* 05/01/30 Suresh Siddha <suresh.b.siddha@intel.com>
* Setup cpu_sibling_map and cpu_core_map
*/
-#include <linux/config.h>
#include <linux/module.h>
#include <linux/acpi.h>
#include <linux/mm.h>
#include <linux/notifier.h>
#include <linux/smp.h>
-#include <linux/smp_lock.h>
#include <linux/spinlock.h>
#include <linux/efi.h>
#include <linux/percpu.h>
#include <asm/cache.h>
#include <asm/current.h>
#include <asm/delay.h>
-#include <asm/ia32.h>
#include <asm/io.h>
#include <asm/irq.h>
#include <asm/machvec.h>
#include <asm/mca.h>
#include <asm/page.h>
+#include <asm/paravirt.h>
#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <asm/processor.h>
#include <asm/system.h>
#include <asm/tlbflush.h>
#include <asm/unistd.h>
+#include <asm/sn/arch.h>
#define SMP_DEBUG 0
#endif
#ifdef CONFIG_HOTPLUG_CPU
+#ifdef CONFIG_PERMIT_BSP_REMOVE
+#define bsp_remove_ok 1
+#else
+#define bsp_remove_ok 0
+#endif
+
/*
* Store all idle threads, this can be reused instead of creating
* a new thread. Also avoids complicated thread destroy functionality
/*
* ITC synchronization related stuff:
*/
-#define MASTER 0
+#define MASTER (0)
#define SLAVE (SMP_CACHE_BYTES/8)
#define NUM_ROUNDS 64 /* magic value */
#define DEBUG_ITC_SYNC 0
-extern void __devinit calibrate_delay (void);
extern void start_ap (void);
extern unsigned long ia64_iobase;
-task_t *task_for_booting_cpu;
+struct task_struct *task_for_booting_cpu;
/*
* State for each CPU
*/
DEFINE_PER_CPU(int, cpu_state);
-/* Bitmasks of currently online, and possible CPUs */
-cpumask_t cpu_online_map;
-EXPORT_SYMBOL(cpu_online_map);
-cpumask_t cpu_possible_map;
-EXPORT_SYMBOL(cpu_possible_map);
-
cpumask_t cpu_core_map[NR_CPUS] __cacheline_aligned;
-cpumask_t cpu_sibling_map[NR_CPUS] __cacheline_aligned;
+EXPORT_SYMBOL(cpu_core_map);
+DEFINE_PER_CPU_SHARED_ALIGNED(cpumask_t, cpu_sibling_map);
+EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
+
int smp_num_siblings = 1;
-int smp_num_cpucores = 1;
/* which logical CPU number maps to which CPU (physical APIC ID) */
volatile int ia64_cpu_to_sapicid[NR_CPUS];
unsigned char smp_int_redirect; /* are INT and IPI redirectable by the chipset? */
+#ifdef CONFIG_FORCE_CPEI_RETARGET
+#define CPEI_OVERRIDE_DEFAULT (1)
+#else
+#define CPEI_OVERRIDE_DEFAULT (0)
+#endif
+
+unsigned int force_cpei_retarget = CPEI_OVERRIDE_DEFAULT;
+
+static int __init
+cmdl_force_cpei(char *str)
+{
+ int value=0;
+
+ get_option (&str, &value);
+ force_cpei_retarget = value;
+
+ return 1;
+}
+
+__setup("force_cpei=", cmdl_force_cpei);
+
static int __init
nointroute (char *str)
{
__setup("nointroute", nointroute);
+static void fix_b0_for_bsp(void)
+{
+#ifdef CONFIG_HOTPLUG_CPU
+ int cpuid;
+ static int fix_bsp_b0 = 1;
+
+ cpuid = smp_processor_id();
+
+ /*
+ * Cache the b0 value on the first AP that comes up
+ */
+ if (!(fix_bsp_b0 && cpuid))
+ return;
+
+ sal_boot_rendez_state[0].br[0] = sal_boot_rendez_state[cpuid].br[0];
+ printk ("Fixed BSP b0 value from CPU %d\n", cpuid);
+
+ fix_bsp_b0 = 0;
+#endif
+}
+
void
sync_master (void *arg)
{
go[MASTER] = 1;
- if (smp_call_function_single(master, sync_master, NULL, 1, 0) < 0) {
+ if (smp_call_function_single(master, sync_master, NULL, 0) < 0) {
printk(KERN_ERR "sync_itc: failed to get attention of CPU %u!\n", master);
return;
}
{
}
-static void __devinit
+static void __cpuinit
smp_callin (void)
{
- int cpuid, phys_id;
+ int cpuid, phys_id, itc_master;
+ struct cpuinfo_ia64 *last_cpuinfo, *this_cpuinfo;
extern void ia64_init_itm(void);
+ extern volatile int time_keeper_id;
#ifdef CONFIG_PERFMON
extern void pfm_init_percpu(void);
cpuid = smp_processor_id();
phys_id = hard_smp_processor_id();
+ itc_master = time_keeper_id;
if (cpu_online(cpuid)) {
printk(KERN_ERR "huh, phys CPU#0x%x, CPU#0x%x already present??\n",
BUG();
}
- lock_ipi_calllock();
+ fix_b0_for_bsp();
+
+ ipi_call_lock_irq();
+ spin_lock(&vector_lock);
+ /* Setup the per cpu irq handling data structures */
+ __setup_vector_irq(cpuid);
+ notify_cpu_starting(cpuid);
cpu_set(cpuid, cpu_online_map);
- unlock_ipi_calllock();
+ per_cpu(cpu_state, cpuid) = CPU_ONLINE;
+ spin_unlock(&vector_lock);
+ ipi_call_unlock_irq();
smp_setup_percpu_timer();
* calls spin_unlock_bh(), which calls spin_unlock_bh(), which calls
* local_bh_enable(), which bugs out if irqs are not enabled...
*/
- Dprintk("Going to syncup ITC with BP.\n");
- ia64_sync_itc(0);
+ Dprintk("Going to syncup ITC with ITC Master.\n");
+ ia64_sync_itc(itc_master);
}
/*
* Get our bogomips.
*/
ia64_init_itm();
- calibrate_delay();
- local_cpu_data->loops_per_jiffy = loops_per_jiffy;
-#ifdef CONFIG_IA32_SUPPORT
- ia32_gdt_init();
-#endif
+ /*
+ * Delay calibration can be skipped if new processor is identical to the
+ * previous processor.
+ */
+ last_cpuinfo = cpu_data(cpuid - 1);
+ this_cpuinfo = local_cpu_data;
+ if (last_cpuinfo->itc_freq != this_cpuinfo->itc_freq ||
+ last_cpuinfo->proc_freq != this_cpuinfo->proc_freq ||
+ last_cpuinfo->features != this_cpuinfo->features ||
+ last_cpuinfo->revision != this_cpuinfo->revision ||
+ last_cpuinfo->family != this_cpuinfo->family ||
+ last_cpuinfo->archrev != this_cpuinfo->archrev ||
+ last_cpuinfo->model != this_cpuinfo->model)
+ calibrate_delay();
+ local_cpu_data->loops_per_jiffy = loops_per_jiffy;
/*
* Allow the master to continue.
/*
* Activate a secondary processor. head.S calls this.
*/
-int __devinit
+int __cpuinit
start_secondary (void *unused)
{
/* Early console may use I/O ports */
ia64_set_kr(IA64_KR_IO_BASE, __pa(ia64_iobase));
+#ifndef CONFIG_PRINTK_TIME
Dprintk("start_secondary: starting CPU 0x%x\n", hard_smp_processor_id());
+#endif
efi_map_pal_code();
cpu_init();
+ preempt_disable();
smp_callin();
cpu_idle();
return 0;
}
-struct pt_regs * __devinit idle_regs(struct pt_regs *regs)
+struct pt_regs * __cpuinit idle_regs(struct pt_regs *regs)
{
return NULL;
}
struct create_idle {
+ struct work_struct work;
struct task_struct *idle;
struct completion done;
int cpu;
};
-void
-do_fork_idle(void *_c_idle)
+void __cpuinit
+do_fork_idle(struct work_struct *work)
{
- struct create_idle *c_idle = _c_idle;
+ struct create_idle *c_idle =
+ container_of(work, struct create_idle, work);
c_idle->idle = fork_idle(c_idle->cpu);
complete(&c_idle->done);
}
-static int __devinit
+static int __cpuinit
do_boot_cpu (int sapicid, int cpu)
{
int timeout;
struct create_idle c_idle = {
+ .work = __WORK_INITIALIZER(c_idle.work, do_fork_idle),
.cpu = cpu,
.done = COMPLETION_INITIALIZER(c_idle.done),
};
- DECLARE_WORK(work, do_fork_idle, &c_idle);
c_idle.idle = get_idle_for_cpu(cpu);
if (c_idle.idle) {
* We can't use kernel_thread since we must avoid to reschedule the child.
*/
if (!keventd_up() || current_is_keventd())
- work.func(work.data);
+ c_idle.work.func(&c_idle.work);
else {
- schedule_work(&work);
+ schedule_work(&c_idle.work);
wait_for_completion(&c_idle.done);
}
for (cpu = 0; cpu < NR_CPUS; cpu++) {
ia64_cpu_to_sapicid[cpu] = -1;
-#ifdef CONFIG_HOTPLUG_CPU
- cpu_set(cpu, cpu_possible_map);
-#endif
}
ia64_cpu_to_sapicid[0] = boot_cpu_id;
cpus_clear(cpu_present_map);
- cpu_set(0, cpu_present_map);
- cpu_set(0, cpu_possible_map);
+ set_cpu_present(0, true);
+ set_cpu_possible(0, true);
for (cpu = 1, i = 0; i < smp_boot_data.cpu_count; i++) {
sapicid = smp_boot_data.cpu_phys_id[i];
if (sapicid == boot_cpu_id)
continue;
- cpu_set(cpu, cpu_present_map);
- cpu_set(cpu, cpu_possible_map);
+ set_cpu_present(cpu, true);
+ set_cpu_possible(cpu, true);
ia64_cpu_to_sapicid[cpu] = sapicid;
cpu++;
}
}
-#ifdef CONFIG_NUMA
-
-/* on which node is each logical CPU (one cacheline even for 64 CPUs) */
-u8 cpu_to_node_map[NR_CPUS] __cacheline_aligned;
-EXPORT_SYMBOL(cpu_to_node_map);
-/* which logical CPUs are on which nodes */
-cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
-
-/*
- * Build cpu to node mapping and initialize the per node cpu masks.
- */
-void __init
-build_cpu_to_node_map (void)
-{
- int cpu, i, node;
-
- for(node=0; node<MAX_NUMNODES; node++)
- cpus_clear(node_to_cpu_mask[node]);
- for(cpu = 0; cpu < NR_CPUS; ++cpu) {
- /*
- * All Itanium NUMA platforms I know use ACPI, so maybe we
- * can drop this ifdef completely. [EF]
- */
-#ifdef CONFIG_ACPI_NUMA
- node = -1;
- for (i = 0; i < NR_CPUS; ++i)
- if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
- node = node_cpuid[i].nid;
- break;
- }
-#else
-# error Fixme: Dunno how to build CPU-to-node map.
-#endif
- cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
- if (node >= 0)
- cpu_set(cpu, node_to_cpu_mask[node]);
- }
-}
-
-#endif /* CONFIG_NUMA */
-
/*
* Cycle through the APs sending Wakeup IPIs to boot each.
*/
*/
if (!max_cpus) {
printk(KERN_INFO "SMP mode deactivated.\n");
- cpus_clear(cpu_online_map);
- cpus_clear(cpu_present_map);
- cpus_clear(cpu_possible_map);
- cpu_set(0, cpu_online_map);
- cpu_set(0, cpu_present_map);
- cpu_set(0, cpu_possible_map);
+ init_cpu_online(cpumask_of(0));
+ init_cpu_present(cpumask_of(0));
+ init_cpu_possible(cpumask_of(0));
return;
}
}
{
cpu_set(smp_processor_id(), cpu_online_map);
cpu_set(smp_processor_id(), cpu_callin_map);
+ per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
+ paravirt_post_smp_prepare_boot_cpu();
}
-/*
- * mt_info[] is a temporary store for all info returned by
- * PAL_LOGICAL_TO_PHYSICAL, to be copied into cpuinfo_ia64 when the
- * specific cpu comes.
- */
-static struct {
- __u32 socket_id;
- __u16 core_id;
- __u16 thread_id;
- __u16 proc_fixed_addr;
- __u8 valid;
-} mt_info[NR_CPUS] __devinitdata;
-
#ifdef CONFIG_HOTPLUG_CPU
static inline void
-remove_from_mtinfo(int cpu)
-{
- int i;
-
- for_each_cpu(i)
- if (mt_info[i].valid && mt_info[i].socket_id ==
- cpu_data(cpu)->socket_id)
- mt_info[i].valid = 0;
-}
-
-static inline void
clear_cpu_sibling_map(int cpu)
{
int i;
- for_each_cpu_mask(i, cpu_sibling_map[cpu])
- cpu_clear(cpu, cpu_sibling_map[i]);
+ for_each_cpu_mask(i, per_cpu(cpu_sibling_map, cpu))
+ cpu_clear(cpu, per_cpu(cpu_sibling_map, i));
for_each_cpu_mask(i, cpu_core_map[cpu])
cpu_clear(cpu, cpu_core_map[i]);
- cpu_sibling_map[cpu] = cpu_core_map[cpu] = CPU_MASK_NONE;
+ per_cpu(cpu_sibling_map, cpu) = cpu_core_map[cpu] = CPU_MASK_NONE;
}
static void
if (cpu_data(cpu)->threads_per_core == 1 &&
cpu_data(cpu)->cores_per_socket == 1) {
cpu_clear(cpu, cpu_core_map[cpu]);
- cpu_clear(cpu, cpu_sibling_map[cpu]);
+ cpu_clear(cpu, per_cpu(cpu_sibling_map, cpu));
return;
}
/* remove it from all sibling map's */
clear_cpu_sibling_map(cpu);
+}
+
+extern void fixup_irqs(void);
- /* if this cpu is the last in the core group, remove all its info
- * from mt_info structure
+int migrate_platform_irqs(unsigned int cpu)
+{
+ int new_cpei_cpu;
+ struct irq_desc *desc = NULL;
+ const struct cpumask *mask;
+ int retval = 0;
+
+ /*
+ * dont permit CPEI target to removed.
*/
- if (last)
- remove_from_mtinfo(cpu);
+ if (cpe_vector > 0 && is_cpu_cpei_target(cpu)) {
+ printk ("CPU (%d) is CPEI Target\n", cpu);
+ if (can_cpei_retarget()) {
+ /*
+ * Now re-target the CPEI to a different processor
+ */
+ new_cpei_cpu = any_online_cpu(cpu_online_map);
+ mask = cpumask_of(new_cpei_cpu);
+ set_cpei_target_cpu(new_cpei_cpu);
+ desc = irq_desc + ia64_cpe_irq;
+ /*
+ * Switch for now, immediately, we need to do fake intr
+ * as other interrupts, but need to study CPEI behaviour with
+ * polling before making changes.
+ */
+ if (desc) {
+ desc->chip->disable(ia64_cpe_irq);
+ desc->chip->set_affinity(ia64_cpe_irq, mask);
+ desc->chip->enable(ia64_cpe_irq);
+ printk ("Re-targetting CPEI to cpu %d\n", new_cpei_cpu);
+ }
+ }
+ if (!desc) {
+ printk ("Unable to retarget CPEI, offline cpu [%d] failed\n", cpu);
+ retval = -EBUSY;
+ }
+ }
+ return retval;
}
-extern void fixup_irqs(void);
/* must be called with cpucontrol mutex held */
int __cpu_disable(void)
{
/*
* dont permit boot processor for now
*/
- if (cpu == 0)
+ if (cpu == 0 && !bsp_remove_ok) {
+ printk ("Your platform does not support removal of BSP\n");
+ return (-EBUSY);
+ }
+
+ if (ia64_platform_is("sn2")) {
+ if (!sn_cpu_disable_allowed(cpu))
+ return -EBUSY;
+ }
+
+ cpu_clear(cpu, cpu_online_map);
+
+ if (migrate_platform_irqs(cpu)) {
+ cpu_set(cpu, cpu_online_map);
return -EBUSY;
+ }
remove_siblinginfo(cpu);
fixup_irqs();
}
printk(KERN_ERR "CPU %u didn't die...\n", cpu);
}
-#else /* !CONFIG_HOTPLUG_CPU */
-int __cpu_disable(void)
-{
- return -ENOSYS;
-}
-
-void __cpu_die(unsigned int cpu)
-{
- /* We said "no" in __cpu_disable */
- BUG();
-}
#endif /* CONFIG_HOTPLUG_CPU */
void
* Allow the user to impress friends.
*/
- for (cpu = 0; cpu < NR_CPUS; cpu++)
- if (cpu_online(cpu))
- bogosum += cpu_data(cpu)->loops_per_jiffy;
+ for_each_online_cpu(cpu) {
+ bogosum += cpu_data(cpu)->loops_per_jiffy;
+ }
printk(KERN_INFO "Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
(int)num_online_cpus(), bogosum/(500000/HZ), (bogosum/(5000/HZ))%100);
cpu_set(i, cpu_core_map[cpu]);
cpu_set(cpu, cpu_core_map[i]);
if (cpu_data(cpu)->core_id == cpu_data(i)->core_id) {
- cpu_set(i, cpu_sibling_map[cpu]);
- cpu_set(cpu, cpu_sibling_map[i]);
+ cpu_set(i, per_cpu(cpu_sibling_map, cpu));
+ cpu_set(cpu, per_cpu(cpu_sibling_map, i));
}
}
}
}
-int __devinit
+int __cpuinit
__cpu_up (unsigned int cpu)
{
int ret;
if (cpu_isset(cpu, cpu_callin_map))
return -EINVAL;
+ per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
/* Processor goes to start_secondary(), sets online flag */
ret = do_boot_cpu(sapicid, cpu);
if (ret < 0)
if (cpu_data(cpu)->threads_per_core == 1 &&
cpu_data(cpu)->cores_per_socket == 1) {
- cpu_set(cpu, cpu_sibling_map[cpu]);
+ cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
cpu_set(cpu, cpu_core_map[cpu]);
return 0;
}
}
/*
- * Assume that CPU's have been discovered by some platform-dependent interface. For
+ * Assume that CPUs have been discovered by some platform-dependent interface. For
* SoftSDV/Lion, that would be ACPI.
*
* Setup of the IPI irq handler is done in irq.c:init_IRQ_SMP().
} *ap_startup;
long sal_ret;
- /* Tell SAL where to drop the AP's. */
+ /* Tell SAL where to drop the APs. */
ap_startup = (struct fptr *) start_ap;
sal_ret = ia64_sal_set_vectors(SAL_VECTOR_OS_BOOT_RENDEZ,
ia64_tpa(ap_startup->fp), ia64_tpa(ap_startup->gp), 0, 0, 0, 0);
ia64_sal_strerror(sal_ret));
}
-static inline int __devinit
-check_for_mtinfo_index(void)
-{
- int i;
-
- for_each_cpu(i)
- if (!mt_info[i].valid)
- return i;
-
- return -1;
-}
-
-/*
- * Search the mt_info to find out if this socket's cid/tid information is
- * cached or not. If the socket exists, fill in the core_id and thread_id
- * in cpuinfo
- */
-static int __devinit
-check_for_new_socket(__u16 logical_address, struct cpuinfo_ia64 *c)
-{
- int i;
- __u32 sid = c->socket_id;
-
- for_each_cpu(i) {
- if (mt_info[i].valid && mt_info[i].proc_fixed_addr == logical_address
- && mt_info[i].socket_id == sid) {
- c->core_id = mt_info[i].core_id;
- c->thread_id = mt_info[i].thread_id;
- return 1; /* not a new socket */
- }
- }
- return 0;
-}
-
/*
* identify_siblings(cpu) gets called from identify_cpu. This populates the
* information related to logical execution units in per_cpu_data structure.
void __devinit
identify_siblings(struct cpuinfo_ia64 *c)
{
- s64 status;
+ long status;
u16 pltid;
- u64 proc_fixed_addr;
- int count, i;
pal_logical_to_physical_t info;
- if (smp_num_cpucores == 1 && smp_num_siblings == 1)
- return;
+ status = ia64_pal_logical_to_phys(-1, &info);
+ if (status != PAL_STATUS_SUCCESS) {
+ if (status != PAL_STATUS_UNIMPLEMENTED) {
+ printk(KERN_ERR
+ "ia64_pal_logical_to_phys failed with %ld\n",
+ status);
+ return;
+ }
- if ((status = ia64_pal_logical_to_phys(0, &info)) != PAL_STATUS_SUCCESS) {
- printk(KERN_ERR "ia64_pal_logical_to_phys failed with %ld\n",
- status);
- return;
+ info.overview_ppid = 0;
+ info.overview_cpp = 1;
+ info.overview_tpc = 1;
}
- if ((status = ia64_sal_physical_id_info(&pltid)) != PAL_STATUS_SUCCESS) {
- printk(KERN_ERR "ia64_sal_pltid failed with %ld\n", status);
- return;
- }
- if ((status = ia64_pal_fixed_addr(&proc_fixed_addr)) != PAL_STATUS_SUCCESS) {
- printk(KERN_ERR "ia64_pal_fixed_addr failed with %ld\n", status);
+
+ status = ia64_sal_physical_id_info(&pltid);
+ if (status != PAL_STATUS_SUCCESS) {
+ if (status != PAL_STATUS_UNIMPLEMENTED)
+ printk(KERN_ERR
+ "ia64_sal_pltid failed with %ld\n",
+ status);
return;
}
c->socket_id = (pltid << 8) | info.overview_ppid;
- c->cores_per_socket = info.overview_cpp;
- c->threads_per_core = info.overview_tpc;
- count = c->num_log = info.overview_num_log;
- /* If the thread and core id information is already cached, then
- * we will simply update cpu_info and return. Otherwise, we will
- * do the PAL calls and cache core and thread id's of all the siblings.
- */
- if (check_for_new_socket(proc_fixed_addr, c))
+ if (info.overview_cpp == 1 && info.overview_tpc == 1)
return;
- for (i = 0; i < count; i++) {
- int index;
-
- if (i && (status = ia64_pal_logical_to_phys(i, &info))
- != PAL_STATUS_SUCCESS) {
- printk(KERN_ERR "ia64_pal_logical_to_phys failed"
- " with %ld\n", status);
- return;
- }
- if (info.log2_la == proc_fixed_addr) {
- c->core_id = info.log1_cid;
- c->thread_id = info.log1_tid;
- }
+ c->cores_per_socket = info.overview_cpp;
+ c->threads_per_core = info.overview_tpc;
+ c->num_log = info.overview_num_log;
- index = check_for_mtinfo_index();
- /* We will not do the mt_info caching optimization in this case.
- */
- if (index < 0)
- continue;
+ c->core_id = info.log1_cid;
+ c->thread_id = info.log1_tid;
+}
- mt_info[index].valid = 1;
- mt_info[index].socket_id = c->socket_id;
- mt_info[index].core_id = info.log1_cid;
- mt_info[index].thread_id = info.log1_tid;
- mt_info[index].proc_fixed_addr = info.log2_la;
+/*
+ * returns non zero, if multi-threading is enabled
+ * on at least one physical package. Due to hotplug cpu
+ * and (maxcpus=), all threads may not necessarily be enabled
+ * even though the processor supports multi-threading.
+ */
+int is_multithreading_enabled(void)
+{
+ int i, j;
+
+ for_each_present_cpu(i) {
+ for_each_present_cpu(j) {
+ if (j == i)
+ continue;
+ if ((cpu_data(j)->socket_id == cpu_data(i)->socket_id)) {
+ if (cpu_data(j)->core_id == cpu_data(i)->core_id)
+ return 1;
+ }
+ }
}
+ return 0;
}
+EXPORT_SYMBOL_GPL(is_multithreading_enabled);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff