linux/arch/powerpc/platforms/pseries/xics.c

944 lines
22 KiB
C
Raw Normal View History

/*
* arch/powerpc/platforms/pseries/xics.c
*
* Copyright 2000 IBM Corporation.
*
* 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.
*/
2006-07-03 11:36:01 +00:00
#include <linux/types.h>
#include <linux/threads.h>
#include <linux/kernel.h>
#include <linux/irq.h>
#include <linux/smp.h>
#include <linux/interrupt.h>
#include <linux/init.h>
#include <linux/radix-tree.h>
#include <linux/cpu.h>
#include <linux/msi.h>
#include <linux/of.h>
#include <linux/percpu.h>
2006-07-03 11:36:01 +00:00
#include <asm/firmware.h>
#include <asm/io.h>
#include <asm/pgtable.h>
#include <asm/smp.h>
#include <asm/rtas.h>
#include <asm/hvcall.h>
#include <asm/machdep.h>
#include "xics.h"
#include "plpar_wrappers.h"
static struct irq_host *xics_host;
#define XICS_IPI 2
#define XICS_IRQ_SPURIOUS 0
/* Want a priority other than 0. Various HW issues require this. */
#define DEFAULT_PRIORITY 5
/*
* Mark IPIs as higher priority so we can take them inside interrupts that
* arent marked IRQF_DISABLED
*/
#define IPI_PRIORITY 4
/* The least favored priority */
#define LOWEST_PRIORITY 0xFF
/* The number of priorities defined above */
#define MAX_NUM_PRIORITIES 3
static unsigned int default_server = 0xFF;
static unsigned int default_distrib_server = 0;
static unsigned int interrupt_server_size = 8;
/* RTAS service tokens */
static int ibm_get_xive;
static int ibm_set_xive;
static int ibm_int_on;
static int ibm_int_off;
struct xics_cppr {
unsigned char stack[MAX_NUM_PRIORITIES];
int index;
};
static DEFINE_PER_CPU(struct xics_cppr, xics_cppr);
/* Direct hardware low level accessors */
/* The part of the interrupt presentation layer that we care about */
struct xics_ipl {
union {
u32 word;
u8 bytes[4];
} xirr_poll;
union {
u32 word;
u8 bytes[4];
} xirr;
u32 dummy;
union {
u32 word;
u8 bytes[4];
} qirr;
};
static struct xics_ipl __iomem *xics_per_cpu[NR_CPUS];
static inline unsigned int direct_xirr_info_get(void)
{
int cpu = smp_processor_id();
return in_be32(&xics_per_cpu[cpu]->xirr.word);
}
static inline void direct_xirr_info_set(unsigned int value)
{
int cpu = smp_processor_id();
out_be32(&xics_per_cpu[cpu]->xirr.word, value);
}
static inline void direct_cppr_info(u8 value)
{
int cpu = smp_processor_id();
out_8(&xics_per_cpu[cpu]->xirr.bytes[0], value);
}
static inline void direct_qirr_info(int n_cpu, u8 value)
{
out_8(&xics_per_cpu[n_cpu]->qirr.bytes[0], value);
}
/* LPAR low level accessors */
static inline unsigned int lpar_xirr_info_get(unsigned char cppr)
{
unsigned long lpar_rc;
unsigned long return_value;
lpar_rc = plpar_xirr(&return_value, cppr);
if (lpar_rc != H_SUCCESS)
panic(" bad return code xirr - rc = %lx\n", lpar_rc);
2006-07-03 11:36:01 +00:00
return (unsigned int)return_value;
}
static inline void lpar_xirr_info_set(unsigned int value)
{
unsigned long lpar_rc;
lpar_rc = plpar_eoi(value);
if (lpar_rc != H_SUCCESS)
panic("bad return code EOI - rc = %ld, value=%x\n", lpar_rc,
value);
}
static inline void lpar_cppr_info(u8 value)
{
unsigned long lpar_rc;
lpar_rc = plpar_cppr(value);
if (lpar_rc != H_SUCCESS)
panic("bad return code cppr - rc = %lx\n", lpar_rc);
}
static inline void lpar_qirr_info(int n_cpu , u8 value)
{
unsigned long lpar_rc;
lpar_rc = plpar_ipi(get_hard_smp_processor_id(n_cpu), value);
if (lpar_rc != H_SUCCESS)
panic("bad return code qirr - rc = %lx\n", lpar_rc);
}
/* Interface to generic irq subsystem */
#ifdef CONFIG_SMP
/*
* For the moment we only implement delivery to all cpus or one cpu.
*
* If the requested affinity is cpu_all_mask, we set global affinity.
* If not we set it to the first cpu in the mask, even if multiple cpus
* are set. This is so things like irqbalance (which set core and package
* wide affinities) do the right thing.
*/
static int get_irq_server(unsigned int virq, const struct cpumask *cpumask,
unsigned int strict_check)
{
if (!distribute_irqs)
return default_server;
if (!cpumask_equal(cpumask, cpu_all_mask)) {
int server = cpumask_first_and(cpu_online_mask, cpumask);
if (server < nr_cpu_ids)
return get_hard_smp_processor_id(server);
if (strict_check)
return -1;
}
/*
* Workaround issue with some versions of JS20 firmware that
* deliver interrupts to cpus which haven't been started. This
* happens when using the maxcpus= boot option.
*/
if (cpumask_equal(cpu_online_mask, cpu_present_mask))
return default_distrib_server;
return default_server;
}
#else
#define get_irq_server(virq, cpumask, strict_check) (default_server)
#endif
static void xics_unmask_irq(unsigned int virq)
{
unsigned int irq;
int call_status;
int server;
pr_devel("xics: unmask virq %d\n", virq);
2006-07-03 11:36:01 +00:00
irq = (unsigned int)irq_map[virq].hwirq;
pr_devel(" -> map to hwirq 0x%x\n", irq);
2006-07-03 11:36:01 +00:00
if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
return;
server = get_irq_server(virq, irq_to_desc(virq)->affinity, 0);
call_status = rtas_call(ibm_set_xive, 3, 1, NULL, irq, server,
DEFAULT_PRIORITY);
if (call_status != 0) {
printk(KERN_ERR
"%s: ibm_set_xive irq %u server %x returned %d\n",
__func__, irq, server, call_status);
return;
}
/* Now unmask the interrupt (often a no-op) */
call_status = rtas_call(ibm_int_on, 1, 1, NULL, irq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_on irq=%u returned %d\n",
__func__, irq, call_status);
return;
}
}
static unsigned int xics_startup(unsigned int virq)
{
/*
* The generic MSI code returns with the interrupt disabled on the
* card, using the MSI mask bits. Firmware doesn't appear to unmask
* at that level, so we do it here by hand.
*/
if (irq_to_desc(virq)->msi_desc)
unmask_msi_irq(virq);
/* unmask it */
xics_unmask_irq(virq);
return 0;
}
static void xics_mask_real_irq(unsigned int irq)
{
int call_status;
if (irq == XICS_IPI)
return;
call_status = rtas_call(ibm_int_off, 1, 1, NULL, irq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_off irq=%u returned %d\n",
__func__, irq, call_status);
return;
}
/* Have to set XIVE to 0xff to be able to remove a slot */
call_status = rtas_call(ibm_set_xive, 3, 1, NULL, irq,
default_server, 0xff);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_set_xive(0xff) irq=%u returned %d\n",
__func__, irq, call_status);
return;
}
}
static void xics_mask_irq(unsigned int virq)
{
unsigned int irq;
pr_devel("xics: mask virq %d\n", virq);
2006-07-03 11:36:01 +00:00
irq = (unsigned int)irq_map[virq].hwirq;
if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
return;
xics_mask_real_irq(irq);
}
static void xics_mask_unknown_vec(unsigned int vec)
{
printk(KERN_ERR "Interrupt %u (real) is invalid, disabling it.\n", vec);
xics_mask_real_irq(vec);
}
static inline unsigned int xics_xirr_vector(unsigned int xirr)
{
/*
* The top byte is the old cppr, to be restored on EOI.
* The remaining 24 bits are the vector.
*/
return xirr & 0x00ffffff;
}
static void push_cppr(unsigned int vec)
{
struct xics_cppr *os_cppr = &__get_cpu_var(xics_cppr);
if (WARN_ON(os_cppr->index >= MAX_NUM_PRIORITIES - 1))
return;
if (vec == XICS_IPI)
os_cppr->stack[++os_cppr->index] = IPI_PRIORITY;
else
os_cppr->stack[++os_cppr->index] = DEFAULT_PRIORITY;
}
static unsigned int xics_get_irq_direct(void)
{
unsigned int xirr = direct_xirr_info_get();
unsigned int vec = xics_xirr_vector(xirr);
unsigned int irq;
if (vec == XICS_IRQ_SPURIOUS)
return NO_IRQ;
irq = irq_radix_revmap_lookup(xics_host, vec);
if (likely(irq != NO_IRQ)) {
push_cppr(vec);
2006-07-03 11:36:01 +00:00
return irq;
}
/* We don't have a linux mapping, so have rtas mask it. */
xics_mask_unknown_vec(vec);
/* We might learn about it later, so EOI it */
direct_xirr_info_set(xirr);
return NO_IRQ;
}
static unsigned int xics_get_irq_lpar(void)
{
struct xics_cppr *os_cppr = &__get_cpu_var(xics_cppr);
unsigned int xirr = lpar_xirr_info_get(os_cppr->stack[os_cppr->index]);
unsigned int vec = xics_xirr_vector(xirr);
unsigned int irq;
if (vec == XICS_IRQ_SPURIOUS)
return NO_IRQ;
irq = irq_radix_revmap_lookup(xics_host, vec);
if (likely(irq != NO_IRQ)) {
push_cppr(vec);
return irq;
}
/* We don't have a linux mapping, so have RTAS mask it. */
xics_mask_unknown_vec(vec);
/* We might learn about it later, so EOI it */
lpar_xirr_info_set(xirr);
return NO_IRQ;
}
static unsigned char pop_cppr(void)
{
struct xics_cppr *os_cppr = &__get_cpu_var(xics_cppr);
if (WARN_ON(os_cppr->index < 1))
return LOWEST_PRIORITY;
return os_cppr->stack[--os_cppr->index];
}
static void xics_eoi_direct(unsigned int virq)
{
unsigned int irq = (unsigned int)irq_map[virq].hwirq;
iosync();
direct_xirr_info_set((pop_cppr() << 24) | irq);
}
static void xics_eoi_lpar(unsigned int virq)
{
unsigned int irq = (unsigned int)irq_map[virq].hwirq;
iosync();
lpar_xirr_info_set((pop_cppr() << 24) | irq);
}
static int xics_set_affinity(unsigned int virq, const struct cpumask *cpumask)
{
unsigned int irq;
int status;
int xics_status[2];
int irq_server;
2006-07-03 11:36:01 +00:00
irq = (unsigned int)irq_map[virq].hwirq;
if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
return -1;
status = rtas_call(ibm_get_xive, 1, 3, xics_status, irq);
if (status) {
printk(KERN_ERR "%s: ibm,get-xive irq=%u returns %d\n",
__func__, irq, status);
return -1;
}
irq_server = get_irq_server(virq, cpumask, 1);
if (irq_server == -1) {
char cpulist[128];
cpumask_scnprintf(cpulist, sizeof(cpulist), cpumask);
printk(KERN_WARNING
"%s: No online cpus in the mask %s for irq %d\n",
__func__, cpulist, virq);
return -1;
}
status = rtas_call(ibm_set_xive, 3, 1, NULL,
irq, irq_server, xics_status[1]);
if (status) {
printk(KERN_ERR "%s: ibm,set-xive irq=%u returns %d\n",
__func__, irq, status);
return -1;
}
return 0;
}
static struct irq_chip xics_pic_direct = {
.name = "XICS",
.startup = xics_startup,
.mask = xics_mask_irq,
.unmask = xics_unmask_irq,
.eoi = xics_eoi_direct,
.set_affinity = xics_set_affinity
};
static struct irq_chip xics_pic_lpar = {
.name = "XICS",
.startup = xics_startup,
.mask = xics_mask_irq,
.unmask = xics_unmask_irq,
.eoi = xics_eoi_lpar,
.set_affinity = xics_set_affinity
};
/* Interface to arch irq controller subsystem layer */
/* Points to the irq_chip we're actually using */
static struct irq_chip *xics_irq_chip;
2006-07-03 11:36:01 +00:00
static int xics_host_match(struct irq_host *h, struct device_node *node)
{
2006-07-03 11:36:01 +00:00
/* IBM machines have interrupt parents of various funky types for things
* like vdevices, events, etc... The trick we use here is to match
* everything here except the legacy 8259 which is compatible "chrp,iic"
*/
return !of_device_is_compatible(node, "chrp,iic");
2006-07-03 11:36:01 +00:00
}
static int xics_host_map(struct irq_host *h, unsigned int virq,
irq_hw_number_t hw)
2006-07-03 11:36:01 +00:00
{
pr_devel("xics: map virq %d, hwirq 0x%lx\n", virq, hw);
2006-07-03 11:36:01 +00:00
/* Insert the interrupt mapping into the radix tree for fast lookup */
irq_radix_revmap_insert(xics_host, virq, hw);
irq_to_desc(virq)->status |= IRQ_LEVEL;
set_irq_chip_and_handler(virq, xics_irq_chip, handle_fasteoi_irq);
2006-07-03 11:36:01 +00:00
return 0;
}
static int xics_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)
{
/* Current xics implementation translates everything
* to level. It is not technically right for MSIs but this
* is irrelevant at this point. We might get smarter in the future
*/
2006-07-03 11:36:01 +00:00
*out_hwirq = intspec[0];
*out_flags = IRQ_TYPE_LEVEL_LOW;
return 0;
}
static struct irq_host_ops xics_host_ops = {
2006-07-03 11:36:01 +00:00
.match = xics_host_match,
.map = xics_host_map,
2006-07-03 11:36:01 +00:00
.xlate = xics_host_xlate,
};
static void __init xics_init_host(void)
{
if (firmware_has_feature(FW_FEATURE_LPAR))
xics_irq_chip = &xics_pic_lpar;
2006-07-03 11:36:01 +00:00
else
xics_irq_chip = &xics_pic_direct;
xics_host = irq_alloc_host(NULL, IRQ_HOST_MAP_TREE, 0, &xics_host_ops,
2006-07-03 11:36:01 +00:00
XICS_IRQ_SPURIOUS);
BUG_ON(xics_host == NULL);
irq_set_default_host(xics_host);
}
/* Inter-processor interrupt support */
#ifdef CONFIG_SMP
/*
* XICS only has a single IPI, so encode the messages per CPU
*/
static DEFINE_PER_CPU_SHARED_ALIGNED(unsigned long, xics_ipi_message);
static inline void smp_xics_do_message(int cpu, int msg)
{
unsigned long *tgt = &per_cpu(xics_ipi_message, cpu);
set_bit(msg, tgt);
mb();
if (firmware_has_feature(FW_FEATURE_LPAR))
lpar_qirr_info(cpu, IPI_PRIORITY);
else
direct_qirr_info(cpu, IPI_PRIORITY);
}
void smp_xics_message_pass(int target, int msg)
{
unsigned int i;
if (target < NR_CPUS) {
smp_xics_do_message(target, msg);
} else {
for_each_online_cpu(i) {
if (target == MSG_ALL_BUT_SELF
&& i == smp_processor_id())
continue;
smp_xics_do_message(i, msg);
}
}
}
static irqreturn_t xics_ipi_dispatch(int cpu)
{
unsigned long *tgt = &per_cpu(xics_ipi_message, cpu);
mb(); /* order mmio clearing qirr */
while (*tgt) {
if (test_and_clear_bit(PPC_MSG_CALL_FUNCTION, tgt)) {
smp_message_recv(PPC_MSG_CALL_FUNCTION);
}
if (test_and_clear_bit(PPC_MSG_RESCHEDULE, tgt)) {
smp_message_recv(PPC_MSG_RESCHEDULE);
}
if (test_and_clear_bit(PPC_MSG_CALL_FUNC_SINGLE, tgt)) {
smp_message_recv(PPC_MSG_CALL_FUNC_SINGLE);
}
#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC)
if (test_and_clear_bit(PPC_MSG_DEBUGGER_BREAK, tgt)) {
smp_message_recv(PPC_MSG_DEBUGGER_BREAK);
}
#endif
}
return IRQ_HANDLED;
}
static irqreturn_t xics_ipi_action_direct(int irq, void *dev_id)
{
int cpu = smp_processor_id();
direct_qirr_info(cpu, 0xff);
return xics_ipi_dispatch(cpu);
}
static irqreturn_t xics_ipi_action_lpar(int irq, void *dev_id)
{
int cpu = smp_processor_id();
lpar_qirr_info(cpu, 0xff);
return xics_ipi_dispatch(cpu);
}
static void xics_request_ipi(void)
{
unsigned int ipi;
int rc;
ipi = irq_create_mapping(xics_host, XICS_IPI);
BUG_ON(ipi == NO_IRQ);
/*
* IPIs are marked IRQF_DISABLED as they must run with irqs
* disabled
*/
set_irq_handler(ipi, handle_percpu_irq);
if (firmware_has_feature(FW_FEATURE_LPAR))
rc = request_irq(ipi, xics_ipi_action_lpar,
IRQF_DISABLED|IRQF_PERCPU, "IPI", NULL);
else
rc = request_irq(ipi, xics_ipi_action_direct,
IRQF_DISABLED|IRQF_PERCPU, "IPI", NULL);
BUG_ON(rc);
}
int __init smp_xics_probe(void)
{
xics_request_ipi();
return cpumask_weight(cpu_possible_mask);
}
#endif /* CONFIG_SMP */
/* Initialization */
static void xics_update_irq_servers(void)
{
int i, j;
struct device_node *np;
u32 ilen;
const u32 *ireg;
u32 hcpuid;
/* Find the server numbers for the boot cpu. */
np = of_get_cpu_node(boot_cpuid, NULL);
BUG_ON(!np);
ireg = of_get_property(np, "ibm,ppc-interrupt-gserver#s", &ilen);
if (!ireg) {
of_node_put(np);
return;
}
i = ilen / sizeof(int);
hcpuid = get_hard_smp_processor_id(boot_cpuid);
/* Global interrupt distribution server is specified in the last
* entry of "ibm,ppc-interrupt-gserver#s" property. Get the last
* entry fom this property for current boot cpu id and use it as
* default distribution server
*/
for (j = 0; j < i; j += 2) {
if (ireg[j] == hcpuid) {
default_server = hcpuid;
default_distrib_server = ireg[j+1];
}
}
of_node_put(np);
}
2006-07-03 11:36:01 +00:00
static void __init xics_map_one_cpu(int hw_id, unsigned long addr,
unsigned long size)
{
int i;
2006-07-03 11:36:01 +00:00
/* This may look gross but it's good enough for now, we don't quite
* have a hard -> linux processor id matching.
*/
for_each_possible_cpu(i) {
if (!cpu_present(i))
continue;
if (hw_id == get_hard_smp_processor_id(i)) {
xics_per_cpu[i] = ioremap(addr, size);
return;
}
}
}
2006-07-03 11:36:01 +00:00
static void __init xics_init_one_node(struct device_node *np,
unsigned int *indx)
{
unsigned int ilen;
const u32 *ireg;
2006-07-03 11:36:01 +00:00
/* This code does the theorically broken assumption that the interrupt
* server numbers are the same as the hard CPU numbers.
* This happens to be the case so far but we are playing with fire...
* should be fixed one of these days. -BenH.
*/
ireg = of_get_property(np, "ibm,interrupt-server-ranges", NULL);
2006-07-03 11:36:01 +00:00
/* Do that ever happen ? we'll know soon enough... but even good'old
* f80 does have that property ..
*/
WARN_ON(ireg == NULL);
if (ireg) {
/*
* set node starting index for this node
*/
2006-07-03 11:36:01 +00:00
*indx = *ireg;
}
ireg = of_get_property(np, "reg", &ilen);
if (!ireg)
panic("xics_init_IRQ: can't find interrupt reg property");
2006-07-03 11:36:01 +00:00
while (ilen >= (4 * sizeof(u32))) {
unsigned long addr, size;
/* XXX Use proper OF parsing code here !!! */
addr = (unsigned long)*ireg++ << 32;
ilen -= sizeof(u32);
addr |= *ireg++;
ilen -= sizeof(u32);
size = (unsigned long)*ireg++ << 32;
ilen -= sizeof(u32);
size |= *ireg++;
ilen -= sizeof(u32);
xics_map_one_cpu(*indx, addr, size);
(*indx)++;
}
}
void __init xics_init_IRQ(void)
{
struct device_node *np;
u32 indx = 0;
2006-07-03 11:36:01 +00:00
int found = 0;
const u32 *isize;
2006-07-03 11:36:01 +00:00
ppc64_boot_msg(0x20, "XICS Init");
ibm_get_xive = rtas_token("ibm,get-xive");
ibm_set_xive = rtas_token("ibm,set-xive");
ibm_int_on = rtas_token("ibm,int-on");
ibm_int_off = rtas_token("ibm,int-off");
for_each_node_by_type(np, "PowerPC-External-Interrupt-Presentation") {
found = 1;
if (firmware_has_feature(FW_FEATURE_LPAR)) {
of_node_put(np);
2006-07-03 11:36:01 +00:00
break;
}
2006-07-03 11:36:01 +00:00
xics_init_one_node(np, &indx);
}
if (found == 0)
return;
/* get the bit size of server numbers */
found = 0;
for_each_compatible_node(np, NULL, "ibm,ppc-xics") {
isize = of_get_property(np, "ibm,interrupt-server#-size", NULL);
if (!isize)
continue;
if (!found) {
interrupt_server_size = *isize;
found = 1;
} else if (*isize != interrupt_server_size) {
printk(KERN_WARNING "XICS: "
"mismatched ibm,interrupt-server#-size\n");
interrupt_server_size = max(*isize,
interrupt_server_size);
}
}
xics_update_irq_servers();
xics_init_host();
2006-07-03 11:36:01 +00:00
if (firmware_has_feature(FW_FEATURE_LPAR))
ppc_md.get_irq = xics_get_irq_lpar;
else
ppc_md.get_irq = xics_get_irq_direct;
xics_setup_cpu();
2006-07-03 11:36:01 +00:00
ppc64_boot_msg(0x21, "XICS Done");
}
/* Cpu startup, shutdown, and hotplug */
static void xics_set_cpu_priority(unsigned char cppr)
{
struct xics_cppr *os_cppr = &__get_cpu_var(xics_cppr);
/*
* we only really want to set the priority when there's
* just one cppr value on the stack
*/
WARN_ON(os_cppr->index != 0);
os_cppr->stack[0] = cppr;
if (firmware_has_feature(FW_FEATURE_LPAR))
lpar_cppr_info(cppr);
else
direct_cppr_info(cppr);
iosync();
}
/* Have the calling processor join or leave the specified global queue */
static void xics_set_cpu_giq(unsigned int gserver, unsigned int join)
{
int index;
int status;
if (!rtas_indicator_present(GLOBAL_INTERRUPT_QUEUE, NULL))
return;
index = (1UL << interrupt_server_size) - 1 - gserver;
status = rtas_set_indicator_fast(GLOBAL_INTERRUPT_QUEUE, index, join);
WARN(status < 0, "set-indicator(%d, %d, %u) returned %d\n",
GLOBAL_INTERRUPT_QUEUE, index, join, status);
}
void xics_setup_cpu(void)
{
xics_set_cpu_priority(LOWEST_PRIORITY);
xics_set_cpu_giq(default_distrib_server, 1);
}
void xics_teardown_cpu(void)
[PATCH] ppc64: kexec support for ppc64 This patch implements the kexec support for ppc64 platforms. A couple of notes: 1) We copy the pages in virtual mode, using the full base kernel and a statically allocated stack. At kexec_prepare time we scan the pages and if any overlap our (0, _end[]) range we return -ETXTBSY. On PowerPC 64 systems running in LPAR (logical partitioning) mode, only a small region of memory, referred to as the RMO, can be accessed in real mode. Since Linux runs with only one zone of memory in the memory allocator, and it can be orders of magnitude more memory than the RMO, looping until we allocate pages in the source region is not feasible. Copying in virtual means we don't have to write a hash table generation and call hypervisor to insert translations, instead we rely on the pinned kernel linear mapping. The kernel already has move to linked location built in, so there is no requirement to load it at 0. If we want to load something other than a kernel, then a stub can be written to copy a linear chunk in real mode. 2) The start entry point gets passed parameters from the kernel. Slaves are started at a fixed address after copying code from the entry point. All CPUs get passed their firmware assigned physical id in r3 (most calling conventions use this register for the first argument). This is used to distinguish each CPU from all other CPUs. Since firmware is not around, there is no other way to obtain this information other than to pass it somewhere. A single CPU, referred to here as the master and the one executing the kexec call, branches to start with the address of start in r4. While this can be calculated, we have to load it through a gpr to branch to this point so defining the register this is contained in is free. A stack of unspecified size is available at r1 (also common calling convention). All remaining running CPUs are sent to start at absolute address 0x60 after copying the first 0x100 bytes from start to address 0. This convention was chosen because it matches what the kernel has been doing itself. (only gpr3 is defined). Note: This is not quite the convention of the kexec bootblock v2 in the kernel. A stub has been written to convert between them, and we may adjust the kernel in the future to allow this directly without any stub. 3) Destination pages can be placed anywhere, even where they would not be accessible in real mode. This will allow us to place ram disks above the RMO if we choose. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: R Sharada <sharada@in.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-25 21:58:10 +00:00
{
struct xics_cppr *os_cppr = &__get_cpu_var(xics_cppr);
[PATCH] ppc64: kexec support for ppc64 This patch implements the kexec support for ppc64 platforms. A couple of notes: 1) We copy the pages in virtual mode, using the full base kernel and a statically allocated stack. At kexec_prepare time we scan the pages and if any overlap our (0, _end[]) range we return -ETXTBSY. On PowerPC 64 systems running in LPAR (logical partitioning) mode, only a small region of memory, referred to as the RMO, can be accessed in real mode. Since Linux runs with only one zone of memory in the memory allocator, and it can be orders of magnitude more memory than the RMO, looping until we allocate pages in the source region is not feasible. Copying in virtual means we don't have to write a hash table generation and call hypervisor to insert translations, instead we rely on the pinned kernel linear mapping. The kernel already has move to linked location built in, so there is no requirement to load it at 0. If we want to load something other than a kernel, then a stub can be written to copy a linear chunk in real mode. 2) The start entry point gets passed parameters from the kernel. Slaves are started at a fixed address after copying code from the entry point. All CPUs get passed their firmware assigned physical id in r3 (most calling conventions use this register for the first argument). This is used to distinguish each CPU from all other CPUs. Since firmware is not around, there is no other way to obtain this information other than to pass it somewhere. A single CPU, referred to here as the master and the one executing the kexec call, branches to start with the address of start in r4. While this can be calculated, we have to load it through a gpr to branch to this point so defining the register this is contained in is free. A stack of unspecified size is available at r1 (also common calling convention). All remaining running CPUs are sent to start at absolute address 0x60 after copying the first 0x100 bytes from start to address 0. This convention was chosen because it matches what the kernel has been doing itself. (only gpr3 is defined). Note: This is not quite the convention of the kexec bootblock v2 in the kernel. A stub has been written to convert between them, and we may adjust the kernel in the future to allow this directly without any stub. 3) Destination pages can be placed anywhere, even where they would not be accessible in real mode. This will allow us to place ram disks above the RMO if we choose. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: R Sharada <sharada@in.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-25 21:58:10 +00:00
int cpu = smp_processor_id();
/*
* we have to reset the cppr index to 0 because we're
* not going to return from the IPI
*/
os_cppr->index = 0;
xics_set_cpu_priority(0);
/* Clear any pending IPI request */
[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
if (firmware_has_feature(FW_FEATURE_LPAR))
lpar_qirr_info(cpu, 0xff);
else
direct_qirr_info(cpu, 0xff);
}
void xics_kexec_teardown_cpu(int secondary)
{
xics_teardown_cpu();
[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 take the ipi irq but and never return so we
* need to EOI the IPI, but want to leave our priority 0
*
* should we check all the other interrupts too?
* should we be flagging idle loop instead?
* or creating some task to be scheduled?
*/
2006-07-03 11:36:01 +00:00
if (firmware_has_feature(FW_FEATURE_LPAR))
lpar_xirr_info_set((0x00 << 24) | XICS_IPI);
else
direct_xirr_info_set((0x00 << 24) | XICS_IPI);
[PATCH] ppc64: kexec support for ppc64 This patch implements the kexec support for ppc64 platforms. A couple of notes: 1) We copy the pages in virtual mode, using the full base kernel and a statically allocated stack. At kexec_prepare time we scan the pages and if any overlap our (0, _end[]) range we return -ETXTBSY. On PowerPC 64 systems running in LPAR (logical partitioning) mode, only a small region of memory, referred to as the RMO, can be accessed in real mode. Since Linux runs with only one zone of memory in the memory allocator, and it can be orders of magnitude more memory than the RMO, looping until we allocate pages in the source region is not feasible. Copying in virtual means we don't have to write a hash table generation and call hypervisor to insert translations, instead we rely on the pinned kernel linear mapping. The kernel already has move to linked location built in, so there is no requirement to load it at 0. If we want to load something other than a kernel, then a stub can be written to copy a linear chunk in real mode. 2) The start entry point gets passed parameters from the kernel. Slaves are started at a fixed address after copying code from the entry point. All CPUs get passed their firmware assigned physical id in r3 (most calling conventions use this register for the first argument). This is used to distinguish each CPU from all other CPUs. Since firmware is not around, there is no other way to obtain this information other than to pass it somewhere. A single CPU, referred to here as the master and the one executing the kexec call, branches to start with the address of start in r4. While this can be calculated, we have to load it through a gpr to branch to this point so defining the register this is contained in is free. A stack of unspecified size is available at r1 (also common calling convention). All remaining running CPUs are sent to start at absolute address 0x60 after copying the first 0x100 bytes from start to address 0. This convention was chosen because it matches what the kernel has been doing itself. (only gpr3 is defined). Note: This is not quite the convention of the kexec bootblock v2 in the kernel. A stub has been written to convert between them, and we may adjust the kernel in the future to allow this directly without any stub. 3) Destination pages can be placed anywhere, even where they would not be accessible in real mode. This will allow us to place ram disks above the RMO if we choose. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: R Sharada <sharada@in.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-25 21:58:10 +00:00
/*
* Some machines need to have at least one cpu in the GIQ,
* so leave the master cpu in the group.
[PATCH] ppc64: kexec support for ppc64 This patch implements the kexec support for ppc64 platforms. A couple of notes: 1) We copy the pages in virtual mode, using the full base kernel and a statically allocated stack. At kexec_prepare time we scan the pages and if any overlap our (0, _end[]) range we return -ETXTBSY. On PowerPC 64 systems running in LPAR (logical partitioning) mode, only a small region of memory, referred to as the RMO, can be accessed in real mode. Since Linux runs with only one zone of memory in the memory allocator, and it can be orders of magnitude more memory than the RMO, looping until we allocate pages in the source region is not feasible. Copying in virtual means we don't have to write a hash table generation and call hypervisor to insert translations, instead we rely on the pinned kernel linear mapping. The kernel already has move to linked location built in, so there is no requirement to load it at 0. If we want to load something other than a kernel, then a stub can be written to copy a linear chunk in real mode. 2) The start entry point gets passed parameters from the kernel. Slaves are started at a fixed address after copying code from the entry point. All CPUs get passed their firmware assigned physical id in r3 (most calling conventions use this register for the first argument). This is used to distinguish each CPU from all other CPUs. Since firmware is not around, there is no other way to obtain this information other than to pass it somewhere. A single CPU, referred to here as the master and the one executing the kexec call, branches to start with the address of start in r4. While this can be calculated, we have to load it through a gpr to branch to this point so defining the register this is contained in is free. A stack of unspecified size is available at r1 (also common calling convention). All remaining running CPUs are sent to start at absolute address 0x60 after copying the first 0x100 bytes from start to address 0. This convention was chosen because it matches what the kernel has been doing itself. (only gpr3 is defined). Note: This is not quite the convention of the kexec bootblock v2 in the kernel. A stub has been written to convert between them, and we may adjust the kernel in the future to allow this directly without any stub. 3) Destination pages can be placed anywhere, even where they would not be accessible in real mode. This will allow us to place ram disks above the RMO if we choose. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: R Sharada <sharada@in.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-25 21:58:10 +00:00
*/
if (secondary)
xics_set_cpu_giq(default_distrib_server, 0);
[PATCH] ppc64: kexec support for ppc64 This patch implements the kexec support for ppc64 platforms. A couple of notes: 1) We copy the pages in virtual mode, using the full base kernel and a statically allocated stack. At kexec_prepare time we scan the pages and if any overlap our (0, _end[]) range we return -ETXTBSY. On PowerPC 64 systems running in LPAR (logical partitioning) mode, only a small region of memory, referred to as the RMO, can be accessed in real mode. Since Linux runs with only one zone of memory in the memory allocator, and it can be orders of magnitude more memory than the RMO, looping until we allocate pages in the source region is not feasible. Copying in virtual means we don't have to write a hash table generation and call hypervisor to insert translations, instead we rely on the pinned kernel linear mapping. The kernel already has move to linked location built in, so there is no requirement to load it at 0. If we want to load something other than a kernel, then a stub can be written to copy a linear chunk in real mode. 2) The start entry point gets passed parameters from the kernel. Slaves are started at a fixed address after copying code from the entry point. All CPUs get passed their firmware assigned physical id in r3 (most calling conventions use this register for the first argument). This is used to distinguish each CPU from all other CPUs. Since firmware is not around, there is no other way to obtain this information other than to pass it somewhere. A single CPU, referred to here as the master and the one executing the kexec call, branches to start with the address of start in r4. While this can be calculated, we have to load it through a gpr to branch to this point so defining the register this is contained in is free. A stack of unspecified size is available at r1 (also common calling convention). All remaining running CPUs are sent to start at absolute address 0x60 after copying the first 0x100 bytes from start to address 0. This convention was chosen because it matches what the kernel has been doing itself. (only gpr3 is defined). Note: This is not quite the convention of the kexec bootblock v2 in the kernel. A stub has been written to convert between them, and we may adjust the kernel in the future to allow this directly without any stub. 3) Destination pages can be placed anywhere, even where they would not be accessible in real mode. This will allow us to place ram disks above the RMO if we choose. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: R Sharada <sharada@in.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-25 21:58:10 +00:00
}
#ifdef CONFIG_HOTPLUG_CPU
/* Interrupts are disabled. */
void xics_migrate_irqs_away(void)
{
int cpu = smp_processor_id(), hw_cpu = hard_smp_processor_id();
unsigned int irq, virq;
/* If we used to be the default server, move to the new "boot_cpuid" */
if (hw_cpu == default_server)
xics_update_irq_servers();
/* Reject any interrupt that was queued to us... */
xics_set_cpu_priority(0);
/* Remove ourselves from the global interrupt queue */
xics_set_cpu_giq(default_distrib_server, 0);
/* Allow IPIs again... */
xics_set_cpu_priority(DEFAULT_PRIORITY);
for_each_irq(virq) {
struct irq_desc *desc;
int xics_status[2];
int status;
unsigned long flags;
/* We cant set affinity on ISA interrupts */
2006-07-03 11:36:01 +00:00
if (virq < NUM_ISA_INTERRUPTS)
continue;
2006-07-03 11:36:01 +00:00
if (irq_map[virq].host != xics_host)
continue;
irq = (unsigned int)irq_map[virq].hwirq;
/* We need to get IPIs still. */
2006-07-03 11:36:01 +00:00
if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
continue;
desc = irq_to_desc(virq);
/* We only need to migrate enabled IRQS */
[PATCH] genirq: rename desc->handler to desc->chip This patch-queue improves the generic IRQ layer to be truly generic, by adding various abstractions and features to it, without impacting existing functionality. While the queue can be best described as "fix and improve everything in the generic IRQ layer that we could think of", and thus it consists of many smaller features and lots of cleanups, the one feature that stands out most is the new 'irq chip' abstraction. The irq-chip abstraction is about describing and coding and IRQ controller driver by mapping its raw hardware capabilities [and quirks, if needed] in a straightforward way, without having to think about "IRQ flow" (level/edge/etc.) type of details. This stands in contrast with the current 'irq-type' model of genirq architectures, which 'mixes' raw hardware capabilities with 'flow' details. The patchset supports both types of irq controller designs at once, and converts i386 and x86_64 to the new irq-chip design. As a bonus side-effect of the irq-chip approach, chained interrupt controllers (master/slave PIC constructs, etc.) are now supported by design as well. The end result of this patchset intends to be simpler architecture-level code and more consolidation between architectures. We reused many bits of code and many concepts from Russell King's ARM IRQ layer, the merging of which was one of the motivations for this patchset. This patch: rename desc->handler to desc->chip. Originally i did not want to do this, because it's a big patch. But having both "desc->handler", "desc->handle_irq" and "action->handler" caused a large degree of confusion and made the code appear alot less clean than it truly is. I have also attempted a dual approach as well by introducing a desc->chip alias - but that just wasnt robust enough and broke frequently. So lets get over with this quickly. The conversion was done automatically via scripts and converts all the code in the kernel. This renaming patch is the first one amongst the patches, so that the remaining patches can stay flexible and can be merged and split up without having some big monolithic patch act as a merge barrier. [akpm@osdl.org: build fix] [akpm@osdl.org: another build fix] Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-29 09:24:36 +00:00
if (desc == NULL || desc->chip == NULL
|| desc->action == NULL
[PATCH] genirq: rename desc->handler to desc->chip This patch-queue improves the generic IRQ layer to be truly generic, by adding various abstractions and features to it, without impacting existing functionality. While the queue can be best described as "fix and improve everything in the generic IRQ layer that we could think of", and thus it consists of many smaller features and lots of cleanups, the one feature that stands out most is the new 'irq chip' abstraction. The irq-chip abstraction is about describing and coding and IRQ controller driver by mapping its raw hardware capabilities [and quirks, if needed] in a straightforward way, without having to think about "IRQ flow" (level/edge/etc.) type of details. This stands in contrast with the current 'irq-type' model of genirq architectures, which 'mixes' raw hardware capabilities with 'flow' details. The patchset supports both types of irq controller designs at once, and converts i386 and x86_64 to the new irq-chip design. As a bonus side-effect of the irq-chip approach, chained interrupt controllers (master/slave PIC constructs, etc.) are now supported by design as well. The end result of this patchset intends to be simpler architecture-level code and more consolidation between architectures. We reused many bits of code and many concepts from Russell King's ARM IRQ layer, the merging of which was one of the motivations for this patchset. This patch: rename desc->handler to desc->chip. Originally i did not want to do this, because it's a big patch. But having both "desc->handler", "desc->handle_irq" and "action->handler" caused a large degree of confusion and made the code appear alot less clean than it truly is. I have also attempted a dual approach as well by introducing a desc->chip alias - but that just wasnt robust enough and broke frequently. So lets get over with this quickly. The conversion was done automatically via scripts and converts all the code in the kernel. This renaming patch is the first one amongst the patches, so that the remaining patches can stay flexible and can be merged and split up without having some big monolithic patch act as a merge barrier. [akpm@osdl.org: build fix] [akpm@osdl.org: another build fix] Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-29 09:24:36 +00:00
|| desc->chip->set_affinity == NULL)
continue;
raw_spin_lock_irqsave(&desc->lock, flags);
status = rtas_call(ibm_get_xive, 1, 3, xics_status, irq);
if (status) {
printk(KERN_ERR "%s: ibm,get-xive irq=%u returns %d\n",
__func__, irq, status);
goto unlock;
}
/*
* We only support delivery to all cpus or to one cpu.
* The irq has to be migrated only in the single cpu
* case.
*/
if (xics_status[0] != hw_cpu)
goto unlock;
/* This is expected during cpu offline. */
if (cpu_online(cpu))
printk(KERN_WARNING "IRQ %u affinity broken off cpu %u\n",
virq, cpu);
/* Reset affinity to all cpus */
cpumask_setall(irq_to_desc(virq)->affinity);
desc->chip->set_affinity(virq, cpu_all_mask);
unlock:
raw_spin_unlock_irqrestore(&desc->lock, flags);
}
}
#endif