blob: ac11754ecec544c965c196f9dc857b2b398b22e0 [file] [log] [blame]
/* MN10300 Arch-specific interrupt handling
*
* Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public Licence
* as published by the Free Software Foundation; either version
* 2 of the Licence, or (at your option) any later version.
*/
#include <linux/module.h>
#include <linux/interrupt.h>
#include <linux/kernel_stat.h>
#include <linux/seq_file.h>
#include <linux/cpumask.h>
#include <asm/setup.h>
#include <asm/serial-regs.h>
unsigned long __mn10300_irq_enabled_epsw[NR_CPUS] __cacheline_aligned_in_smp = {
[0 ... NR_CPUS - 1] = EPSW_IE | EPSW_IM_7
};
EXPORT_SYMBOL(__mn10300_irq_enabled_epsw);
#ifdef CONFIG_SMP
static char irq_affinity_online[NR_IRQS] = {
[0 ... NR_IRQS - 1] = 0
};
#define NR_IRQ_WORDS ((NR_IRQS + 31) / 32)
static unsigned long irq_affinity_request[NR_IRQ_WORDS] = {
[0 ... NR_IRQ_WORDS - 1] = 0
};
#endif /* CONFIG_SMP */
atomic_t irq_err_count;
/*
* MN10300 interrupt controller operations
*/
static void mn10300_cpupic_ack(unsigned int irq)
{
unsigned long flags;
u16 tmp;
flags = arch_local_cli_save();
GxICR_u8(irq) = GxICR_DETECT;
tmp = GxICR(irq);
arch_local_irq_restore(flags);
}
static void __mask_and_set_icr(unsigned int irq,
unsigned int mask, unsigned int set)
{
unsigned long flags;
u16 tmp;
flags = arch_local_cli_save();
tmp = GxICR(irq);
GxICR(irq) = (tmp & mask) | set;
tmp = GxICR(irq);
arch_local_irq_restore(flags);
}
static void mn10300_cpupic_mask(unsigned int irq)
{
__mask_and_set_icr(irq, GxICR_LEVEL, 0);
}
static void mn10300_cpupic_mask_ack(unsigned int irq)
{
#ifdef CONFIG_SMP
unsigned long flags;
u16 tmp;
flags = arch_local_cli_save();
if (!test_and_clear_bit(irq, irq_affinity_request)) {
tmp = GxICR(irq);
GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_DETECT;
tmp = GxICR(irq);
} else {
u16 tmp2;
tmp = GxICR(irq);
GxICR(irq) = (tmp & GxICR_LEVEL);
tmp2 = GxICR(irq);
irq_affinity_online[irq] =
any_online_cpu(*irq_desc[irq].affinity);
CROSS_GxICR(irq, irq_affinity_online[irq]) =
(tmp & (GxICR_LEVEL | GxICR_ENABLE)) | GxICR_DETECT;
tmp = CROSS_GxICR(irq, irq_affinity_online[irq]);
}
arch_local_irq_restore(flags);
#else /* CONFIG_SMP */
__mask_and_set_icr(irq, GxICR_LEVEL, GxICR_DETECT);
#endif /* CONFIG_SMP */
}
static void mn10300_cpupic_unmask(unsigned int irq)
{
__mask_and_set_icr(irq, GxICR_LEVEL, GxICR_ENABLE);
}
static void mn10300_cpupic_unmask_clear(unsigned int irq)
{
/* the MN10300 PIC latches its interrupt request bit, even after the
* device has ceased to assert its interrupt line and the interrupt
* channel has been disabled in the PIC, so for level-triggered
* interrupts we need to clear the request bit when we re-enable */
#ifdef CONFIG_SMP
unsigned long flags;
u16 tmp;
flags = arch_local_cli_save();
if (!test_and_clear_bit(irq, irq_affinity_request)) {
tmp = GxICR(irq);
GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_ENABLE | GxICR_DETECT;
tmp = GxICR(irq);
} else {
tmp = GxICR(irq);
irq_affinity_online[irq] = any_online_cpu(*irq_desc[irq].affinity);
CROSS_GxICR(irq, irq_affinity_online[irq]) = (tmp & GxICR_LEVEL) | GxICR_ENABLE | GxICR_DETECT;
tmp = CROSS_GxICR(irq, irq_affinity_online[irq]);
}
arch_local_irq_restore(flags);
#else /* CONFIG_SMP */
__mask_and_set_icr(irq, GxICR_LEVEL, GxICR_ENABLE | GxICR_DETECT);
#endif /* CONFIG_SMP */
}
#ifdef CONFIG_SMP
static int
mn10300_cpupic_setaffinity(unsigned int irq, const struct cpumask *mask)
{
unsigned long flags;
int err;
flags = arch_local_cli_save();
/* check irq no */
switch (irq) {
case TMJCIRQ:
case RESCHEDULE_IPI:
case CALL_FUNC_SINGLE_IPI:
case LOCAL_TIMER_IPI:
case FLUSH_CACHE_IPI:
case CALL_FUNCTION_NMI_IPI:
case GDB_NMI_IPI:
#ifdef CONFIG_MN10300_TTYSM0
case SC0RXIRQ:
case SC0TXIRQ:
#ifdef CONFIG_MN10300_TTYSM0_TIMER8
case TM8IRQ:
#elif CONFIG_MN10300_TTYSM0_TIMER2
case TM2IRQ:
#endif /* CONFIG_MN10300_TTYSM0_TIMER8 */
#endif /* CONFIG_MN10300_TTYSM0 */
#ifdef CONFIG_MN10300_TTYSM1
case SC1RXIRQ:
case SC1TXIRQ:
#ifdef CONFIG_MN10300_TTYSM1_TIMER12
case TM12IRQ:
#elif CONFIG_MN10300_TTYSM1_TIMER9
case TM9IRQ:
#elif CONFIG_MN10300_TTYSM1_TIMER3
case TM3IRQ:
#endif /* CONFIG_MN10300_TTYSM1_TIMER12 */
#endif /* CONFIG_MN10300_TTYSM1 */
#ifdef CONFIG_MN10300_TTYSM2
case SC2RXIRQ:
case SC2TXIRQ:
case TM10IRQ:
#endif /* CONFIG_MN10300_TTYSM2 */
err = -1;
break;
default:
set_bit(irq, irq_affinity_request);
err = 0;
break;
}
arch_local_irq_restore(flags);
return err;
}
#endif /* CONFIG_SMP */
/*
* MN10300 PIC level-triggered IRQ handling.
*
* The PIC has no 'ACK' function per se. It is possible to clear individual
* channel latches, but each latch relatches whether or not the channel is
* masked, so we need to clear the latch when we unmask the channel.
*
* Also for this reason, we don't supply an ack() op (it's unused anyway if
* mask_ack() is provided), and mask_ack() just masks.
*/
static struct irq_chip mn10300_cpu_pic_level = {
.name = "cpu_l",
.disable = mn10300_cpupic_mask,
.enable = mn10300_cpupic_unmask_clear,
.ack = NULL,
.mask = mn10300_cpupic_mask,
.mask_ack = mn10300_cpupic_mask,
.unmask = mn10300_cpupic_unmask_clear,
#ifdef CONFIG_SMP
.set_affinity = mn10300_cpupic_setaffinity,
#endif
};
/*
* MN10300 PIC edge-triggered IRQ handling.
*
* We use the latch clearing function of the PIC as the 'ACK' function.
*/
static struct irq_chip mn10300_cpu_pic_edge = {
.name = "cpu_e",
.disable = mn10300_cpupic_mask,
.enable = mn10300_cpupic_unmask,
.ack = mn10300_cpupic_ack,
.mask = mn10300_cpupic_mask,
.mask_ack = mn10300_cpupic_mask_ack,
.unmask = mn10300_cpupic_unmask,
#ifdef CONFIG_SMP
.set_affinity = mn10300_cpupic_setaffinity,
#endif
};
/*
* 'what should we do if we get a hw irq event on an illegal vector'.
* each architecture has to answer this themselves.
*/
void ack_bad_irq(int irq)
{
printk(KERN_WARNING "unexpected IRQ trap at vector %02x\n", irq);
}
/*
* change the level at which an IRQ executes
* - must not be called whilst interrupts are being processed!
*/
void set_intr_level(int irq, u16 level)
{
BUG_ON(in_interrupt());
__mask_and_set_icr(irq, GxICR_ENABLE, level);
}
void mn10300_intc_set_level(unsigned int irq, unsigned int level)
{
set_intr_level(irq, NUM2GxICR_LEVEL(level) & GxICR_LEVEL);
}
void mn10300_intc_clear(unsigned int irq)
{
__mask_and_set_icr(irq, GxICR_LEVEL | GxICR_ENABLE, GxICR_DETECT);
}
void mn10300_intc_set(unsigned int irq)
{
__mask_and_set_icr(irq, 0, GxICR_REQUEST | GxICR_DETECT);
}
void mn10300_intc_enable(unsigned int irq)
{
mn10300_cpupic_unmask(irq);
}
void mn10300_intc_disable(unsigned int irq)
{
mn10300_cpupic_mask(irq);
}
/*
* mark an interrupt to be ACK'd after interrupt handlers have been run rather
* than before
* - see Documentation/mn10300/features.txt
*/
void mn10300_set_lateack_irq_type(int irq)
{
set_irq_chip_and_handler(irq, &mn10300_cpu_pic_level,
handle_level_irq);
}
/*
* initialise the interrupt system
*/
void __init init_IRQ(void)
{
int irq;
for (irq = 0; irq < NR_IRQS; irq++)
if (irq_desc[irq].chip == &no_irq_chip)
/* due to the PIC latching interrupt requests, even
* when the IRQ is disabled, IRQ_PENDING is superfluous
* and we can use handle_level_irq() for edge-triggered
* interrupts */
set_irq_chip_and_handler(irq, &mn10300_cpu_pic_edge,
handle_level_irq);
unit_init_IRQ();
}
/*
* handle normal device IRQs
*/
asmlinkage void do_IRQ(void)
{
unsigned long sp, epsw, irq_disabled_epsw, old_irq_enabled_epsw;
unsigned int cpu_id = smp_processor_id();
int irq;
sp = current_stack_pointer();
BUG_ON(sp - (sp & ~(THREAD_SIZE - 1)) < STACK_WARN);
/* make sure local_irq_enable() doesn't muck up the interrupt priority
* setting in EPSW */
old_irq_enabled_epsw = __mn10300_irq_enabled_epsw[cpu_id];
local_save_flags(epsw);
__mn10300_irq_enabled_epsw[cpu_id] = EPSW_IE | (EPSW_IM & epsw);
irq_disabled_epsw = EPSW_IE | MN10300_CLI_LEVEL;
#ifdef CONFIG_MN10300_WD_TIMER
__IRQ_STAT(cpu_id, __irq_count)++;
#endif
irq_enter();
for (;;) {
/* ask the interrupt controller for the next IRQ to process
* - the result we get depends on EPSW.IM
*/
irq = IAGR & IAGR_GN;
if (!irq)
break;
local_irq_restore(irq_disabled_epsw);
generic_handle_irq(irq >> 2);
/* restore IRQ controls for IAGR access */
local_irq_restore(epsw);
}
__mn10300_irq_enabled_epsw[cpu_id] = old_irq_enabled_epsw;
irq_exit();
}
/*
* Display interrupt management information through /proc/interrupts
*/
int show_interrupts(struct seq_file *p, void *v)
{
int i = *(loff_t *) v, j, cpu;
struct irqaction *action;
unsigned long flags;
switch (i) {
/* display column title bar naming CPUs */
case 0:
seq_printf(p, " ");
for (j = 0; j < NR_CPUS; j++)
if (cpu_online(j))
seq_printf(p, "CPU%d ", j);
seq_putc(p, '\n');
break;
/* display information rows, one per active CPU */
case 1 ... NR_IRQS - 1:
raw_spin_lock_irqsave(&irq_desc[i].lock, flags);
action = irq_desc[i].action;
if (action) {
seq_printf(p, "%3d: ", i);
for_each_present_cpu(cpu)
seq_printf(p, "%10u ", kstat_irqs_cpu(i, cpu));
if (i < NR_CPU_IRQS)
seq_printf(p, " %14s.%u",
irq_desc[i].chip->name,
(GxICR(i) & GxICR_LEVEL) >>
GxICR_LEVEL_SHIFT);
else
seq_printf(p, " %14s",
irq_desc[i].chip->name);
seq_printf(p, " %s", action->name);
for (action = action->next;
action;
action = action->next)
seq_printf(p, ", %s", action->name);
seq_putc(p, '\n');
}
raw_spin_unlock_irqrestore(&irq_desc[i].lock, flags);
break;
/* polish off with NMI and error counters */
case NR_IRQS:
#ifdef CONFIG_MN10300_WD_TIMER
seq_printf(p, "NMI: ");
for (j = 0; j < NR_CPUS; j++)
if (cpu_online(j))
seq_printf(p, "%10u ", nmi_count(j));
seq_putc(p, '\n');
#endif
seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
break;
}
return 0;
}
#ifdef CONFIG_HOTPLUG_CPU
void migrate_irqs(void)
{
irq_desc_t *desc;
int irq;
unsigned int self, new;
unsigned long flags;
self = smp_processor_id();
for (irq = 0; irq < NR_IRQS; irq++) {
desc = irq_desc + irq;
if (desc->status == IRQ_PER_CPU)
continue;
if (cpu_isset(self, irq_desc[irq].affinity) &&
!cpus_intersects(irq_affinity[irq], cpu_online_map)) {
int cpu_id;
cpu_id = first_cpu(cpu_online_map);
cpu_set(cpu_id, irq_desc[irq].affinity);
}
/* We need to operate irq_affinity_online atomically. */
arch_local_cli_save(flags);
if (irq_affinity_online[irq] == self) {
u16 x, tmp;
x = GxICR(irq);
GxICR(irq) = x & GxICR_LEVEL;
tmp = GxICR(irq);
new = any_online_cpu(irq_desc[irq].affinity);
irq_affinity_online[irq] = new;
CROSS_GxICR(irq, new) =
(x & GxICR_LEVEL) | GxICR_DETECT;
tmp = CROSS_GxICR(irq, new);
x &= GxICR_LEVEL | GxICR_ENABLE;
if (GxICR(irq) & GxICR_REQUEST)
x |= GxICR_REQUEST | GxICR_DETECT;
CROSS_GxICR(irq, new) = x;
tmp = CROSS_GxICR(irq, new);
}
arch_local_irq_restore(flags);
}
}
#endif /* CONFIG_HOTPLUG_CPU */