBUSY;

free_and_enomem:
	kfree(action);
	spin_unlock_irqrestore(&irq_action_lock, flags);
	return -ENOMEM;
}

EXPORT_SYMBOL(request_irq);

void free_irq(unsigned int irq, void *dev_id)
{
	struct irqaction *action;
	struct irqaction *tmp = NULL;
	unsigned long flags;
	struct ino_bucket *bucket = __bucket(irq), *bp;

	if ((bucket != &pil0_dummy_bucket) &&
	    (bucket < &ivector_table[0] ||
	     bucket >= &ivector_table[NUM_IVECS])) {
		unsigned int *caller;

		__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
		printk(KERN_CRIT "free_irq: Old style IRQ removal attempt "
		       "from %p, irq %08x.\n", caller, irq);
		return;
	}
	
	spin_lock_irqsave(&irq_action_lock, flags);

	action = *(bucket->pil + irq_action);
	if (!action->handler) {
		printk("Freeing free IRQ %d\n", bucket->pil);
		return;
	}
	if (dev_id) {
		for ( ; action; action = action->next) {
			if (action->dev_id == dev_id)
				break;
			tmp = action;
		}
		if (!action) {
			printk("Trying to free free shared IRQ %d\n", bucket->pil);
			spin_unlock_irqrestore(&irq_action_lock, flags);
			return;
		}
	} else if (action->flags & SA_SHIRQ) {
		printk("Trying to free shared IRQ %d with NULL device ID\n", bucket->pil);
		spin_unlock_irqrestore(&irq_action_lock, flags);
		return;
	}

	if (action->flags & SA_STATIC_ALLOC) {
		printk("Attempt to free statically allocated IRQ %d (%s)\n",
		       bucket->pil, action->name);
		spin_unlock_irqrestore(&irq_action_lock, flags);
		return;
	}

	if (action && tmp)
		tmp->next = action->next;
	else
		*(bucket->pil + irq_action) = action->next;

	spin_unlock_irqrestore(&irq_action_lock, flags);

	synchronize_irq(irq);

	spin_lock_irqsave(&irq_action_lock, flags);

	if (bucket != &pil0_dummy_bucket) {
		unsigned long imap = bucket->imap;
		void **vector, *orig;
		int ent;

		orig = bucket->irq_info;
		vector = (void **)orig;

		if ((bucket->flags & IBF_MULTI) != 0) {
			int other = 0;
			void *orphan = NULL;
			for (ent = 0; ent < 4; ent++) {
				if (vector[ent] == action)
					vector[ent] = NULL;
				else if (vector[ent] != NULL) {
					orphan = vector[ent];
					other++;
				}
			}

			/* Only free when no other shared irq
			 * uses this bucket.
			 */
			if (other) {
				if (other == 1) {
					/* Convert back to non-shared bucket. */
					bucket->irq_info = orphan;
					bucket->flags &= ~(IBF_MULTI);
					kfree(vector);
				}
				goto out;
			}
		} else {
			bucket->irq_info = NULL;
		}

		/* This unique interrupt source is now inactive. */
		bucket->flags &= ~IBF_ACTIVE;

		/* See if any other buckets share this bucket's IMAP
		 * and are still active.
		 */
		for (ent = 0; ent < NUM_IVECS; ent++) {
			bp = &ivector_table[ent];
			if (bp != bucket	&&
			    bp->imap == imap	&&
			    (bp->flags & IBF_ACTIVE) != 0)
				break;
		}

		/* Only disable when no other sub-irq levels of
		 * the same IMAP are active.
		 */
		if (ent == NUM_IVECS)
			disable_irq(irq);
	}

out:
	kfree(action);
	spin_unlock_irqrestore(&irq_action_lock, flags);
}

EXPORT_SYMBOL(free_irq);

#ifdef CONFIG_SMP
void synchronize_irq(unsigned int irq)
{
	struct ino_bucket *bucket = __bucket(irq);

#if 0
	/* The following is how I wish I could implement this.
	 * Unfortunately the ICLR registers are read-only, you can
	 * only write ICLR_foo values to them.  To get the current
	 * IRQ status you would need to get at the IRQ diag registers
	 * in the PCI/SBUS controller and the layout of those vary
	 * from one controller to the next, sigh... -DaveM
	 */
	unsigned long iclr = bucket->iclr;

	while (1) {
		u32 tmp = upa_readl(iclr);
		
		if (tmp == ICLR_TRANSMIT ||
		    tmp == ICLR_PENDING) {
			cpu_relax();
			continue;
		}
		break;
	}
#else
	/* So we have to do this with a INPROGRESS bit just like x86.  */
	while (bucket->flags & IBF_INPROGRESS)
		cpu_relax();
#endif
}
#endif /* CONFIG_SMP */

void catch_disabled_ivec(struct pt_regs *regs)
{
	int cpu = smp_processor_id();
	struct ino_bucket *bucket = __bucket(*irq_work(cpu, 0));

	/* We can actually see this on Ultra/PCI PCI cards, which are bridges
	 * to other devices.  Here a single IMAP enabled potentially multiple
	 * unique interrupt sources (which each do have a unique ICLR register.
	 *
	 * So what we do is just register that the IVEC arrived, when registered
	 * for real the request_irq() code will check the bit and signal
	 * a local CPU interrupt for it.
	 */
#if 0
	printk("IVEC: Spurious interrupt vector (%x) received at (%016lx)\n",
	       bucket - &ivector_table[0], regs->tpc);
#endif
	*irq_work(cpu, 0) = 0;
	bucket->pending = 1;
}

/* Tune this... */
#define FORWARD_VOLUME		12

#ifdef CONFIG_SMP

static inline void redirect_intr(int cpu, struct ino_bucket *bp)
{
	/* Ok, here is what is going on:
	 * 1) Retargeting IRQs on Starfire is very
	 *    expensive so just forget about it on them.
	 * 2) Moving around very high priority interrupts
	 *    is a losing game.
	 * 3) If the current cpu is idle, interrupts are
	 *    useful work, so keep them here.  But do not
	 *    pass to our neighbour if he is not very idle.
	 * 4) If sysadmin explicitly asks for directed intrs,
	 *    Just Do It.
	 */
	struct irqaction *ap = bp->irq_info;
	cpumask_t cpu_mask;
	unsigned int buddy, ticks;

	cpu_mask = get_smpaff_in_irqaction(ap);
	cpus_and(cpu_mask, cpu_mask, cpu_online_map);
	if (cpus_empty(cpu_mask))
		cpu_mask = cpu_online_map;

	if (this_is_starfire != 0 ||
	    bp->pil >= 10 || current->pid == 0)
		goto out;

	/* 'cpu' is the MID (ie. UPAID), calculate the MID
	 * of our buddy.
	 */
	buddy = cpu + 1;
	if (buddy >= NR_CPUS)
		buddy = 0;

	ticks = 0;
	while (!cpu_isset(buddy, cpu_mask)) {
		if (++buddy >= NR_CPUS)
			buddy = 0;
		if (++ticks > NR_CPUS) {
			put_smpaff_in_irqaction(ap, CPU_MASK_NONE);
			goto out;
		}
	}

	if (buddy == cpu)
		goto out;

	/* Voo-doo programming. */
	if (cpu_data(buddy).idle_volume < FORWARD_VOLUME)
		goto out;

	/* This just so happens to be correct on Cheetah
	 * at the moment.
	 */
	buddy <<= 26;

	/* Push it to our buddy. */
	upa_writel(buddy | IMAP_VALID, bp->imap);

out:
	return;
}

#endif

void handler_irq(int irq, struct pt_regs *regs)
{
	struct ino_bucket *bp, *nbp;
	int cpu = smp_processor_id();

#ifndef CONFIG_SMP
	/*
	 * Check for TICK_INT on level 14 softint.
	 */
	{
		unsigned long clr_mask = 1 << irq;
		unsigned long tick_mask = tick_ops->softint_mask;

		if ((irq == 14) && (get_softint() & tick_mask)) {
			irq = 0;
			clr_mask = tick_mask;
		}
		clear_softint(clr_mask);
	}
#else
	int should_forward = 1;

	clear_softint(1 << irq);
#endif

	irq_enter();
	kstat_this_cpu.irqs[irq]++;

	/* Sliiiick... */
#ifndef CONFIG_SMP
	bp = ((irq != 0) ?
	      __bucket(xchg32(irq_work(cpu, irq), 0)) :
	      &pil0_dummy_bucket);
#else
	bp = __bucket(xchg32(irq_work(cpu, irq), 0));
#endif
	for ( ; bp != NULL; bp = nbp) {
		unsigned char flags = bp->flags;
		unsigned char random = 0;

		nbp = __bucket(bp->irq_chain);
		bp->irq_chain = 0;

		bp->flags |= IBF_INPROGRESS;

		if ((flags & IBF_ACTIVE) != 0) {
#ifdef CONFIG_PCI
			if ((flags & IBF_DMA_SYNC) != 0) {
				upa_readl(dma_sync_reg_table[bp->synctab_ent]);
				upa_readq(pci_dma_wsync);
			}
#endif
			if ((flags & IBF_MULTI) == 0) {
				struct irqaction *ap = bp->irq_info;
				int ret;

				ret = ap->handler(__irq(bp), ap->dev_id, regs);
				if (ret == IRQ_HANDLED)
					random |= ap->flags;
			} else {
				void **vector = (void **)bp->irq_info;
				int ent;
				for (ent = 0; ent < 4; ent++) {
					struct irqaction *ap = vector[ent];
					if (ap != NULL) {
						int ret;

						ret = ap->handler(__irq(bp),
								  ap->dev_id,
								  regs);
						if (ret == IRQ_HANDLED)
							random |= ap->flags;
					}
				}
			}
			/* Only the dummy bucket lacks IMAP/ICLR. */
			if (bp->pil != 0) {
#ifdef CONFIG_SMP
				if (should_forward) {
					redirect_intr(cpu, bp);
					should_forward = 0;
				}
#endif
				upa_writel(ICLR_IDLE, bp->iclr);

				/* Test and add entropy */
				if (random & SA_SAMPLE_RANDOM)
					add_interrupt_randomness(irq);
			}
		} else
			bp->pending = 1;

		bp->flags &= ~IBF_INPROGRESS;
	}
	irq_exit();
}

#ifdef CONFIG_BLK_DEV_FD
extern void floppy_interrupt(int irq, void *dev_cookie, struct pt_regs *regs);

void sparc_floppy_irq(int irq, void *dev_cookie, struct pt_regs *regs)
{
	struct irqaction *action = *(irq + irq_action);
	struct ino_bucket *bucket;
	int cpu = smp_processor_id();

	irq_enter();
	kstat_this_cpu.irqs[irq]++;

	*(irq_work(cpu, irq)) = 0;
	bucket = get_ino_in_irqaction(action) + ivector_table;

	bucket->flags |= IBF_INPROGRESS;

	floppy_interrupt(irq, dev_cookie, regs);
	upa_writel(ICLR_IDLE, bucket->iclr);

	bucket->flags &= ~IBF_INPROGRESS;

	irq_exit();
}
#endif

/* The following assumes that the branch lies before the place we
 * are branching to.  This is the case for a trap vector...
 * You have been warned.
 */
#define SPARC_BRANCH(dest_addr, inst_addr) \
          (0x10800000 | ((((dest_addr)-(inst_addr))>>2)&0x3fffff))

#define SPARC_NOP (0x01000000)

static void install_fast_irq(unsigned int cpu_irq,
			     irqreturn_t (*handler)(int, void *, struct pt_regs *))
{
	extern unsigned long sparc64_ttable_tl0;
	unsigned long ttent = (unsigned long) &sparc64_ttable_tl0;
	unsigned int *insns;

	ttent += 0x820;
	ttent += (cpu_irq - 1) << 5;
	insns = (unsigned int *) ttent;
	insns[0] = SPARC_BRANCH(((unsigned long) handler),
				((unsigned long)&insns[0]));
	insns[1] = SPARC_NOP;
	__asm__ __volatile__("membar #StoreStore; flush %0" : : "r" (ttent));
}

int request_fast_irq(unsigned int irq,
		     irqreturn_t (*handler)(int, void *, struct pt_regs *),
		     unsigned long irqflags, const char *name, void *dev_id)
{
	struct irqaction *action;
	struct ino_bucket *bucket = __bucket(irq);
	unsigned long flags;

	/* No pil0 dummy buckets allowed here. */
	if (bucket < &ivector_table[0] ||
	    bucket >= &ivector_table[NUM_IVECS]) {
		unsigned int *caller;

		__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
		printk(KERN_CRIT "request_fast_irq: Old style IRQ registry attempt "
		       "from %p, irq %08x.\n", caller, irq);
		return -EINVAL;
	}	
	
	if (!handler)
		return -EINVAL;

	if ((bucket->pil == 0) || (bucket->pil == 14)) {
		printk("request_fast_irq: Trying to register shared IRQ 0 or 14.\n");
		return -EBUSY;
	}

	spin_lock_irqsave(&irq_action_lock, flags);

	action = *(bucket->pil + irq_action);
	if (action) {
		if (action->flags & SA_SHIRQ)
			panic("Trying to register fast irq when already shared.\n");
		if (irqflags & SA_SHIRQ)
			panic("Trying to register fast irq as shared.\n");
		printk("request_fast_irq: Trying to register yet already owned.\n");
		spin_unlock_irqrestore(&irq_action_lock, flags);
		return -EBUSY;
	}

	/*
	 * We do not check for SA_SAMPLE_RANDOM in this path. Neither do we
	 * support smp intr affinity in this path.
	 */
	if (irqflags & SA_STATIC_ALLOC) {
		if (static_irq_count < MAX_STATIC_ALLOC)
			action = &static_irqaction[static_irq_count++];
		else
			printk("Request for IRQ%d (%s) SA_STATIC_ALLOC failed "
			       "using kmalloc\n", bucket->pil, name);
	}
	if (action == NULL)
		action = (struct irqaction *)kmalloc(sizeof(struct irqaction),
						     GFP_ATOMIC);
	if (!action) {
		spin_unlock_irqrestore(&irq_action_lock, flags);
		return -ENOMEM;
	}
	install_fast_irq(bucket->pil, handler);

	bucket->irq_info = action;
	bucket->flags |= IBF_ACTIVE;

	action->handler = handler;
	action->flags = irqflags;
	action->dev_id = NULL;
	action->name = name;
	action->next = NULL;
	put_ino_in_irqaction(action, irq);
	put_smpaff_in_irqaction(action, CPU_MASK_NONE);

	*(bucket->pil + irq_action) = action;
	enable_irq(irq);

	spin_unlock_irqrestore(&irq_action_lock, flags);

#ifdef CONFIG_SMP
	distribute_irqs();
#endif
	return 0;
}

/* We really don't need these at all on the Sparc.  We only have
 * stubs here because they are exported to modules.
 */
unsigned long probe_irq_on(void)
{
	return 0;
}

EXPORT_SYMBOL(probe_irq_on);

int probe_irq_off(unsigned long mask)
{
	return 0;
}

EXPORT_SYMBOL(probe_irq_off);

#ifdef CONFIG_SMP
static int retarget_one_irq(struct irqaction *p, int goal_cpu)
{
	struct ino_bucket *bucket = get_ino_in_irqaction(p) + ivector_table;
	unsigned long imap = bucket->imap;
	unsigned int tid;

	while (!cpu_online(goal_cpu)) {
		if (++goal_cpu >= NR_CPUS)
			goal_cpu = 0;
	}

	if (tlb_type == cheetah || tlb_type == cheetah_plus) {
		tid = goal_cpu << 26;
		tid &= IMAP_AID_SAFARI;
	} else if (this_is_starfire == 0) {
		tid = goal_cpu << 26;
		tid &= IMAP_TID_UPA;
	} else {
		tid = (starfire_translate(imap, goal_cpu) << 26);
		tid &= IMAP_TID_UPA;
	}
	upa_writel(tid | IMAP_VALID, imap);

	while (!cpu_online(goal_cpu)) {
		if (++goal_cpu >= NR_CPUS)
			goal_cpu = 0;
	}

	return goal_cpu;
}

/* Called from request_irq. */
static void distribute_irqs(void)
{
	unsigned long flags;
	int cpu, level;

	spin_lock_irqsave(&irq_action_lock, flags);
	cpu = 0;

	/*
	 * Skip the timer at [0], and very rare error/power intrs at [15].
	 * Also level [12], it causes problems on Ex000 systems.
	 */
	for (level = 1; level < NR_IRQS; level++) {
		struct irqaction *p = irq_action[level];
		if (level == 12) continue;
		while(p) {
			cpu = retarget_one_irq(p, cpu);
			p = p->next;
		}
	}
	spin_unlock_irqrestore(&irq_action_lock, flags);
}
#endif


struct sun5_timer *prom_timers;
static u64 prom_limit0, prom_limit1;

static void map_prom_timers(void)
{
	unsigned int addr[3];
	int tnode, err;

	/* PROM timer node hangs out in the top level of device siblings... */
	tnode = prom_finddevice("/counter-timer");

	/* Assume if node is not present, PROM uses different tick mechanism
	 * which we should not care about.
	 */
	if (tnode == 0 || tnode == -1) {
		prom_timers = (struct sun5_timer *) 0;
		return;
	}

	/* If PROM is really using this, it must be mapped by him. */
	err = prom_getproperty(tnode, "address", (char *)addr, sizeof(addr));
	if (err == -1) {
		prom_printf("PROM does not have timer mapped, trying to continue.\n");
		prom_timers = (struct sun5_timer *) 0;
		return;
	}
	prom_timers = (struct sun5_timer *) ((unsigned long)addr[0]);
}

static void kill_prom_timer(void)
{
	if (!prom_timers)
		return;

	/* Save them away for later. */
	prom_limit0 = prom_timers->limit0;
	prom_limit1 = prom_timers->limit1;

	/* Just as in sun4c/sun4m PROM uses timer which ticks at IRQ 14.
	 * We turn both off here just to be paranoid.
	 */
	prom_timers->limit0 = 0;
	prom_timers->limit1 = 0;

	/* Wheee, eat the interrupt packet too... */
	__asm__ __volatile__(
"	mov	0x40, %%g2\n"
"	ldxa	[%%g0] %0, %%g1\n"
"	ldxa	[%%g2] %1, %%g1\n"
"	stxa	%%g0, [%%g0] %0\n"
"	membar	#Sync\n"
	: /* no outputs */
	: "i" (ASI_INTR_RECEIVE), "i" (ASI_INTR_R)
	: "g1", "g2");
}

void enable_prom_timer(void)
{
	if (!prom_timers)
		return;

	/* Set it to whatever was there before. */
	prom_timers->limit1 = prom_limit1;
	prom_timers->count1 = 0;
	prom_timers->limit0 = prom_limit0;
	prom_timers->count0 = 0;
}

void init_irqwork_curcpu(void)
{
	register struct irq_work_struct *workp asm("o2");
	register unsigned long tmp asm("o3");
	int cpu = hard_smp_processor_id();

	memset(__irq_work + cpu, 0, sizeof(*workp));

	/* Make sure we are called with PSTATE_IE disabled.  */
	__asm__ __volatile__("rdpr	%%pstate, %0\n\t"
			     : "=r" (tmp));
	if (tmp & PSTATE_IE) {
		prom_printf("BUG: init_irqwork_curcpu() called with "
			    "PSTATE_IE enabled, bailing.\n");
		__asm__ __volatile__("mov	%%i7, %0\n\t"
				     : "=r" (tmp));
		prom_printf("BUG: Called from %lx\n", tmp);
		prom_halt();
	}

	/* Set interrupt globals.  */
	workp = &__irq_work[cpu];
	__asm__ __volatile__(
	"rdpr	%%pstate, %0\n\t"
	"wrpr	%0, %1, %%pstate\n\t"
	"mov	%2, %%g6\n\t"
	"wrpr	%0, 0x0, %%pstate\n\t"
	: "=&r" (tmp)
	: "i" (PSTATE_IG), "r" (workp));
}

/* Only invoked on boot processor. */
void __init init_IRQ(void)
{
	map_prom_timers();
	kill_prom_timer();
	memset(&ivector_table[0], 0, sizeof(ivector_table));

	/* We need to clear any IRQ's pending in the soft interrupt
	 * registers, a spurious one could be left around from the
	 * PROM timer which we just disabled.
	 */
	clear_softint(get_softint());

	/* Now that ivector table is initialized, it is safe
	 * to receive IRQ vector traps.  We will normally take
	 * one or two right now, in case some device PROM used
	 * to boot us wants to speak to us.  We just ignore them.
	 */
	__asm__ __volatile__("rdpr	%%pstate, %%g1\n\t"
			     "or	%%g1, %0, %%g1\n\t"
			     "wrpr	%%g1, 0x0, %%pstate"
			     : /* No outputs */
			     : "i" (PSTATE_IE)
			     : "g1");
}

static struct proc_dir_entry * root_irq_dir;
static struct proc_dir_entry * irq_dir [NUM_IVECS];

#ifdef CONFIG_SMP

static int irq_affinity_read_proc (char *page, char **start, off_t off,
			int count, int *eof, void *data)
{
	struct ino_bucket *bp = ivector_table + (long)data;
	struct irqaction *ap = bp->irq_info;
	cpumask_t mask;
	int len;

	mask = get_smpaff_in_irqaction(ap);
	if (cpus_empty(mask))
		mask = cpu_online_map;

	len = cpumask_scnprintf(page, count, mask);
	if (count - len < 2)
		return -EINVAL;
	len += sprintf(page + len, "\n");
	return len;
}

static inline void set_intr_affinity(int irq, cpumask_t hw_aff)
{
	struct ino_bucket *bp = ivector_table + irq;

	/* Users specify affinity in terms of hw cpu ids.
	 * As soon as we do this, handler_irq() might see and take action.
	 */
	put_smpaff_in_irqaction((struct irqaction *)bp->irq_info, hw_aff);

	/* Migration is simply done by the next cpu to service this
	 * interrupt.
	 */
}

static int irq_affinity_write_proc (struct file *file, const char __user *buffer,
					unsigned long count, void *data)
{
	int irq = (long) data, full_count = count, err;
	cpumask_t new_value;

	err = cpumask_parse(buffer, count, new_value);

	/*
	 * Do not allow disabling IRQs completely - it's a too easy
	 * way to make the system unusable accidentally :-) At least
	 * one online CPU still has to be targeted.
	 */
	cpus_and(new_value, new_value, cpu_online_map);
	if (cpus_empty(new_value))
		return -EINVAL;

	set_intr_affinity(irq, new_value);

	return full_count;
}

#endif

#define MAX_NAMELEN 10

static void register_irq_proc (unsigned int irq)
{
	char name [MAX_NAMELEN];

	if (!root_irq_dir || irq_dir[irq])
		return;

	memset(name, 0, MAX_NAMELEN);
	sprintf(name, "%x", irq);

	/* create /proc/irq/1234 */
	irq_dir[irq] = proc_mkdir(name, root_irq_dir);

#ifdef CONFIG_SMP
	/* XXX SMP affinity not supported on starfire yet. */
	if (this_is_starfire == 0) {
		struct proc_dir_entry *entry;

		/* create /proc/irq/1234/smp_affinity */
		entry = create_proc_entry("smp_affinity", 0600, irq_dir[irq]);

		if (entry) {
			entry->nlink = 1;
			entry->data = (void *)(long)irq;
			entry->read_proc = irq_affinity_read_proc;
			entry->write_proc = irq_affinity_write_proc;
		}
	}
#endif
}

void init_irq_proc (void)
{
	/* create /proc/irq */
	root_irq_dir = proc_mkdir("irq", NULL);
}