OpenCloudOS-Kernel/arch/blackfin/mach-common/ints-priority.c

1367 lines
31 KiB
C

/*
* Set up the interrupt priorities
*
* Copyright 2004-2009 Analog Devices Inc.
* 2003 Bas Vermeulen <bas@buyways.nl>
* 2002 Arcturus Networks Inc. MaTed <mated@sympatico.ca>
* 2000-2001 Lineo, Inc. D. Jefff Dionne <jeff@lineo.ca>
* 1999 D. Jeff Dionne <jeff@uclinux.org>
* 1996 Roman Zippel
*
* Licensed under the GPL-2
*/
#include <linux/module.h>
#include <linux/kernel_stat.h>
#include <linux/seq_file.h>
#include <linux/irq.h>
#include <linux/sched.h>
#include <linux/sched/debug.h>
#include <linux/syscore_ops.h>
#include <asm/delay.h>
#ifdef CONFIG_IPIPE
#include <linux/ipipe.h>
#endif
#include <asm/traps.h>
#include <asm/blackfin.h>
#include <asm/irq_handler.h>
#include <asm/dpmc.h>
#include <asm/traps.h>
#include <asm/gpio.h>
/*
* NOTES:
* - we have separated the physical Hardware interrupt from the
* levels that the LINUX kernel sees (see the description in irq.h)
* -
*/
#ifndef CONFIG_SMP
/* Initialize this to an actual value to force it into the .data
* section so that we know it is properly initialized at entry into
* the kernel but before bss is initialized to zero (which is where
* it would live otherwise). The 0x1f magic represents the IRQs we
* cannot actually mask out in hardware.
*/
unsigned long bfin_irq_flags = 0x1f;
EXPORT_SYMBOL(bfin_irq_flags);
#endif
#ifdef CONFIG_PM
unsigned long bfin_sic_iwr[3]; /* Up to 3 SIC_IWRx registers */
unsigned vr_wakeup;
#endif
#ifndef SEC_GCTL
static struct ivgx {
/* irq number for request_irq, available in mach-bf5xx/irq.h */
unsigned int irqno;
/* corresponding bit in the SIC_ISR register */
unsigned int isrflag;
} ivg_table[NR_PERI_INTS];
static struct ivg_slice {
/* position of first irq in ivg_table for given ivg */
struct ivgx *ifirst;
struct ivgx *istop;
} ivg7_13[IVG13 - IVG7 + 1];
/*
* Search SIC_IAR and fill tables with the irqvalues
* and their positions in the SIC_ISR register.
*/
static void __init search_IAR(void)
{
unsigned ivg, irq_pos = 0;
for (ivg = 0; ivg <= IVG13 - IVG7; ivg++) {
int irqN;
ivg7_13[ivg].istop = ivg7_13[ivg].ifirst = &ivg_table[irq_pos];
for (irqN = 0; irqN < NR_PERI_INTS; irqN += 4) {
int irqn;
u32 iar =
bfin_read32((unsigned long *)SIC_IAR0 +
#if defined(CONFIG_BF51x) || defined(CONFIG_BF52x) || \
defined(CONFIG_BF538) || defined(CONFIG_BF539)
((irqN % 32) >> 3) + ((irqN / 32) * ((SIC_IAR4 - SIC_IAR0) / 4))
#else
(irqN >> 3)
#endif
);
for (irqn = irqN; irqn < irqN + 4; ++irqn) {
int iar_shift = (irqn & 7) * 4;
if (ivg == (0xf & (iar >> iar_shift))) {
ivg_table[irq_pos].irqno = IVG7 + irqn;
ivg_table[irq_pos].isrflag = 1 << (irqn % 32);
ivg7_13[ivg].istop++;
irq_pos++;
}
}
}
}
}
#endif
/*
* This is for core internal IRQs
*/
void bfin_ack_noop(struct irq_data *d)
{
/* Dummy function. */
}
static void bfin_core_mask_irq(struct irq_data *d)
{
bfin_irq_flags &= ~(1 << d->irq);
if (!hard_irqs_disabled())
hard_local_irq_enable();
}
static void bfin_core_unmask_irq(struct irq_data *d)
{
bfin_irq_flags |= 1 << d->irq;
/*
* If interrupts are enabled, IMASK must contain the same value
* as bfin_irq_flags. Make sure that invariant holds. If interrupts
* are currently disabled we need not do anything; one of the
* callers will take care of setting IMASK to the proper value
* when reenabling interrupts.
* local_irq_enable just does "STI bfin_irq_flags", so it's exactly
* what we need.
*/
if (!hard_irqs_disabled())
hard_local_irq_enable();
return;
}
#ifndef SEC_GCTL
void bfin_internal_mask_irq(unsigned int irq)
{
unsigned long flags = hard_local_irq_save();
#ifdef SIC_IMASK0
unsigned mask_bank = BFIN_SYSIRQ(irq) / 32;
unsigned mask_bit = BFIN_SYSIRQ(irq) % 32;
bfin_write_SIC_IMASK(mask_bank, bfin_read_SIC_IMASK(mask_bank) &
~(1 << mask_bit));
# if defined(CONFIG_SMP) || defined(CONFIG_ICC)
bfin_write_SICB_IMASK(mask_bank, bfin_read_SICB_IMASK(mask_bank) &
~(1 << mask_bit));
# endif
#else
bfin_write_SIC_IMASK(bfin_read_SIC_IMASK() &
~(1 << BFIN_SYSIRQ(irq)));
#endif /* end of SIC_IMASK0 */
hard_local_irq_restore(flags);
}
static void bfin_internal_mask_irq_chip(struct irq_data *d)
{
bfin_internal_mask_irq(d->irq);
}
#ifdef CONFIG_SMP
void bfin_internal_unmask_irq_affinity(unsigned int irq,
const struct cpumask *affinity)
#else
void bfin_internal_unmask_irq(unsigned int irq)
#endif
{
unsigned long flags = hard_local_irq_save();
#ifdef SIC_IMASK0
unsigned mask_bank = BFIN_SYSIRQ(irq) / 32;
unsigned mask_bit = BFIN_SYSIRQ(irq) % 32;
# ifdef CONFIG_SMP
if (cpumask_test_cpu(0, affinity))
# endif
bfin_write_SIC_IMASK(mask_bank,
bfin_read_SIC_IMASK(mask_bank) |
(1 << mask_bit));
# ifdef CONFIG_SMP
if (cpumask_test_cpu(1, affinity))
bfin_write_SICB_IMASK(mask_bank,
bfin_read_SICB_IMASK(mask_bank) |
(1 << mask_bit));
# endif
#else
bfin_write_SIC_IMASK(bfin_read_SIC_IMASK() |
(1 << BFIN_SYSIRQ(irq)));
#endif
hard_local_irq_restore(flags);
}
#ifdef CONFIG_SMP
static void bfin_internal_unmask_irq_chip(struct irq_data *d)
{
bfin_internal_unmask_irq_affinity(d->irq,
irq_data_get_affinity_mask(d));
}
static int bfin_internal_set_affinity(struct irq_data *d,
const struct cpumask *mask, bool force)
{
bfin_internal_mask_irq(d->irq);
bfin_internal_unmask_irq_affinity(d->irq, mask);
return 0;
}
#else
static void bfin_internal_unmask_irq_chip(struct irq_data *d)
{
bfin_internal_unmask_irq(d->irq);
}
#endif
#if defined(CONFIG_PM)
int bfin_internal_set_wake(unsigned int irq, unsigned int state)
{
u32 bank, bit, wakeup = 0;
unsigned long flags;
bank = BFIN_SYSIRQ(irq) / 32;
bit = BFIN_SYSIRQ(irq) % 32;
switch (irq) {
#ifdef IRQ_RTC
case IRQ_RTC:
wakeup |= WAKE;
break;
#endif
#ifdef IRQ_CAN0_RX
case IRQ_CAN0_RX:
wakeup |= CANWE;
break;
#endif
#ifdef IRQ_CAN1_RX
case IRQ_CAN1_RX:
wakeup |= CANWE;
break;
#endif
#ifdef IRQ_USB_INT0
case IRQ_USB_INT0:
wakeup |= USBWE;
break;
#endif
#ifdef CONFIG_BF54x
case IRQ_CNT:
wakeup |= ROTWE;
break;
#endif
default:
break;
}
flags = hard_local_irq_save();
if (state) {
bfin_sic_iwr[bank] |= (1 << bit);
vr_wakeup |= wakeup;
} else {
bfin_sic_iwr[bank] &= ~(1 << bit);
vr_wakeup &= ~wakeup;
}
hard_local_irq_restore(flags);
return 0;
}
static int bfin_internal_set_wake_chip(struct irq_data *d, unsigned int state)
{
return bfin_internal_set_wake(d->irq, state);
}
#else
inline int bfin_internal_set_wake(unsigned int irq, unsigned int state)
{
return 0;
}
# define bfin_internal_set_wake_chip NULL
#endif
#else /* SEC_GCTL */
static void bfin_sec_preflow_handler(struct irq_data *d)
{
unsigned long flags = hard_local_irq_save();
unsigned int sid = BFIN_SYSIRQ(d->irq);
bfin_write_SEC_SCI(0, SEC_CSID, sid);
hard_local_irq_restore(flags);
}
static void bfin_sec_mask_ack_irq(struct irq_data *d)
{
unsigned long flags = hard_local_irq_save();
unsigned int sid = BFIN_SYSIRQ(d->irq);
bfin_write_SEC_SCI(0, SEC_CSID, sid);
hard_local_irq_restore(flags);
}
static void bfin_sec_unmask_irq(struct irq_data *d)
{
unsigned long flags = hard_local_irq_save();
unsigned int sid = BFIN_SYSIRQ(d->irq);
bfin_write32(SEC_END, sid);
hard_local_irq_restore(flags);
}
static void bfin_sec_enable_ssi(unsigned int sid)
{
unsigned long flags = hard_local_irq_save();
uint32_t reg_sctl = bfin_read_SEC_SCTL(sid);
reg_sctl |= SEC_SCTL_SRC_EN;
bfin_write_SEC_SCTL(sid, reg_sctl);
hard_local_irq_restore(flags);
}
static void bfin_sec_disable_ssi(unsigned int sid)
{
unsigned long flags = hard_local_irq_save();
uint32_t reg_sctl = bfin_read_SEC_SCTL(sid);
reg_sctl &= ((uint32_t)~SEC_SCTL_SRC_EN);
bfin_write_SEC_SCTL(sid, reg_sctl);
hard_local_irq_restore(flags);
}
static void bfin_sec_set_ssi_coreid(unsigned int sid, unsigned int coreid)
{
unsigned long flags = hard_local_irq_save();
uint32_t reg_sctl = bfin_read_SEC_SCTL(sid);
reg_sctl &= ((uint32_t)~SEC_SCTL_CTG);
bfin_write_SEC_SCTL(sid, reg_sctl | ((coreid << 20) & SEC_SCTL_CTG));
hard_local_irq_restore(flags);
}
static void bfin_sec_enable_sci(unsigned int sid)
{
unsigned long flags = hard_local_irq_save();
uint32_t reg_sctl = bfin_read_SEC_SCTL(sid);
if (sid == BFIN_SYSIRQ(IRQ_WATCH0))
reg_sctl |= SEC_SCTL_FAULT_EN;
else
reg_sctl |= SEC_SCTL_INT_EN;
bfin_write_SEC_SCTL(sid, reg_sctl);
hard_local_irq_restore(flags);
}
static void bfin_sec_disable_sci(unsigned int sid)
{
unsigned long flags = hard_local_irq_save();
uint32_t reg_sctl = bfin_read_SEC_SCTL(sid);
reg_sctl &= ((uint32_t)~SEC_SCTL_INT_EN);
bfin_write_SEC_SCTL(sid, reg_sctl);
hard_local_irq_restore(flags);
}
static void bfin_sec_enable(struct irq_data *d)
{
unsigned long flags = hard_local_irq_save();
unsigned int sid = BFIN_SYSIRQ(d->irq);
bfin_sec_enable_sci(sid);
bfin_sec_enable_ssi(sid);
hard_local_irq_restore(flags);
}
static void bfin_sec_disable(struct irq_data *d)
{
unsigned long flags = hard_local_irq_save();
unsigned int sid = BFIN_SYSIRQ(d->irq);
bfin_sec_disable_sci(sid);
bfin_sec_disable_ssi(sid);
hard_local_irq_restore(flags);
}
static void bfin_sec_set_priority(unsigned int sec_int_levels, u8 *sec_int_priority)
{
unsigned long flags = hard_local_irq_save();
uint32_t reg_sctl;
int i;
bfin_write_SEC_SCI(0, SEC_CPLVL, sec_int_levels);
for (i = 0; i < SYS_IRQS - BFIN_IRQ(0); i++) {
reg_sctl = bfin_read_SEC_SCTL(i) & ~SEC_SCTL_PRIO;
reg_sctl |= sec_int_priority[i] << SEC_SCTL_PRIO_OFFSET;
bfin_write_SEC_SCTL(i, reg_sctl);
}
hard_local_irq_restore(flags);
}
void bfin_sec_raise_irq(unsigned int irq)
{
unsigned long flags = hard_local_irq_save();
unsigned int sid = BFIN_SYSIRQ(irq);
bfin_write32(SEC_RAISE, sid);
hard_local_irq_restore(flags);
}
static void init_software_driven_irq(void)
{
bfin_sec_set_ssi_coreid(34, 0);
bfin_sec_set_ssi_coreid(35, 1);
bfin_sec_enable_sci(35);
bfin_sec_enable_ssi(35);
bfin_sec_set_ssi_coreid(36, 0);
bfin_sec_set_ssi_coreid(37, 1);
bfin_sec_enable_sci(37);
bfin_sec_enable_ssi(37);
}
void handle_sec_sfi_fault(uint32_t gstat)
{
}
void handle_sec_sci_fault(uint32_t gstat)
{
uint32_t core_id;
uint32_t cstat;
core_id = gstat & SEC_GSTAT_SCI;
cstat = bfin_read_SEC_SCI(core_id, SEC_CSTAT);
if (cstat & SEC_CSTAT_ERR) {
switch (cstat & SEC_CSTAT_ERRC) {
case SEC_CSTAT_ACKERR:
printk(KERN_DEBUG "sec ack err\n");
break;
default:
printk(KERN_DEBUG "sec sci unknown err\n");
}
}
}
void handle_sec_ssi_fault(uint32_t gstat)
{
uint32_t sid;
uint32_t sstat;
sid = gstat & SEC_GSTAT_SID;
sstat = bfin_read_SEC_SSTAT(sid);
}
void handle_sec_fault(uint32_t sec_gstat)
{
if (sec_gstat & SEC_GSTAT_ERR) {
switch (sec_gstat & SEC_GSTAT_ERRC) {
case 0:
handle_sec_sfi_fault(sec_gstat);
break;
case SEC_GSTAT_SCIERR:
handle_sec_sci_fault(sec_gstat);
break;
case SEC_GSTAT_SSIERR:
handle_sec_ssi_fault(sec_gstat);
break;
}
}
}
static struct irqaction bfin_fault_irq = {
.name = "Blackfin fault",
};
static irqreturn_t bfin_fault_routine(int irq, void *data)
{
struct pt_regs *fp = get_irq_regs();
switch (irq) {
case IRQ_C0_DBL_FAULT:
double_fault_c(fp);
break;
case IRQ_C0_HW_ERR:
dump_bfin_process(fp);
dump_bfin_mem(fp);
show_regs(fp);
printk(KERN_NOTICE "Kernel Stack\n");
show_stack(current, NULL);
print_modules();
panic("Core 0 hardware error");
break;
case IRQ_C0_NMI_L1_PARITY_ERR:
panic("Core 0 NMI L1 parity error");
break;
case IRQ_SEC_ERR:
pr_err("SEC error\n");
handle_sec_fault(bfin_read32(SEC_GSTAT));
break;
default:
panic("Unknown fault %d", irq);
}
return IRQ_HANDLED;
}
#endif /* SEC_GCTL */
static struct irq_chip bfin_core_irqchip = {
.name = "CORE",
.irq_mask = bfin_core_mask_irq,
.irq_unmask = bfin_core_unmask_irq,
};
#ifndef SEC_GCTL
static struct irq_chip bfin_internal_irqchip = {
.name = "INTN",
.irq_mask = bfin_internal_mask_irq_chip,
.irq_unmask = bfin_internal_unmask_irq_chip,
.irq_disable = bfin_internal_mask_irq_chip,
.irq_enable = bfin_internal_unmask_irq_chip,
#ifdef CONFIG_SMP
.irq_set_affinity = bfin_internal_set_affinity,
#endif
.irq_set_wake = bfin_internal_set_wake_chip,
};
#else
static struct irq_chip bfin_sec_irqchip = {
.name = "SEC",
.irq_mask_ack = bfin_sec_mask_ack_irq,
.irq_mask = bfin_sec_mask_ack_irq,
.irq_unmask = bfin_sec_unmask_irq,
.irq_eoi = bfin_sec_unmask_irq,
.irq_disable = bfin_sec_disable,
.irq_enable = bfin_sec_enable,
};
#endif
void bfin_handle_irq(unsigned irq)
{
#ifdef CONFIG_IPIPE
struct pt_regs regs; /* Contents not used. */
ipipe_trace_irq_entry(irq);
__ipipe_handle_irq(irq, &regs);
ipipe_trace_irq_exit(irq);
#else /* !CONFIG_IPIPE */
generic_handle_irq(irq);
#endif /* !CONFIG_IPIPE */
}
#if defined(CONFIG_BFIN_MAC) || defined(CONFIG_BFIN_MAC_MODULE)
static int mac_stat_int_mask;
static void bfin_mac_status_ack_irq(unsigned int irq)
{
switch (irq) {
case IRQ_MAC_MMCINT:
bfin_write_EMAC_MMC_TIRQS(
bfin_read_EMAC_MMC_TIRQE() &
bfin_read_EMAC_MMC_TIRQS());
bfin_write_EMAC_MMC_RIRQS(
bfin_read_EMAC_MMC_RIRQE() &
bfin_read_EMAC_MMC_RIRQS());
break;
case IRQ_MAC_RXFSINT:
bfin_write_EMAC_RX_STKY(
bfin_read_EMAC_RX_IRQE() &
bfin_read_EMAC_RX_STKY());
break;
case IRQ_MAC_TXFSINT:
bfin_write_EMAC_TX_STKY(
bfin_read_EMAC_TX_IRQE() &
bfin_read_EMAC_TX_STKY());
break;
case IRQ_MAC_WAKEDET:
bfin_write_EMAC_WKUP_CTL(
bfin_read_EMAC_WKUP_CTL() | MPKS | RWKS);
break;
default:
/* These bits are W1C */
bfin_write_EMAC_SYSTAT(1L << (irq - IRQ_MAC_PHYINT));
break;
}
}
static void bfin_mac_status_mask_irq(struct irq_data *d)
{
unsigned int irq = d->irq;
mac_stat_int_mask &= ~(1L << (irq - IRQ_MAC_PHYINT));
#ifdef BF537_FAMILY
switch (irq) {
case IRQ_MAC_PHYINT:
bfin_write_EMAC_SYSCTL(bfin_read_EMAC_SYSCTL() & ~PHYIE);
break;
default:
break;
}
#else
if (!mac_stat_int_mask)
bfin_internal_mask_irq(IRQ_MAC_ERROR);
#endif
bfin_mac_status_ack_irq(irq);
}
static void bfin_mac_status_unmask_irq(struct irq_data *d)
{
unsigned int irq = d->irq;
#ifdef BF537_FAMILY
switch (irq) {
case IRQ_MAC_PHYINT:
bfin_write_EMAC_SYSCTL(bfin_read_EMAC_SYSCTL() | PHYIE);
break;
default:
break;
}
#else
if (!mac_stat_int_mask)
bfin_internal_unmask_irq(IRQ_MAC_ERROR);
#endif
mac_stat_int_mask |= 1L << (irq - IRQ_MAC_PHYINT);
}
#ifdef CONFIG_PM
int bfin_mac_status_set_wake(struct irq_data *d, unsigned int state)
{
#ifdef BF537_FAMILY
return bfin_internal_set_wake(IRQ_GENERIC_ERROR, state);
#else
return bfin_internal_set_wake(IRQ_MAC_ERROR, state);
#endif
}
#else
# define bfin_mac_status_set_wake NULL
#endif
static struct irq_chip bfin_mac_status_irqchip = {
.name = "MACST",
.irq_mask = bfin_mac_status_mask_irq,
.irq_unmask = bfin_mac_status_unmask_irq,
.irq_set_wake = bfin_mac_status_set_wake,
};
void bfin_demux_mac_status_irq(struct irq_desc *inta_desc)
{
int i, irq = 0;
u32 status = bfin_read_EMAC_SYSTAT();
for (i = 0; i <= (IRQ_MAC_STMDONE - IRQ_MAC_PHYINT); i++)
if (status & (1L << i)) {
irq = IRQ_MAC_PHYINT + i;
break;
}
if (irq) {
if (mac_stat_int_mask & (1L << (irq - IRQ_MAC_PHYINT))) {
bfin_handle_irq(irq);
} else {
bfin_mac_status_ack_irq(irq);
pr_debug("IRQ %d:"
" MASKED MAC ERROR INTERRUPT ASSERTED\n",
irq);
}
} else
printk(KERN_ERR
"%s : %s : LINE %d :\nIRQ ?: MAC ERROR"
" INTERRUPT ASSERTED BUT NO SOURCE FOUND"
"(EMAC_SYSTAT=0x%X)\n",
__func__, __FILE__, __LINE__, status);
}
#endif
static inline void bfin_set_irq_handler(struct irq_data *d, irq_flow_handler_t handle)
{
#ifdef CONFIG_IPIPE
handle = handle_level_irq;
#endif
irq_set_handler_locked(d, handle);
}
#ifdef CONFIG_GPIO_ADI
static DECLARE_BITMAP(gpio_enabled, MAX_BLACKFIN_GPIOS);
static void bfin_gpio_ack_irq(struct irq_data *d)
{
/* AFAIK ack_irq in case mask_ack is provided
* get's only called for edge sense irqs
*/
set_gpio_data(irq_to_gpio(d->irq), 0);
}
static void bfin_gpio_mask_ack_irq(struct irq_data *d)
{
unsigned int irq = d->irq;
u32 gpionr = irq_to_gpio(irq);
if (!irqd_is_level_type(d))
set_gpio_data(gpionr, 0);
set_gpio_maska(gpionr, 0);
}
static void bfin_gpio_mask_irq(struct irq_data *d)
{
set_gpio_maska(irq_to_gpio(d->irq), 0);
}
static void bfin_gpio_unmask_irq(struct irq_data *d)
{
set_gpio_maska(irq_to_gpio(d->irq), 1);
}
static unsigned int bfin_gpio_irq_startup(struct irq_data *d)
{
u32 gpionr = irq_to_gpio(d->irq);
if (__test_and_set_bit(gpionr, gpio_enabled))
bfin_gpio_irq_prepare(gpionr);
bfin_gpio_unmask_irq(d);
return 0;
}
static void bfin_gpio_irq_shutdown(struct irq_data *d)
{
u32 gpionr = irq_to_gpio(d->irq);
bfin_gpio_mask_irq(d);
__clear_bit(gpionr, gpio_enabled);
bfin_gpio_irq_free(gpionr);
}
static int bfin_gpio_irq_type(struct irq_data *d, unsigned int type)
{
unsigned int irq = d->irq;
int ret;
char buf[16];
u32 gpionr = irq_to_gpio(irq);
if (type == IRQ_TYPE_PROBE) {
/* only probe unenabled GPIO interrupt lines */
if (test_bit(gpionr, gpio_enabled))
return 0;
type = IRQ_TYPE_EDGE_RISING | IRQ_TYPE_EDGE_FALLING;
}
if (type & (IRQ_TYPE_EDGE_RISING | IRQ_TYPE_EDGE_FALLING |
IRQ_TYPE_LEVEL_HIGH | IRQ_TYPE_LEVEL_LOW)) {
snprintf(buf, 16, "gpio-irq%d", irq);
ret = bfin_gpio_irq_request(gpionr, buf);
if (ret)
return ret;
if (__test_and_set_bit(gpionr, gpio_enabled))
bfin_gpio_irq_prepare(gpionr);
} else {
__clear_bit(gpionr, gpio_enabled);
return 0;
}
set_gpio_inen(gpionr, 0);
set_gpio_dir(gpionr, 0);
if ((type & (IRQ_TYPE_EDGE_RISING | IRQ_TYPE_EDGE_FALLING))
== (IRQ_TYPE_EDGE_RISING | IRQ_TYPE_EDGE_FALLING))
set_gpio_both(gpionr, 1);
else
set_gpio_both(gpionr, 0);
if ((type & (IRQ_TYPE_EDGE_FALLING | IRQ_TYPE_LEVEL_LOW)))
set_gpio_polar(gpionr, 1); /* low or falling edge denoted by one */
else
set_gpio_polar(gpionr, 0); /* high or rising edge denoted by zero */
if (type & (IRQ_TYPE_EDGE_RISING | IRQ_TYPE_EDGE_FALLING)) {
set_gpio_edge(gpionr, 1);
set_gpio_inen(gpionr, 1);
set_gpio_data(gpionr, 0);
} else {
set_gpio_edge(gpionr, 0);
set_gpio_inen(gpionr, 1);
}
if (type & (IRQ_TYPE_EDGE_RISING | IRQ_TYPE_EDGE_FALLING))
bfin_set_irq_handler(d, handle_edge_irq);
else
bfin_set_irq_handler(d, handle_level_irq);
return 0;
}
static void bfin_demux_gpio_block(unsigned int irq)
{
unsigned int gpio, mask;
gpio = irq_to_gpio(irq);
mask = get_gpiop_data(gpio) & get_gpiop_maska(gpio);
while (mask) {
if (mask & 1)
bfin_handle_irq(irq);
irq++;
mask >>= 1;
}
}
void bfin_demux_gpio_irq(struct irq_desc *desc)
{
unsigned int inta_irq = irq_desc_get_irq(desc);
unsigned int irq;
switch (inta_irq) {
#if defined(BF537_FAMILY)
case IRQ_PF_INTA_PG_INTA:
bfin_demux_gpio_block(IRQ_PF0);
irq = IRQ_PG0;
break;
case IRQ_PH_INTA_MAC_RX:
irq = IRQ_PH0;
break;
#elif defined(BF533_FAMILY)
case IRQ_PROG_INTA:
irq = IRQ_PF0;
break;
#elif defined(BF538_FAMILY)
case IRQ_PORTF_INTA:
irq = IRQ_PF0;
break;
#elif defined(CONFIG_BF52x) || defined(CONFIG_BF51x)
case IRQ_PORTF_INTA:
irq = IRQ_PF0;
break;
case IRQ_PORTG_INTA:
irq = IRQ_PG0;
break;
case IRQ_PORTH_INTA:
irq = IRQ_PH0;
break;
#elif defined(CONFIG_BF561)
case IRQ_PROG0_INTA:
irq = IRQ_PF0;
break;
case IRQ_PROG1_INTA:
irq = IRQ_PF16;
break;
case IRQ_PROG2_INTA:
irq = IRQ_PF32;
break;
#endif
default:
BUG();
return;
}
bfin_demux_gpio_block(irq);
}
#ifdef CONFIG_PM
static int bfin_gpio_set_wake(struct irq_data *d, unsigned int state)
{
return bfin_gpio_pm_wakeup_ctrl(irq_to_gpio(d->irq), state);
}
#else
# define bfin_gpio_set_wake NULL
#endif
static struct irq_chip bfin_gpio_irqchip = {
.name = "GPIO",
.irq_ack = bfin_gpio_ack_irq,
.irq_mask = bfin_gpio_mask_irq,
.irq_mask_ack = bfin_gpio_mask_ack_irq,
.irq_unmask = bfin_gpio_unmask_irq,
.irq_disable = bfin_gpio_mask_irq,
.irq_enable = bfin_gpio_unmask_irq,
.irq_set_type = bfin_gpio_irq_type,
.irq_startup = bfin_gpio_irq_startup,
.irq_shutdown = bfin_gpio_irq_shutdown,
.irq_set_wake = bfin_gpio_set_wake,
};
#endif
#ifdef CONFIG_PM
#ifdef SEC_GCTL
static u32 save_pint_sec_ctl[NR_PINT_SYS_IRQS];
static int sec_suspend(void)
{
u32 bank;
for (bank = 0; bank < NR_PINT_SYS_IRQS; bank++)
save_pint_sec_ctl[bank] = bfin_read_SEC_SCTL(bank + BFIN_SYSIRQ(IRQ_PINT0));
return 0;
}
static void sec_resume(void)
{
u32 bank;
bfin_write_SEC_SCI(0, SEC_CCTL, SEC_CCTL_RESET);
udelay(100);
bfin_write_SEC_GCTL(SEC_GCTL_EN);
bfin_write_SEC_SCI(0, SEC_CCTL, SEC_CCTL_EN | SEC_CCTL_NMI_EN);
for (bank = 0; bank < NR_PINT_SYS_IRQS; bank++)
bfin_write_SEC_SCTL(bank + BFIN_SYSIRQ(IRQ_PINT0), save_pint_sec_ctl[bank]);
}
static struct syscore_ops sec_pm_syscore_ops = {
.suspend = sec_suspend,
.resume = sec_resume,
};
#endif
#endif
void init_exception_vectors(void)
{
/* cannot program in software:
* evt0 - emulation (jtag)
* evt1 - reset
*/
bfin_write_EVT2(evt_nmi);
bfin_write_EVT3(trap);
bfin_write_EVT5(evt_ivhw);
bfin_write_EVT6(evt_timer);
bfin_write_EVT7(evt_evt7);
bfin_write_EVT8(evt_evt8);
bfin_write_EVT9(evt_evt9);
bfin_write_EVT10(evt_evt10);
bfin_write_EVT11(evt_evt11);
bfin_write_EVT12(evt_evt12);
bfin_write_EVT13(evt_evt13);
bfin_write_EVT14(evt_evt14);
bfin_write_EVT15(evt_system_call);
CSYNC();
}
#ifndef SEC_GCTL
/*
* This function should be called during kernel startup to initialize
* the BFin IRQ handling routines.
*/
int __init init_arch_irq(void)
{
int irq;
unsigned long ilat = 0;
/* Disable all the peripheral intrs - page 4-29 HW Ref manual */
#ifdef SIC_IMASK0
bfin_write_SIC_IMASK0(SIC_UNMASK_ALL);
bfin_write_SIC_IMASK1(SIC_UNMASK_ALL);
# ifdef SIC_IMASK2
bfin_write_SIC_IMASK2(SIC_UNMASK_ALL);
# endif
# if defined(CONFIG_SMP) || defined(CONFIG_ICC)
bfin_write_SICB_IMASK0(SIC_UNMASK_ALL);
bfin_write_SICB_IMASK1(SIC_UNMASK_ALL);
# endif
#else
bfin_write_SIC_IMASK(SIC_UNMASK_ALL);
#endif
local_irq_disable();
for (irq = 0; irq <= SYS_IRQS; irq++) {
if (irq <= IRQ_CORETMR)
irq_set_chip(irq, &bfin_core_irqchip);
else
irq_set_chip(irq, &bfin_internal_irqchip);
switch (irq) {
#if !BFIN_GPIO_PINT
#if defined(BF537_FAMILY)
case IRQ_PH_INTA_MAC_RX:
case IRQ_PF_INTA_PG_INTA:
#elif defined(BF533_FAMILY)
case IRQ_PROG_INTA:
#elif defined(CONFIG_BF52x) || defined(CONFIG_BF51x)
case IRQ_PORTF_INTA:
case IRQ_PORTG_INTA:
case IRQ_PORTH_INTA:
#elif defined(CONFIG_BF561)
case IRQ_PROG0_INTA:
case IRQ_PROG1_INTA:
case IRQ_PROG2_INTA:
#elif defined(BF538_FAMILY)
case IRQ_PORTF_INTA:
#endif
irq_set_chained_handler(irq, bfin_demux_gpio_irq);
break;
#endif
#if defined(CONFIG_BFIN_MAC) || defined(CONFIG_BFIN_MAC_MODULE)
case IRQ_MAC_ERROR:
irq_set_chained_handler(irq,
bfin_demux_mac_status_irq);
break;
#endif
#if defined(CONFIG_SMP) || defined(CONFIG_ICC)
case IRQ_SUPPLE_0:
case IRQ_SUPPLE_1:
irq_set_handler(irq, handle_percpu_irq);
break;
#endif
#ifdef CONFIG_TICKSOURCE_CORETMR
case IRQ_CORETMR:
# ifdef CONFIG_SMP
irq_set_handler(irq, handle_percpu_irq);
# else
irq_set_handler(irq, handle_simple_irq);
# endif
break;
#endif
#ifdef CONFIG_TICKSOURCE_GPTMR0
case IRQ_TIMER0:
irq_set_handler(irq, handle_simple_irq);
break;
#endif
default:
#ifdef CONFIG_IPIPE
irq_set_handler(irq, handle_level_irq);
#else
irq_set_handler(irq, handle_simple_irq);
#endif
break;
}
}
init_mach_irq();
#if (defined(CONFIG_BFIN_MAC) || defined(CONFIG_BFIN_MAC_MODULE))
for (irq = IRQ_MAC_PHYINT; irq <= IRQ_MAC_STMDONE; irq++)
irq_set_chip_and_handler(irq, &bfin_mac_status_irqchip,
handle_level_irq);
#endif
/* if configured as edge, then will be changed to do_edge_IRQ */
#ifdef CONFIG_GPIO_ADI
for (irq = GPIO_IRQ_BASE;
irq < (GPIO_IRQ_BASE + MAX_BLACKFIN_GPIOS); irq++)
irq_set_chip_and_handler(irq, &bfin_gpio_irqchip,
handle_level_irq);
#endif
bfin_write_IMASK(0);
CSYNC();
ilat = bfin_read_ILAT();
CSYNC();
bfin_write_ILAT(ilat);
CSYNC();
printk(KERN_INFO "Configuring Blackfin Priority Driven Interrupts\n");
/* IMASK=xxx is equivalent to STI xx or bfin_irq_flags=xx,
* local_irq_enable()
*/
program_IAR();
/* Therefore it's better to setup IARs before interrupts enabled */
search_IAR();
/* Enable interrupts IVG7-15 */
bfin_irq_flags |= IMASK_IVG15 |
IMASK_IVG14 | IMASK_IVG13 | IMASK_IVG12 | IMASK_IVG11 |
IMASK_IVG10 | IMASK_IVG9 | IMASK_IVG8 | IMASK_IVG7 | IMASK_IVGHW;
/* This implicitly covers ANOMALY_05000171
* Boot-ROM code modifies SICA_IWRx wakeup registers
*/
#ifdef SIC_IWR0
bfin_write_SIC_IWR0(IWR_DISABLE_ALL);
# ifdef SIC_IWR1
/* BF52x/BF51x system reset does not properly reset SIC_IWR1 which
* will screw up the bootrom as it relies on MDMA0/1 waking it
* up from IDLE instructions. See this report for more info:
* http://blackfin.uclinux.org/gf/tracker/4323
*/
if (ANOMALY_05000435)
bfin_write_SIC_IWR1(IWR_ENABLE(10) | IWR_ENABLE(11));
else
bfin_write_SIC_IWR1(IWR_DISABLE_ALL);
# endif
# ifdef SIC_IWR2
bfin_write_SIC_IWR2(IWR_DISABLE_ALL);
# endif
#else
bfin_write_SIC_IWR(IWR_DISABLE_ALL);
#endif
return 0;
}
#ifdef CONFIG_DO_IRQ_L1
__attribute__((l1_text))
#endif
static int vec_to_irq(int vec)
{
struct ivgx *ivg = ivg7_13[vec - IVG7].ifirst;
struct ivgx *ivg_stop = ivg7_13[vec - IVG7].istop;
unsigned long sic_status[3];
if (likely(vec == EVT_IVTMR_P))
return IRQ_CORETMR;
#ifdef SIC_ISR
sic_status[0] = bfin_read_SIC_IMASK() & bfin_read_SIC_ISR();
#else
if (smp_processor_id()) {
# ifdef SICB_ISR0
/* This will be optimized out in UP mode. */
sic_status[0] = bfin_read_SICB_ISR0() & bfin_read_SICB_IMASK0();
sic_status[1] = bfin_read_SICB_ISR1() & bfin_read_SICB_IMASK1();
# endif
} else {
sic_status[0] = bfin_read_SIC_ISR0() & bfin_read_SIC_IMASK0();
sic_status[1] = bfin_read_SIC_ISR1() & bfin_read_SIC_IMASK1();
}
#endif
#ifdef SIC_ISR2
sic_status[2] = bfin_read_SIC_ISR2() & bfin_read_SIC_IMASK2();
#endif
for (;; ivg++) {
if (ivg >= ivg_stop)
return -1;
#ifdef SIC_ISR
if (sic_status[0] & ivg->isrflag)
#else
if (sic_status[(ivg->irqno - IVG7) / 32] & ivg->isrflag)
#endif
return ivg->irqno;
}
}
#else /* SEC_GCTL */
/*
* This function should be called during kernel startup to initialize
* the BFin IRQ handling routines.
*/
int __init init_arch_irq(void)
{
int irq;
unsigned long ilat = 0;
bfin_write_SEC_GCTL(SEC_GCTL_RESET);
local_irq_disable();
for (irq = 0; irq <= SYS_IRQS; irq++) {
if (irq <= IRQ_CORETMR) {
irq_set_chip_and_handler(irq, &bfin_core_irqchip,
handle_simple_irq);
#if defined(CONFIG_TICKSOURCE_CORETMR) && defined(CONFIG_SMP)
if (irq == IRQ_CORETMR)
irq_set_handler(irq, handle_percpu_irq);
#endif
} else if (irq >= BFIN_IRQ(34) && irq <= BFIN_IRQ(37)) {
irq_set_chip_and_handler(irq, &bfin_sec_irqchip,
handle_percpu_irq);
} else {
irq_set_chip(irq, &bfin_sec_irqchip);
irq_set_handler(irq, handle_fasteoi_irq);
__irq_set_preflow_handler(irq, bfin_sec_preflow_handler);
}
}
bfin_write_IMASK(0);
CSYNC();
ilat = bfin_read_ILAT();
CSYNC();
bfin_write_ILAT(ilat);
CSYNC();
printk(KERN_INFO "Configuring Blackfin Priority Driven Interrupts\n");
bfin_sec_set_priority(CONFIG_SEC_IRQ_PRIORITY_LEVELS, sec_int_priority);
/* Enable interrupts IVG7-15 */
bfin_irq_flags |= IMASK_IVG15 |
IMASK_IVG14 | IMASK_IVG13 | IMASK_IVG12 | IMASK_IVG11 |
IMASK_IVG10 | IMASK_IVG9 | IMASK_IVG8 | IMASK_IVG7 | IMASK_IVGHW;
bfin_write_SEC_FCTL(SEC_FCTL_EN | SEC_FCTL_SYSRST_EN | SEC_FCTL_FLTIN_EN);
bfin_sec_enable_sci(BFIN_SYSIRQ(IRQ_WATCH0));
bfin_sec_enable_ssi(BFIN_SYSIRQ(IRQ_WATCH0));
bfin_write_SEC_SCI(0, SEC_CCTL, SEC_CCTL_RESET);
udelay(100);
bfin_write_SEC_GCTL(SEC_GCTL_EN);
bfin_write_SEC_SCI(0, SEC_CCTL, SEC_CCTL_EN | SEC_CCTL_NMI_EN);
bfin_write_SEC_SCI(1, SEC_CCTL, SEC_CCTL_EN | SEC_CCTL_NMI_EN);
init_software_driven_irq();
#ifdef CONFIG_PM
register_syscore_ops(&sec_pm_syscore_ops);
#endif
bfin_fault_irq.handler = bfin_fault_routine;
#ifdef CONFIG_L1_PARITY_CHECK
setup_irq(IRQ_C0_NMI_L1_PARITY_ERR, &bfin_fault_irq);
#endif
setup_irq(IRQ_C0_DBL_FAULT, &bfin_fault_irq);
setup_irq(IRQ_SEC_ERR, &bfin_fault_irq);
return 0;
}
#ifdef CONFIG_DO_IRQ_L1
__attribute__((l1_text))
#endif
static int vec_to_irq(int vec)
{
if (likely(vec == EVT_IVTMR_P))
return IRQ_CORETMR;
return BFIN_IRQ(bfin_read_SEC_SCI(0, SEC_CSID));
}
#endif /* SEC_GCTL */
#ifdef CONFIG_DO_IRQ_L1
__attribute__((l1_text))
#endif
void do_irq(int vec, struct pt_regs *fp)
{
int irq = vec_to_irq(vec);
if (irq == -1)
return;
asm_do_IRQ(irq, fp);
}
#ifdef CONFIG_IPIPE
int __ipipe_get_irq_priority(unsigned irq)
{
int ient, prio;
if (irq <= IRQ_CORETMR)
return irq;
#ifdef SEC_GCTL
if (irq >= BFIN_IRQ(0))
return IVG11;
#else
for (ient = 0; ient < NR_PERI_INTS; ient++) {
struct ivgx *ivg = ivg_table + ient;
if (ivg->irqno == irq) {
for (prio = 0; prio <= IVG13-IVG7; prio++) {
if (ivg7_13[prio].ifirst <= ivg &&
ivg7_13[prio].istop > ivg)
return IVG7 + prio;
}
}
}
#endif
return IVG15;
}
/* Hw interrupts are disabled on entry (check SAVE_CONTEXT). */
#ifdef CONFIG_DO_IRQ_L1
__attribute__((l1_text))
#endif
asmlinkage int __ipipe_grab_irq(int vec, struct pt_regs *regs)
{
struct ipipe_percpu_domain_data *p = ipipe_root_cpudom_ptr();
struct ipipe_domain *this_domain = __ipipe_current_domain;
int irq, s = 0;
irq = vec_to_irq(vec);
if (irq == -1)
return 0;
if (irq == IRQ_SYSTMR) {
#if !defined(CONFIG_GENERIC_CLOCKEVENTS) || defined(CONFIG_TICKSOURCE_GPTMR0)
bfin_write_TIMER_STATUS(1); /* Latch TIMIL0 */
#endif
/* This is basically what we need from the register frame. */
__this_cpu_write(__ipipe_tick_regs.ipend, regs->ipend);
__this_cpu_write(__ipipe_tick_regs.pc, regs->pc);
if (this_domain != ipipe_root_domain)
__this_cpu_and(__ipipe_tick_regs.ipend, ~0x10);
else
__this_cpu_or(__ipipe_tick_regs.ipend, 0x10);
}
/*
* We don't want Linux interrupt handlers to run at the
* current core priority level (i.e. < EVT15), since this
* might delay other interrupts handled by a high priority
* domain. Here is what we do instead:
*
* - we raise the SYNCDEFER bit to prevent
* __ipipe_handle_irq() to sync the pipeline for the root
* stage for the incoming interrupt. Upon return, that IRQ is
* pending in the interrupt log.
*
* - we raise the TIF_IRQ_SYNC bit for the current thread, so
* that _schedule_and_signal_from_int will eventually sync the
* pipeline from EVT15.
*/
if (this_domain == ipipe_root_domain) {
s = __test_and_set_bit(IPIPE_SYNCDEFER_FLAG, &p->status);
barrier();
}
ipipe_trace_irq_entry(irq);
__ipipe_handle_irq(irq, regs);
ipipe_trace_irq_exit(irq);
if (user_mode(regs) &&
!ipipe_test_foreign_stack() &&
(current->ipipe_flags & PF_EVTRET) != 0) {
/*
* Testing for user_regs() does NOT fully eliminate
* foreign stack contexts, because of the forged
* interrupt returns we do through
* __ipipe_call_irqtail. In that case, we might have
* preempted a foreign stack context in a high
* priority domain, with a single interrupt level now
* pending after the irqtail unwinding is done. In
* which case user_mode() is now true, and the event
* gets dispatched spuriously.
*/
current->ipipe_flags &= ~PF_EVTRET;
__ipipe_dispatch_event(IPIPE_EVENT_RETURN, regs);
}
if (this_domain == ipipe_root_domain) {
set_thread_flag(TIF_IRQ_SYNC);
if (!s) {
__clear_bit(IPIPE_SYNCDEFER_FLAG, &p->status);
return !test_bit(IPIPE_STALL_FLAG, &p->status);
}
}
return 0;
}
#endif /* CONFIG_IPIPE */