linux/arch/powerpc/platforms/powermac/pic.c

748 lines
20 KiB
C
Raw Normal View History

/*
* Support for the interrupt controllers found on Power Macintosh,
* currently Apple's "Grand Central" interrupt controller in all
* it's incarnations. OpenPIC support used on newer machines is
* in a separate file
*
* Copyright (C) 1997 Paul Mackerras (paulus@samba.org)
* Copyright (C) 2005 Benjamin Herrenschmidt (benh@kernel.crashing.org)
* IBM, Corp.
*
* 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/stddef.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/signal.h>
#include <linux/pci.h>
#include <linux/interrupt.h>
#include <linux/sysdev.h>
#include <linux/adb.h>
#include <linux/pmu.h>
#include <linux/module.h>
#include <asm/sections.h>
#include <asm/io.h>
#include <asm/smp.h>
#include <asm/prom.h>
#include <asm/pci-bridge.h>
#include <asm/time.h>
#include <asm/pmac_feature.h>
#include <asm/mpic.h>
#include <asm/xmon.h>
#include "pmac.h"
#ifdef CONFIG_PPC32
struct pmac_irq_hw {
unsigned int event;
unsigned int enable;
unsigned int ack;
unsigned int level;
};
/* Workaround flags for 32bit powermac machines */
unsigned int of_irq_workarounds;
struct device_node *of_irq_dflt_pic;
/* Default addresses */
static volatile struct pmac_irq_hw __iomem *pmac_irq_hw[4];
#define GC_LEVEL_MASK 0x3ff00000
#define OHARE_LEVEL_MASK 0x1ff00000
#define HEATHROW_LEVEL_MASK 0x1ff00000
static int max_irqs;
static int max_real_irqs;
static u32 level_mask[4];
static DEFINE_RAW_SPINLOCK(pmac_pic_lock);
#define NR_MASK_WORDS ((NR_IRQS + 31) / 32)
static unsigned long ppc_lost_interrupts[NR_MASK_WORDS];
static unsigned long ppc_cached_irq_mask[NR_MASK_WORDS];
static int pmac_irq_cascade = -1;
2006-07-03 11:36:01 +00:00
static struct irq_host *pmac_pic_host;
static void __pmac_retrigger(unsigned int irq_nr)
{
if (irq_nr >= max_real_irqs && pmac_irq_cascade > 0) {
__set_bit(irq_nr, ppc_lost_interrupts);
irq_nr = pmac_irq_cascade;
mb();
}
if (!__test_and_set_bit(irq_nr, ppc_lost_interrupts)) {
atomic_inc(&ppc_n_lost_interrupts);
set_dec(1);
}
}
static void pmac_mask_and_ack_irq(struct irq_data *d)
{
unsigned int src = irqd_to_hwirq(d);
unsigned long bit = 1UL << (src & 0x1f);
int i = src >> 5;
unsigned long flags;
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
2006-07-03 11:36:01 +00:00
__clear_bit(src, ppc_cached_irq_mask);
if (__test_and_clear_bit(src, ppc_lost_interrupts))
atomic_dec(&ppc_n_lost_interrupts);
out_le32(&pmac_irq_hw[i]->enable, ppc_cached_irq_mask[i]);
out_le32(&pmac_irq_hw[i]->ack, bit);
do {
/* make sure ack gets to controller before we enable
interrupts */
mb();
} while((in_le32(&pmac_irq_hw[i]->enable) & bit)
!= (ppc_cached_irq_mask[i] & bit));
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
}
static void pmac_ack_irq(struct irq_data *d)
{
unsigned int src = irqd_to_hwirq(d);
2006-07-03 11:36:01 +00:00
unsigned long bit = 1UL << (src & 0x1f);
int i = src >> 5;
unsigned long flags;
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
2006-07-03 11:36:01 +00:00
if (__test_and_clear_bit(src, ppc_lost_interrupts))
atomic_dec(&ppc_n_lost_interrupts);
out_le32(&pmac_irq_hw[i]->ack, bit);
(void)in_le32(&pmac_irq_hw[i]->ack);
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
}
static void __pmac_set_irq_mask(unsigned int irq_nr, int nokicklost)
{
unsigned long bit = 1UL << (irq_nr & 0x1f);
int i = irq_nr >> 5;
if ((unsigned)irq_nr >= max_irqs)
return;
/* enable unmasked interrupts */
out_le32(&pmac_irq_hw[i]->enable, ppc_cached_irq_mask[i]);
do {
/* make sure mask gets to controller before we
return to user */
mb();
} while((in_le32(&pmac_irq_hw[i]->enable) & bit)
!= (ppc_cached_irq_mask[i] & bit));
/*
* Unfortunately, setting the bit in the enable register
* when the device interrupt is already on *doesn't* set
* the bit in the flag register or request another interrupt.
*/
if (bit & ppc_cached_irq_mask[i] & in_le32(&pmac_irq_hw[i]->level))
__pmac_retrigger(irq_nr);
}
/* When an irq gets requested for the first client, if it's an
* edge interrupt, we clear any previous one on the controller
*/
static unsigned int pmac_startup_irq(struct irq_data *d)
{
unsigned long flags;
unsigned int src = irqd_to_hwirq(d);
2006-07-03 11:36:01 +00:00
unsigned long bit = 1UL << (src & 0x1f);
int i = src >> 5;
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
if (!irqd_is_level_type(d))
out_le32(&pmac_irq_hw[i]->ack, bit);
2006-07-03 11:36:01 +00:00
__set_bit(src, ppc_cached_irq_mask);
__pmac_set_irq_mask(src, 0);
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
return 0;
}
static void pmac_mask_irq(struct irq_data *d)
{
unsigned long flags;
unsigned int src = irqd_to_hwirq(d);
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
2006-07-03 11:36:01 +00:00
__clear_bit(src, ppc_cached_irq_mask);
__pmac_set_irq_mask(src, 1);
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
}
static void pmac_unmask_irq(struct irq_data *d)
{
unsigned long flags;
unsigned int src = irqd_to_hwirq(d);
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
2006-07-03 11:36:01 +00:00
__set_bit(src, ppc_cached_irq_mask);
__pmac_set_irq_mask(src, 0);
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
}
static int pmac_retrigger(struct irq_data *d)
{
unsigned long flags;
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
__pmac_retrigger(irqd_to_hwirq(d));
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
return 1;
}
static struct irq_chip pmac_pic = {
.name = "PMAC-PIC",
.irq_startup = pmac_startup_irq,
.irq_mask = pmac_mask_irq,
.irq_ack = pmac_ack_irq,
.irq_mask_ack = pmac_mask_and_ack_irq,
.irq_unmask = pmac_unmask_irq,
.irq_retrigger = pmac_retrigger,
};
static irqreturn_t gatwick_action(int cpl, void *dev_id)
{
unsigned long flags;
int irq, bits;
int rc = IRQ_NONE;
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
for (irq = max_irqs; (irq -= 32) >= max_real_irqs; ) {
int i = irq >> 5;
bits = in_le32(&pmac_irq_hw[i]->event) | ppc_lost_interrupts[i];
/* We must read level interrupts from the level register */
bits |= (in_le32(&pmac_irq_hw[i]->level) & level_mask[i]);
bits &= ppc_cached_irq_mask[i];
if (bits == 0)
continue;
irq += __ilog2(bits);
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
generic_handle_irq(irq);
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
rc = IRQ_HANDLED;
}
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
return rc;
}
static unsigned int pmac_pic_get_irq(void)
{
int irq;
unsigned long bits = 0;
unsigned long flags;
#ifdef CONFIG_PPC_PMAC32_PSURGE
void psurge_smp_message_recv(void);
/* IPI's are a hack on the powersurge -- Cort */
if ( smp_processor_id() != 0 ) {
psurge_smp_message_recv();
2006-07-03 11:36:01 +00:00
return NO_IRQ_IGNORE; /* ignore, already handled */
}
#endif /* CONFIG_PPC_PMAC32_PSURGE */
raw_spin_lock_irqsave(&pmac_pic_lock, flags);
for (irq = max_real_irqs; (irq -= 32) >= 0; ) {
int i = irq >> 5;
bits = in_le32(&pmac_irq_hw[i]->event) | ppc_lost_interrupts[i];
/* We must read level interrupts from the level register */
bits |= (in_le32(&pmac_irq_hw[i]->level) & level_mask[i]);
bits &= ppc_cached_irq_mask[i];
if (bits == 0)
continue;
irq += __ilog2(bits);
break;
}
raw_spin_unlock_irqrestore(&pmac_pic_lock, flags);
2006-07-03 11:36:01 +00:00
if (unlikely(irq < 0))
return NO_IRQ;
return irq_linear_revmap(pmac_pic_host, irq);
}
#ifdef CONFIG_XMON
static struct irqaction xmon_action = {
.handler = xmon_irq,
.flags = 0,
.name = "NMI - XMON"
};
#endif
static struct irqaction gatwick_cascade_action = {
.handler = gatwick_action,
.flags = IRQF_DISABLED,
.name = "cascade",
};
2006-07-03 11:36:01 +00:00
static int pmac_pic_host_match(struct irq_host *h, struct device_node *node)
{
/* We match all, we don't always have a node anyway */
return 1;
}
static int pmac_pic_host_map(struct irq_host *h, unsigned int virq,
[PATCH] powerpc: fix trigger handling in the new irq code This patch slightly reworks the new irq code to fix a small design error. I removed the passing of the trigger to the map() calls entirely, it was not a good idea to have one call do two different things. It also fixes a couple of corner cases. Mapping a linux virtual irq to a physical irq now does only that. Setting the trigger is a different action which has a different call. The main changes are: - I no longer call host->ops->map() for an already mapped irq, I just return the virtual number that was already mapped. It was called before to give an opportunity to change the trigger, but that was causing issues as that could happen while the interrupt was in use by a device, and because of the trigger change, map would potentially muck around with things in a racy way. That was causing much burden on a given's controller implementation of map() to get it right. This is much simpler now. map() is only called on the initial mapping of an irq, meaning that you know that this irq is _not_ being used. You can initialize the hardware if you want (though you don't have to). - Controllers that can handle different type of triggers (level/edge/etc...) now implement the standard irq_chip->set_type() call as defined by the generic code. That means that you can use the standard set_irq_type() to configure an irq line manually if you wish or (though I don't like that interface), pass explicit trigger flags to request_irq() as defined by the generic kernel interfaces. Also, using those interfaces guarantees that your controller set_type callback is called with the descriptor lock held, thus providing locking against activity on the same interrupt (including mask/unmask/etc...) automatically. A result is that, for example, MPIC's own map() implementation calls irq_set_type(NONE) to configure the hardware to the default triggers. - To allow the above, the irq_map array entry for the new mapped interrupt is now set before map() callback is called for the controller. - The irq_create_of_mapping() (also used by irq_of_parse_and_map()) function for mapping interrupts from the device-tree now also call the separate set_irq_type(), and only does so if there is a change in the trigger type. - While I was at it, I changed pci_read_irq_line() (which is the helper I would expect most archs to use in their pcibios_fixup() to get the PCI interrupt routing from the device tree) to also handle a fallback when the DT mapping fails consisting of reading the PCI_INTERRUPT_PIN to know wether the device has an interrupt at all, and the the PCI_INTERRUPT_LINE to get an interrupt number from the device. That number is then mapped using the default controller, and the trigger is set to level low. That default behaviour works for several platforms that don't have a proper interrupt tree like Pegasos. If it doesn't work for your platform, then either provide a proper interrupt tree from the firmware so that fallback isn't needed, or don't call pci_read_irq_line() - Add back a bit that got dropped by my main rework patch for properly clearing pending IPIs on pSeries when using a kexec Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-07-10 11:44:42 +00:00
irq_hw_number_t hw)
2006-07-03 11:36:01 +00:00
{
int level;
if (hw >= max_irqs)
return -EINVAL;
/* Mark level interrupts, set delayed disable for edge ones and set
* handlers
*/
level = !!(level_mask[hw >> 5] & (1UL << (hw & 0x1f)));
if (level)
irq_set_status_flags(virq, IRQ_LEVEL);
irq_set_chip_and_handler(virq, &pmac_pic,
level ? handle_level_irq : handle_edge_irq);
2006-07-03 11:36:01 +00:00
return 0;
}
static int pmac_pic_host_xlate(struct irq_host *h, struct device_node *ct,
const u32 *intspec, unsigned int intsize,
2006-07-03 11:36:01 +00:00
irq_hw_number_t *out_hwirq,
unsigned int *out_flags)
{
[PATCH] powerpc: fix trigger handling in the new irq code This patch slightly reworks the new irq code to fix a small design error. I removed the passing of the trigger to the map() calls entirely, it was not a good idea to have one call do two different things. It also fixes a couple of corner cases. Mapping a linux virtual irq to a physical irq now does only that. Setting the trigger is a different action which has a different call. The main changes are: - I no longer call host->ops->map() for an already mapped irq, I just return the virtual number that was already mapped. It was called before to give an opportunity to change the trigger, but that was causing issues as that could happen while the interrupt was in use by a device, and because of the trigger change, map would potentially muck around with things in a racy way. That was causing much burden on a given's controller implementation of map() to get it right. This is much simpler now. map() is only called on the initial mapping of an irq, meaning that you know that this irq is _not_ being used. You can initialize the hardware if you want (though you don't have to). - Controllers that can handle different type of triggers (level/edge/etc...) now implement the standard irq_chip->set_type() call as defined by the generic code. That means that you can use the standard set_irq_type() to configure an irq line manually if you wish or (though I don't like that interface), pass explicit trigger flags to request_irq() as defined by the generic kernel interfaces. Also, using those interfaces guarantees that your controller set_type callback is called with the descriptor lock held, thus providing locking against activity on the same interrupt (including mask/unmask/etc...) automatically. A result is that, for example, MPIC's own map() implementation calls irq_set_type(NONE) to configure the hardware to the default triggers. - To allow the above, the irq_map array entry for the new mapped interrupt is now set before map() callback is called for the controller. - The irq_create_of_mapping() (also used by irq_of_parse_and_map()) function for mapping interrupts from the device-tree now also call the separate set_irq_type(), and only does so if there is a change in the trigger type. - While I was at it, I changed pci_read_irq_line() (which is the helper I would expect most archs to use in their pcibios_fixup() to get the PCI interrupt routing from the device tree) to also handle a fallback when the DT mapping fails consisting of reading the PCI_INTERRUPT_PIN to know wether the device has an interrupt at all, and the the PCI_INTERRUPT_LINE to get an interrupt number from the device. That number is then mapped using the default controller, and the trigger is set to level low. That default behaviour works for several platforms that don't have a proper interrupt tree like Pegasos. If it doesn't work for your platform, then either provide a proper interrupt tree from the firmware so that fallback isn't needed, or don't call pci_read_irq_line() - Add back a bit that got dropped by my main rework patch for properly clearing pending IPIs on pSeries when using a kexec Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-07-10 11:44:42 +00:00
*out_flags = IRQ_TYPE_NONE;
2006-07-03 11:36:01 +00:00
*out_hwirq = *intspec;
return 0;
}
static struct irq_host_ops pmac_pic_host_ops = {
.match = pmac_pic_host_match,
.map = pmac_pic_host_map,
.xlate = pmac_pic_host_xlate,
};
static void __init pmac_pic_probe_oldstyle(void)
{
int i;
struct device_node *master = NULL;
struct device_node *slave = NULL;
u8 __iomem *addr;
struct resource r;
/* Set our get_irq function */
2006-07-03 11:36:01 +00:00
ppc_md.get_irq = pmac_pic_get_irq;
/*
* Find the interrupt controller type & node
*/
if ((master = of_find_node_by_name(NULL, "gc")) != NULL) {
max_irqs = max_real_irqs = 32;
level_mask[0] = GC_LEVEL_MASK;
} else if ((master = of_find_node_by_name(NULL, "ohare")) != NULL) {
max_irqs = max_real_irqs = 32;
level_mask[0] = OHARE_LEVEL_MASK;
/* We might have a second cascaded ohare */
slave = of_find_node_by_name(NULL, "pci106b,7");
if (slave) {
max_irqs = 64;
level_mask[1] = OHARE_LEVEL_MASK;
}
} else if ((master = of_find_node_by_name(NULL, "mac-io")) != NULL) {
max_irqs = max_real_irqs = 64;
level_mask[0] = HEATHROW_LEVEL_MASK;
level_mask[1] = 0;
/* We might have a second cascaded heathrow */
slave = of_find_node_by_name(master, "mac-io");
/* Check ordering of master & slave */
if (of_device_is_compatible(master, "gatwick")) {
struct device_node *tmp;
BUG_ON(slave == NULL);
tmp = master;
master = slave;
slave = tmp;
}
/* We found a slave */
if (slave) {
max_irqs = 128;
level_mask[2] = HEATHROW_LEVEL_MASK;
level_mask[3] = 0;
}
}
BUG_ON(master == NULL);
2006-07-03 11:36:01 +00:00
/*
* Allocate an irq host
*/
pmac_pic_host = irq_alloc_host(master, IRQ_HOST_MAP_LINEAR, max_irqs,
2006-07-03 11:36:01 +00:00
&pmac_pic_host_ops,
max_irqs);
BUG_ON(pmac_pic_host == NULL);
irq_set_default_host(pmac_pic_host);
/* Get addresses of first controller if we have a node for it */
BUG_ON(of_address_to_resource(master, 0, &r));
/* Map interrupts of primary controller */
addr = (u8 __iomem *) ioremap(r.start, 0x40);
i = 0;
pmac_irq_hw[i++] = (volatile struct pmac_irq_hw __iomem *)
(addr + 0x20);
if (max_real_irqs > 32)
pmac_irq_hw[i++] = (volatile struct pmac_irq_hw __iomem *)
(addr + 0x10);
of_node_put(master);
printk(KERN_INFO "irq: Found primary Apple PIC %s for %d irqs\n",
master->full_name, max_real_irqs);
/* Map interrupts of cascaded controller */
if (slave && !of_address_to_resource(slave, 0, &r)) {
addr = (u8 __iomem *)ioremap(r.start, 0x40);
pmac_irq_hw[i++] = (volatile struct pmac_irq_hw __iomem *)
(addr + 0x20);
if (max_irqs > 64)
pmac_irq_hw[i++] =
(volatile struct pmac_irq_hw __iomem *)
(addr + 0x10);
2006-07-03 11:36:01 +00:00
pmac_irq_cascade = irq_of_parse_and_map(slave, 0);
printk(KERN_INFO "irq: Found slave Apple PIC %s for %d irqs"
" cascade: %d\n", slave->full_name,
max_irqs - max_real_irqs, pmac_irq_cascade);
}
of_node_put(slave);
/* Disable all interrupts in all controllers */
for (i = 0; i * 32 < max_irqs; ++i)
out_le32(&pmac_irq_hw[i]->enable, 0);
/* Hookup cascade irq */
2006-07-03 11:36:01 +00:00
if (slave && pmac_irq_cascade != NO_IRQ)
setup_irq(pmac_irq_cascade, &gatwick_cascade_action);
printk(KERN_INFO "irq: System has %d possible interrupts\n", max_irqs);
#ifdef CONFIG_XMON
[PATCH] powerpc: fix trigger handling in the new irq code This patch slightly reworks the new irq code to fix a small design error. I removed the passing of the trigger to the map() calls entirely, it was not a good idea to have one call do two different things. It also fixes a couple of corner cases. Mapping a linux virtual irq to a physical irq now does only that. Setting the trigger is a different action which has a different call. The main changes are: - I no longer call host->ops->map() for an already mapped irq, I just return the virtual number that was already mapped. It was called before to give an opportunity to change the trigger, but that was causing issues as that could happen while the interrupt was in use by a device, and because of the trigger change, map would potentially muck around with things in a racy way. That was causing much burden on a given's controller implementation of map() to get it right. This is much simpler now. map() is only called on the initial mapping of an irq, meaning that you know that this irq is _not_ being used. You can initialize the hardware if you want (though you don't have to). - Controllers that can handle different type of triggers (level/edge/etc...) now implement the standard irq_chip->set_type() call as defined by the generic code. That means that you can use the standard set_irq_type() to configure an irq line manually if you wish or (though I don't like that interface), pass explicit trigger flags to request_irq() as defined by the generic kernel interfaces. Also, using those interfaces guarantees that your controller set_type callback is called with the descriptor lock held, thus providing locking against activity on the same interrupt (including mask/unmask/etc...) automatically. A result is that, for example, MPIC's own map() implementation calls irq_set_type(NONE) to configure the hardware to the default triggers. - To allow the above, the irq_map array entry for the new mapped interrupt is now set before map() callback is called for the controller. - The irq_create_of_mapping() (also used by irq_of_parse_and_map()) function for mapping interrupts from the device-tree now also call the separate set_irq_type(), and only does so if there is a change in the trigger type. - While I was at it, I changed pci_read_irq_line() (which is the helper I would expect most archs to use in their pcibios_fixup() to get the PCI interrupt routing from the device tree) to also handle a fallback when the DT mapping fails consisting of reading the PCI_INTERRUPT_PIN to know wether the device has an interrupt at all, and the the PCI_INTERRUPT_LINE to get an interrupt number from the device. That number is then mapped using the default controller, and the trigger is set to level low. That default behaviour works for several platforms that don't have a proper interrupt tree like Pegasos. If it doesn't work for your platform, then either provide a proper interrupt tree from the firmware so that fallback isn't needed, or don't call pci_read_irq_line() - Add back a bit that got dropped by my main rework patch for properly clearing pending IPIs on pSeries when using a kexec Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-07-10 11:44:42 +00:00
setup_irq(irq_create_mapping(NULL, 20), &xmon_action);
#endif
}
int of_irq_map_oldworld(struct device_node *device, int index,
struct of_irq *out_irq)
{
const u32 *ints = NULL;
int intlen;
/*
* Old machines just have a list of interrupt numbers
* and no interrupt-controller nodes. We also have dodgy
* cases where the APPL,interrupts property is completely
* missing behind pci-pci bridges and we have to get it
* from the parent (the bridge itself, as apple just wired
* everything together on these)
*/
while (device) {
ints = of_get_property(device, "AAPL,interrupts", &intlen);
if (ints != NULL)
break;
device = device->parent;
if (device && strcmp(device->type, "pci") != 0)
break;
}
if (ints == NULL)
return -EINVAL;
intlen /= sizeof(u32);
if (index >= intlen)
return -EINVAL;
out_irq->controller = NULL;
out_irq->specifier[0] = ints[index];
out_irq->size = 1;
return 0;
}
#endif /* CONFIG_PPC32 */
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static void pmac_u3_cascade(unsigned int irq, struct irq_desc *desc)
{
struct irq_chip *chip = irq_desc_get_chip(desc);
struct mpic *mpic = irq_desc_get_handler_data(desc);
unsigned int cascade_irq = mpic_get_one_irq(mpic);
2006-07-03 11:36:01 +00:00
if (cascade_irq != NO_IRQ)
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
generic_handle_irq(cascade_irq);
chip->irq_eoi(&desc->irq_data);
}
static void __init pmac_pic_setup_mpic_nmi(struct mpic *mpic)
{
#if defined(CONFIG_XMON) && defined(CONFIG_PPC32)
struct device_node* pswitch;
int nmi_irq;
pswitch = of_find_node_by_name(NULL, "programmer-switch");
2006-07-03 11:36:01 +00:00
if (pswitch) {
nmi_irq = irq_of_parse_and_map(pswitch, 0);
if (nmi_irq != NO_IRQ) {
mpic_irq_set_priority(nmi_irq, 9);
setup_irq(nmi_irq, &xmon_action);
}
of_node_put(pswitch);
}
#endif /* defined(CONFIG_XMON) && defined(CONFIG_PPC32) */
}
static struct mpic * __init pmac_setup_one_mpic(struct device_node *np,
int master)
{
const char *name = master ? " MPIC 1 " : " MPIC 2 ";
struct resource r;
struct mpic *mpic;
unsigned int flags = master ? MPIC_PRIMARY : 0;
int rc;
rc = of_address_to_resource(np, 0, &r);
if (rc)
return NULL;
pmac_call_feature(PMAC_FTR_ENABLE_MPIC, np, 0, 0);
flags |= MPIC_WANTS_RESET;
if (of_get_property(np, "big-endian", NULL))
flags |= MPIC_BIG_ENDIAN;
/* Primary Big Endian means HT interrupts. This is quite dodgy
* but works until I find a better way
*/
if (master && (flags & MPIC_BIG_ENDIAN))
flags |= MPIC_U3_HT_IRQS;
2006-07-03 11:36:01 +00:00
mpic = mpic_alloc(np, r.start, flags, 0, 0, name);
if (mpic == NULL)
return NULL;
mpic_init(mpic);
return mpic;
}
static int __init pmac_pic_probe_mpic(void)
{
struct mpic *mpic1, *mpic2;
struct device_node *np, *master = NULL, *slave = NULL;
2006-07-03 11:36:01 +00:00
unsigned int cascade;
/* We can have up to 2 MPICs cascaded */
for (np = NULL; (np = of_find_node_by_type(np, "open-pic"))
!= NULL;) {
if (master == NULL &&
of_get_property(np, "interrupts", NULL) == NULL)
master = of_node_get(np);
else if (slave == NULL)
slave = of_node_get(np);
if (master && slave)
break;
}
/* Check for bogus setups */
if (master == NULL && slave != NULL) {
master = slave;
slave = NULL;
}
/* Not found, default to good old pmac pic */
if (master == NULL)
return -ENODEV;
/* Set master handler */
ppc_md.get_irq = mpic_get_irq;
/* Setup master */
mpic1 = pmac_setup_one_mpic(master, 1);
BUG_ON(mpic1 == NULL);
/* Install NMI if any */
pmac_pic_setup_mpic_nmi(mpic1);
of_node_put(master);
/* No slave, let's go out */
2006-07-03 11:36:01 +00:00
if (slave == NULL)
return 0;
/* Get/Map slave interrupt */
cascade = irq_of_parse_and_map(slave, 0);
if (cascade == NO_IRQ) {
printk(KERN_ERR "Failed to map cascade IRQ\n");
return 0;
2006-07-03 11:36:01 +00:00
}
mpic2 = pmac_setup_one_mpic(slave, 0);
if (mpic2 == NULL) {
printk(KERN_ERR "Failed to setup slave MPIC\n");
of_node_put(slave);
return 0;
}
irq_set_handler_data(cascade, mpic2);
irq_set_chained_handler(cascade, pmac_u3_cascade);
of_node_put(slave);
return 0;
}
void __init pmac_pic_init(void)
{
2006-07-03 11:36:01 +00:00
/* We configure the OF parsing based on our oldworld vs. newworld
* platform type and wether we were booted by BootX.
*/
#ifdef CONFIG_PPC32
if (!pmac_newworld)
of_irq_workarounds |= OF_IMAP_OLDWORLD_MAC;
if (of_get_property(of_chosen, "linux,bootx", NULL) != NULL)
of_irq_workarounds |= OF_IMAP_NO_PHANDLE;
2006-07-03 11:36:01 +00:00
/* If we don't have phandles on a newworld, then try to locate a
* default interrupt controller (happens when booting with BootX).
* We do a first match here, hopefully, that only ever happens on
* machines with one controller.
*/
if (pmac_newworld && (of_irq_workarounds & OF_IMAP_NO_PHANDLE)) {
struct device_node *np;
for_each_node_with_property(np, "interrupt-controller") {
/* Skip /chosen/interrupt-controller */
if (strcmp(np->name, "chosen") == 0)
continue;
/* It seems like at least one person wants
* to use BootX on a machine with an AppleKiwi
* controller which happens to pretend to be an
* interrupt controller too. */
if (strcmp(np->name, "AppleKiwi") == 0)
continue;
/* I think we found one ! */
of_irq_dflt_pic = np;
break;
}
}
#endif /* CONFIG_PPC32 */
[PATCH] powerpc: fix trigger handling in the new irq code This patch slightly reworks the new irq code to fix a small design error. I removed the passing of the trigger to the map() calls entirely, it was not a good idea to have one call do two different things. It also fixes a couple of corner cases. Mapping a linux virtual irq to a physical irq now does only that. Setting the trigger is a different action which has a different call. The main changes are: - I no longer call host->ops->map() for an already mapped irq, I just return the virtual number that was already mapped. It was called before to give an opportunity to change the trigger, but that was causing issues as that could happen while the interrupt was in use by a device, and because of the trigger change, map would potentially muck around with things in a racy way. That was causing much burden on a given's controller implementation of map() to get it right. This is much simpler now. map() is only called on the initial mapping of an irq, meaning that you know that this irq is _not_ being used. You can initialize the hardware if you want (though you don't have to). - Controllers that can handle different type of triggers (level/edge/etc...) now implement the standard irq_chip->set_type() call as defined by the generic code. That means that you can use the standard set_irq_type() to configure an irq line manually if you wish or (though I don't like that interface), pass explicit trigger flags to request_irq() as defined by the generic kernel interfaces. Also, using those interfaces guarantees that your controller set_type callback is called with the descriptor lock held, thus providing locking against activity on the same interrupt (including mask/unmask/etc...) automatically. A result is that, for example, MPIC's own map() implementation calls irq_set_type(NONE) to configure the hardware to the default triggers. - To allow the above, the irq_map array entry for the new mapped interrupt is now set before map() callback is called for the controller. - The irq_create_of_mapping() (also used by irq_of_parse_and_map()) function for mapping interrupts from the device-tree now also call the separate set_irq_type(), and only does so if there is a change in the trigger type. - While I was at it, I changed pci_read_irq_line() (which is the helper I would expect most archs to use in their pcibios_fixup() to get the PCI interrupt routing from the device tree) to also handle a fallback when the DT mapping fails consisting of reading the PCI_INTERRUPT_PIN to know wether the device has an interrupt at all, and the the PCI_INTERRUPT_LINE to get an interrupt number from the device. That number is then mapped using the default controller, and the trigger is set to level low. That default behaviour works for several platforms that don't have a proper interrupt tree like Pegasos. If it doesn't work for your platform, then either provide a proper interrupt tree from the firmware so that fallback isn't needed, or don't call pci_read_irq_line() - Add back a bit that got dropped by my main rework patch for properly clearing pending IPIs on pSeries when using a kexec Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-07-10 11:44:42 +00:00
/* We first try to detect Apple's new Core99 chipset, since mac-io
* is quite different on those machines and contains an IBM MPIC2.
*/
if (pmac_pic_probe_mpic() == 0)
return;
#ifdef CONFIG_PPC32
pmac_pic_probe_oldstyle();
#endif
}
#if defined(CONFIG_PM) && defined(CONFIG_PPC32)
/*
* These procedures are used in implementing sleep on the powerbooks.
* sleep_save_intrs() saves the states of all interrupt enables
* and disables all interrupts except for the nominated one.
* sleep_restore_intrs() restores the states of all interrupt enables.
*/
unsigned long sleep_save_mask[2];
/* This used to be passed by the PMU driver but that link got
* broken with the new driver model. We use this tweak for now...
2006-07-03 11:36:01 +00:00
* We really want to do things differently though...
*/
static int pmacpic_find_viaint(void)
{
int viaint = -1;
#ifdef CONFIG_ADB_PMU
struct device_node *np;
if (pmu_get_model() != PMU_OHARE_BASED)
goto not_found;
np = of_find_node_by_name(NULL, "via-pmu");
if (np == NULL)
goto not_found;
viaint = irq_of_parse_and_map(np, 0);
not_found:
#endif /* CONFIG_ADB_PMU */
return viaint;
}
static int pmacpic_suspend(struct sys_device *sysdev, pm_message_t state)
{
int viaint = pmacpic_find_viaint();
sleep_save_mask[0] = ppc_cached_irq_mask[0];
sleep_save_mask[1] = ppc_cached_irq_mask[1];
ppc_cached_irq_mask[0] = 0;
ppc_cached_irq_mask[1] = 0;
if (viaint > 0)
set_bit(viaint, ppc_cached_irq_mask);
out_le32(&pmac_irq_hw[0]->enable, ppc_cached_irq_mask[0]);
if (max_real_irqs > 32)
out_le32(&pmac_irq_hw[1]->enable, ppc_cached_irq_mask[1]);
(void)in_le32(&pmac_irq_hw[0]->event);
/* make sure mask gets to controller before we return to caller */
mb();
(void)in_le32(&pmac_irq_hw[0]->enable);
return 0;
}
static int pmacpic_resume(struct sys_device *sysdev)
{
int i;
out_le32(&pmac_irq_hw[0]->enable, 0);
if (max_real_irqs > 32)
out_le32(&pmac_irq_hw[1]->enable, 0);
mb();
for (i = 0; i < max_real_irqs; ++i)
if (test_bit(i, sleep_save_mask))
pmac_unmask_irq(irq_get_irq_data(i));
return 0;
}
#endif /* CONFIG_PM && CONFIG_PPC32 */
static struct sysdev_class pmacpic_sysclass = {
.name = "pmac_pic",
};
static struct sys_device device_pmacpic = {
.id = 0,
.cls = &pmacpic_sysclass,
};
static struct sysdev_driver driver_pmacpic = {
#if defined(CONFIG_PM) && defined(CONFIG_PPC32)
.suspend = &pmacpic_suspend,
.resume = &pmacpic_resume,
#endif /* CONFIG_PM && CONFIG_PPC32 */
};
static int __init init_pmacpic_sysfs(void)
{
#ifdef CONFIG_PPC32
if (max_irqs == 0)
return -ENODEV;
#endif
printk(KERN_DEBUG "Registering pmac pic with sysfs...\n");
sysdev_class_register(&pmacpic_sysclass);
sysdev_register(&device_pmacpic);
sysdev_driver_register(&pmacpic_sysclass, &driver_pmacpic);
return 0;
}
machine_subsys_initcall(powermac, init_pmacpic_sysfs);