[SPARC64]: dr-cpu unconfigure support.
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/arch/sparc64/kernel/ds.c b/arch/sparc64/kernel/ds.c
index b633171..1c58710 100644
--- a/arch/sparc64/kernel/ds.c
+++ b/arch/sparc64/kernel/ds.c
@@ -20,6 +20,7 @@
#include <asm/power.h>
#include <asm/mdesc.h>
#include <asm/head.h>
+#include <asm/irq.h>
#define DRV_MODULE_NAME "ds"
#define PFX DRV_MODULE_NAME ": "
@@ -559,6 +560,9 @@
kfree(resp);
+ /* Redistribute IRQs, taking into account the new cpus. */
+ fixup_irqs();
+
return 0;
}
@@ -566,7 +570,8 @@
cpumask_t *mask)
{
struct ds_data *resp;
- int resp_len, ncpus;
+ int resp_len, ncpus, cpu;
+ unsigned long flags;
ncpus = cpus_weight(*mask);
resp_len = dr_cpu_size_response(ncpus);
@@ -578,9 +583,25 @@
resp_len, ncpus, mask,
DR_CPU_STAT_UNCONFIGURED);
+ for_each_cpu_mask(cpu, *mask) {
+ int err;
+
+ printk(KERN_INFO PFX "CPU[%d]: Shutting down cpu %d...\n",
+ smp_processor_id(), cpu);
+ err = cpu_down(cpu);
+ if (err)
+ dr_cpu_mark(resp, cpu, ncpus,
+ DR_CPU_RES_FAILURE,
+ DR_CPU_STAT_CONFIGURED);
+ }
+
+ spin_lock_irqsave(&ds_lock, flags);
+ ds_send(ds_info->lp, resp, resp_len);
+ spin_unlock_irqrestore(&ds_lock, flags);
+
kfree(resp);
- return -EOPNOTSUPP;
+ return 0;
}
static void process_dr_cpu_list(struct ds_cap_state *cp)
diff --git a/arch/sparc64/kernel/irq.c b/arch/sparc64/kernel/irq.c
index a1c916f..8cb3358 100644
--- a/arch/sparc64/kernel/irq.c
+++ b/arch/sparc64/kernel/irq.c
@@ -803,6 +803,26 @@
set_irq_regs(old_regs);
}
+#ifdef CONFIG_HOTPLUG_CPU
+void fixup_irqs(void)
+{
+ unsigned int irq;
+
+ for (irq = 0; irq < NR_IRQS; irq++) {
+ unsigned long flags;
+
+ spin_lock_irqsave(&irq_desc[irq].lock, flags);
+ if (irq_desc[irq].action &&
+ !(irq_desc[irq].status & IRQ_PER_CPU)) {
+ if (irq_desc[irq].chip->set_affinity)
+ irq_desc[irq].chip->set_affinity(irq,
+ irq_desc[irq].affinity);
+ }
+ spin_unlock_irqrestore(&irq_desc[irq].lock, flags);
+ }
+}
+#endif
+
struct sun5_timer {
u64 count0;
u64 limit0;
diff --git a/arch/sparc64/kernel/process.c b/arch/sparc64/kernel/process.c
index f5f97e2..9355750 100644
--- a/arch/sparc64/kernel/process.c
+++ b/arch/sparc64/kernel/process.c
@@ -29,6 +29,7 @@
#include <linux/compat.h>
#include <linux/tick.h>
#include <linux/init.h>
+#include <linux/cpu.h>
#include <asm/oplib.h>
#include <asm/uaccess.h>
@@ -49,7 +50,7 @@
/* #define VERBOSE_SHOWREGS */
-static void sparc64_yield(void)
+static void sparc64_yield(int cpu)
{
if (tlb_type != hypervisor)
return;
@@ -57,7 +58,7 @@
clear_thread_flag(TIF_POLLING_NRFLAG);
smp_mb__after_clear_bit();
- while (!need_resched()) {
+ while (!need_resched() && !cpu_is_offline(cpu)) {
unsigned long pstate;
/* Disable interrupts. */
@@ -68,7 +69,7 @@
: "=&r" (pstate)
: "i" (PSTATE_IE));
- if (!need_resched())
+ if (!need_resched() && !cpu_is_offline(cpu))
sun4v_cpu_yield();
/* Re-enable interrupts. */
@@ -86,15 +87,25 @@
/* The idle loop on sparc64. */
void cpu_idle(void)
{
+ int cpu = smp_processor_id();
+
set_thread_flag(TIF_POLLING_NRFLAG);
while(1) {
tick_nohz_stop_sched_tick();
- while (!need_resched())
- sparc64_yield();
+
+ while (!need_resched() && !cpu_is_offline(cpu))
+ sparc64_yield(cpu);
+
tick_nohz_restart_sched_tick();
preempt_enable_no_resched();
+
+#ifdef CONFIG_HOTPLUG_CPU
+ if (cpu_is_offline(cpu))
+ cpu_play_dead();
+#endif
+
schedule();
preempt_disable();
}
diff --git a/arch/sparc64/kernel/smp.c b/arch/sparc64/kernel/smp.c
index e038ae6..b448d33 100644
--- a/arch/sparc64/kernel/smp.c
+++ b/arch/sparc64/kernel/smp.c
@@ -44,6 +44,7 @@
#include <asm/prom.h>
#include <asm/mdesc.h>
#include <asm/ldc.h>
+#include <asm/hypervisor.h>
extern void calibrate_delay(void);
@@ -62,7 +63,6 @@
EXPORT_SYMBOL(cpu_core_map);
static cpumask_t smp_commenced_mask;
-static cpumask_t cpu_callout_map;
void smp_info(struct seq_file *m)
{
@@ -83,6 +83,8 @@
i, cpu_data(i).clock_tick);
}
+static __cacheline_aligned_in_smp DEFINE_SPINLOCK(call_lock);
+
extern void setup_sparc64_timer(void);
static volatile unsigned long callin_flag = 0;
@@ -121,7 +123,9 @@
while (!cpu_isset(cpuid, smp_commenced_mask))
rmb();
+ spin_lock(&call_lock);
cpu_set(cpuid, cpu_online_map);
+ spin_unlock(&call_lock);
/* idle thread is expected to have preempt disabled */
preempt_disable();
@@ -324,6 +328,9 @@
hv_err = sun4v_cpu_start(cpu, trampoline_ra,
kimage_addr_to_ra(&sparc64_ttable_tl0),
__pa(hdesc));
+ if (hv_err)
+ printk(KERN_ERR "ldom_startcpu_cpuid: sun4v_cpu_start() "
+ "gives error %lu\n", hv_err);
}
#endif
@@ -350,7 +357,6 @@
p = fork_idle(cpu);
callin_flag = 0;
cpu_new_thread = task_thread_info(p);
- cpu_set(cpu, cpu_callout_map);
if (tlb_type == hypervisor) {
/* Alloc the mondo queues, cpu will load them. */
@@ -379,7 +385,6 @@
ret = 0;
} else {
printk("Processor %d is stuck.\n", cpu);
- cpu_clear(cpu, cpu_callout_map);
ret = -ENODEV;
}
cpu_new_thread = NULL;
@@ -791,7 +796,6 @@
int wait;
};
-static __cacheline_aligned_in_smp DEFINE_SPINLOCK(call_lock);
static struct call_data_struct *call_data;
extern unsigned long xcall_call_function;
@@ -1241,7 +1245,7 @@
{
unsigned int i;
- for_each_possible_cpu(i) {
+ for_each_present_cpu(i) {
unsigned int j;
cpus_clear(cpu_core_map[i]);
@@ -1250,14 +1254,14 @@
continue;
}
- for_each_possible_cpu(j) {
+ for_each_present_cpu(j) {
if (cpu_data(i).core_id ==
cpu_data(j).core_id)
cpu_set(j, cpu_core_map[i]);
}
}
- for_each_possible_cpu(i) {
+ for_each_present_cpu(i) {
unsigned int j;
cpus_clear(cpu_sibling_map[i]);
@@ -1266,7 +1270,7 @@
continue;
}
- for_each_possible_cpu(j) {
+ for_each_present_cpu(j) {
if (cpu_data(i).proc_id ==
cpu_data(j).proc_id)
cpu_set(j, cpu_sibling_map[i]);
@@ -1296,16 +1300,106 @@
}
#ifdef CONFIG_HOTPLUG_CPU
+void cpu_play_dead(void)
+{
+ int cpu = smp_processor_id();
+ unsigned long pstate;
+
+ idle_task_exit();
+
+ if (tlb_type == hypervisor) {
+ struct trap_per_cpu *tb = &trap_block[cpu];
+
+ sun4v_cpu_qconf(HV_CPU_QUEUE_CPU_MONDO,
+ tb->cpu_mondo_pa, 0);
+ sun4v_cpu_qconf(HV_CPU_QUEUE_DEVICE_MONDO,
+ tb->dev_mondo_pa, 0);
+ sun4v_cpu_qconf(HV_CPU_QUEUE_RES_ERROR,
+ tb->resum_mondo_pa, 0);
+ sun4v_cpu_qconf(HV_CPU_QUEUE_NONRES_ERROR,
+ tb->nonresum_mondo_pa, 0);
+ }
+
+ cpu_clear(cpu, smp_commenced_mask);
+ membar_safe("#Sync");
+
+ local_irq_disable();
+
+ __asm__ __volatile__(
+ "rdpr %%pstate, %0\n\t"
+ "wrpr %0, %1, %%pstate"
+ : "=r" (pstate)
+ : "i" (PSTATE_IE));
+
+ while (1)
+ barrier();
+}
+
int __cpu_disable(void)
{
- printk(KERN_ERR "SMP: __cpu_disable() on cpu %d\n",
- smp_processor_id());
- return -ENODEV;
+ int cpu = smp_processor_id();
+ cpuinfo_sparc *c;
+ int i;
+
+ for_each_cpu_mask(i, cpu_core_map[cpu])
+ cpu_clear(cpu, cpu_core_map[i]);
+ cpus_clear(cpu_core_map[cpu]);
+
+ for_each_cpu_mask(i, cpu_sibling_map[cpu])
+ cpu_clear(cpu, cpu_sibling_map[i]);
+ cpus_clear(cpu_sibling_map[cpu]);
+
+ c = &cpu_data(cpu);
+
+ c->core_id = 0;
+ c->proc_id = -1;
+
+ spin_lock(&call_lock);
+ cpu_clear(cpu, cpu_online_map);
+ spin_unlock(&call_lock);
+
+ smp_wmb();
+
+ /* Make sure no interrupts point to this cpu. */
+ fixup_irqs();
+
+ local_irq_enable();
+ mdelay(1);
+ local_irq_disable();
+
+ return 0;
}
void __cpu_die(unsigned int cpu)
{
- printk(KERN_ERR "SMP: __cpu_die(%u)\n", cpu);
+ int i;
+
+ for (i = 0; i < 100; i++) {
+ smp_rmb();
+ if (!cpu_isset(cpu, smp_commenced_mask))
+ break;
+ msleep(100);
+ }
+ if (cpu_isset(cpu, smp_commenced_mask)) {
+ printk(KERN_ERR "CPU %u didn't die...\n", cpu);
+ } else {
+#if defined(CONFIG_SUN_LDOMS)
+ unsigned long hv_err;
+ int limit = 100;
+
+ do {
+ hv_err = sun4v_cpu_stop(cpu);
+ if (hv_err == HV_EOK) {
+ cpu_clear(cpu, cpu_present_map);
+ break;
+ }
+ } while (--limit > 0);
+ if (limit <= 0) {
+ printk(KERN_ERR "sun4v_cpu_stop() fails err=%lu\n",
+ hv_err);
+ }
+#endif
+ }
}
#endif
