This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / arch / powerpc / kernel / smp.c
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
new file mode 100644 (file)
index 0000000..530f7db
--- /dev/null
@@ -0,0 +1,595 @@
+/*
+ * SMP support for ppc.
+ *
+ * Written by Cort Dougan (cort@cs.nmt.edu) borrowing a great
+ * deal of code from the sparc and intel versions.
+ *
+ * Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
+ *
+ * PowerPC-64 Support added by Dave Engebretsen, Peter Bergner, and
+ * Mike Corrigan {engebret|bergner|mikec}@us.ibm.com
+ *
+ *      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.
+ */
+
+#undef DEBUG
+
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/interrupt.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/spinlock.h>
+#include <linux/cache.h>
+#include <linux/err.h>
+#include <linux/sysdev.h>
+#include <linux/cpu.h>
+#include <linux/notifier.h>
+#include <linux/topology.h>
+
+#include <asm/ptrace.h>
+#include <asm/atomic.h>
+#include <asm/irq.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/prom.h>
+#include <asm/smp.h>
+#include <asm/time.h>
+#include <asm/machdep.h>
+#include <asm/cputable.h>
+#include <asm/system.h>
+#include <asm/mpic.h>
+#include <asm/vdso_datapage.h>
+#ifdef CONFIG_PPC64
+#include <asm/paca.h>
+#endif
+
+#ifdef DEBUG
+#include <asm/udbg.h>
+#define DBG(fmt...) udbg_printf(fmt)
+#else
+#define DBG(fmt...)
+#endif
+
+int smp_hw_index[NR_CPUS];
+struct thread_info *secondary_ti;
+
+cpumask_t cpu_possible_map = CPU_MASK_NONE;
+cpumask_t cpu_online_map = CPU_MASK_NONE;
+cpumask_t cpu_sibling_map[NR_CPUS] = { [0 ... NR_CPUS-1] = CPU_MASK_NONE };
+
+EXPORT_SYMBOL(cpu_online_map);
+EXPORT_SYMBOL(cpu_possible_map);
+
+/* SMP operations for this machine */
+struct smp_ops_t *smp_ops;
+
+static volatile unsigned int cpu_callin_map[NR_CPUS];
+
+void smp_call_function_interrupt(void);
+
+int smt_enabled_at_boot = 1;
+
+static void (*crash_ipi_function_ptr)(struct pt_regs *) = NULL;
+
+#ifdef CONFIG_MPIC
+int __init smp_mpic_probe(void)
+{
+       int nr_cpus;
+
+       DBG("smp_mpic_probe()...\n");
+
+       nr_cpus = cpus_weight(cpu_possible_map);
+
+       DBG("nr_cpus: %d\n", nr_cpus);
+
+       if (nr_cpus > 1)
+               mpic_request_ipis();
+
+       return nr_cpus;
+}
+
+void __devinit smp_mpic_setup_cpu(int cpu)
+{
+       mpic_setup_this_cpu();
+}
+#endif /* CONFIG_MPIC */
+
+#ifdef CONFIG_PPC64
+void __devinit smp_generic_kick_cpu(int nr)
+{
+       BUG_ON(nr < 0 || nr >= NR_CPUS);
+
+       /*
+        * The processor is currently spinning, waiting for the
+        * cpu_start field to become non-zero After we set cpu_start,
+        * the processor will continue on to secondary_start
+        */
+       paca[nr].cpu_start = 1;
+       smp_mb();
+}
+#endif
+
+void smp_message_recv(int msg, struct pt_regs *regs)
+{
+       switch(msg) {
+       case PPC_MSG_CALL_FUNCTION:
+               smp_call_function_interrupt();
+               break;
+       case PPC_MSG_RESCHEDULE:
+               /* XXX Do we have to do this? */
+               set_need_resched();
+               break;
+       case PPC_MSG_DEBUGGER_BREAK:
+               if (crash_ipi_function_ptr) {
+                       crash_ipi_function_ptr(regs);
+                       break;
+               }
+#ifdef CONFIG_DEBUGGER
+               debugger_ipi(regs);
+               break;
+#endif /* CONFIG_DEBUGGER */
+               /* FALLTHROUGH */
+       default:
+               printk("SMP %d: smp_message_recv(): unknown msg %d\n",
+                      smp_processor_id(), msg);
+               break;
+       }
+}
+
+void smp_send_reschedule(int cpu)
+{
+       smp_ops->message_pass(cpu, PPC_MSG_RESCHEDULE);
+}
+
+#ifdef CONFIG_DEBUGGER
+void smp_send_debugger_break(int cpu)
+{
+       smp_ops->message_pass(cpu, PPC_MSG_DEBUGGER_BREAK);
+}
+#endif
+
+#ifdef CONFIG_KEXEC
+void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *))
+{
+       crash_ipi_function_ptr = crash_ipi_callback;
+       if (crash_ipi_callback) {
+               mb();
+               smp_ops->message_pass(MSG_ALL_BUT_SELF, PPC_MSG_DEBUGGER_BREAK);
+       }
+}
+#endif
+
+static void stop_this_cpu(void *dummy)
+{
+       local_irq_disable();
+       while (1)
+               ;
+}
+
+void smp_send_stop(void)
+{
+       smp_call_function(stop_this_cpu, NULL, 1, 0);
+}
+
+/*
+ * Structure and data for smp_call_function(). This is designed to minimise
+ * static memory requirements. It also looks cleaner.
+ * Stolen from the i386 version.
+ */
+static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(call_lock);
+
+static struct call_data_struct {
+       void (*func) (void *info);
+       void *info;
+       atomic_t started;
+       atomic_t finished;
+       int wait;
+} *call_data;
+
+/* delay of at least 8 seconds */
+#define SMP_CALL_TIMEOUT       8
+
+/*
+ * This function sends a 'generic call function' IPI to all other CPUs
+ * in the system.
+ *
+ * [SUMMARY] Run a function on all other CPUs.
+ * <func> The function to run. This must be fast and non-blocking.
+ * <info> An arbitrary pointer to pass to the function.
+ * <nonatomic> currently unused.
+ * <wait> If true, wait (atomically) until function has completed on other CPUs.
+ * [RETURNS] 0 on success, else a negative status code. Does not return until
+ * remote CPUs are nearly ready to execute <<func>> or are or have executed.
+ *
+ * You must not call this function with disabled interrupts or from a
+ * hardware interrupt handler or from a bottom half handler.
+ */
+int smp_call_function (void (*func) (void *info), void *info, int nonatomic,
+                      int wait)
+{ 
+       struct call_data_struct data;
+       int ret = -1, cpus;
+       u64 timeout;
+
+       /* Can deadlock when called with interrupts disabled */
+       WARN_ON(irqs_disabled());
+
+       data.func = func;
+       data.info = info;
+       atomic_set(&data.started, 0);
+       data.wait = wait;
+       if (wait)
+               atomic_set(&data.finished, 0);
+
+       spin_lock(&call_lock);
+       /* Must grab online cpu count with preempt disabled, otherwise
+        * it can change. */
+       cpus = num_online_cpus() - 1;
+       if (!cpus) {
+               ret = 0;
+               goto out;
+       }
+
+       call_data = &data;
+       smp_wmb();
+       /* Send a message to all other CPUs and wait for them to respond */
+       smp_ops->message_pass(MSG_ALL_BUT_SELF, PPC_MSG_CALL_FUNCTION);
+
+       timeout = get_tb() + (u64) SMP_CALL_TIMEOUT * tb_ticks_per_sec;
+
+       /* Wait for response */
+       while (atomic_read(&data.started) != cpus) {
+               HMT_low();
+               if (get_tb() >= timeout) {
+                       printk("smp_call_function on cpu %d: other cpus not "
+                              "responding (%d)\n", smp_processor_id(),
+                              atomic_read(&data.started));
+                       debugger(NULL);
+                       goto out;
+               }
+       }
+
+       if (wait) {
+               while (atomic_read(&data.finished) != cpus) {
+                       HMT_low();
+                       if (get_tb() >= timeout) {
+                               printk("smp_call_function on cpu %d: other "
+                                      "cpus not finishing (%d/%d)\n",
+                                      smp_processor_id(),
+                                      atomic_read(&data.finished),
+                                      atomic_read(&data.started));
+                               debugger(NULL);
+                               goto out;
+                       }
+               }
+       }
+
+       ret = 0;
+
+ out:
+       call_data = NULL;
+       HMT_medium();
+       spin_unlock(&call_lock);
+       return ret;
+}
+
+EXPORT_SYMBOL(smp_call_function);
+
+void smp_call_function_interrupt(void)
+{
+       void (*func) (void *info);
+       void *info;
+       int wait;
+
+       /* call_data will be NULL if the sender timed out while
+        * waiting on us to receive the call.
+        */
+       if (!call_data)
+               return;
+
+       func = call_data->func;
+       info = call_data->info;
+       wait = call_data->wait;
+
+       if (!wait)
+               smp_mb__before_atomic_inc();
+
+       /*
+        * Notify initiating CPU that I've grabbed the data and am
+        * about to execute the function
+        */
+       atomic_inc(&call_data->started);
+       /*
+        * At this point the info structure may be out of scope unless wait==1
+        */
+       (*func)(info);
+       if (wait) {
+               smp_mb__before_atomic_inc();
+               atomic_inc(&call_data->finished);
+       }
+}
+
+extern struct gettimeofday_struct do_gtod;
+
+struct thread_info *current_set[NR_CPUS];
+
+DECLARE_PER_CPU(unsigned int, pvr);
+
+static void __devinit smp_store_cpu_info(int id)
+{
+       per_cpu(pvr, id) = mfspr(SPRN_PVR);
+}
+
+static void __init smp_create_idle(unsigned int cpu)
+{
+       struct task_struct *p;
+
+       /* create a process for the processor */
+       p = fork_idle(cpu);
+       if (IS_ERR(p))
+               panic("failed fork for CPU %u: %li", cpu, PTR_ERR(p));
+#ifdef CONFIG_PPC64
+       paca[cpu].__current = p;
+#endif
+       current_set[cpu] = task_thread_info(p);
+       task_thread_info(p)->cpu = cpu;
+}
+
+void __init smp_prepare_cpus(unsigned int max_cpus)
+{
+       unsigned int cpu;
+
+       DBG("smp_prepare_cpus\n");
+
+       /* 
+        * setup_cpu may need to be called on the boot cpu. We havent
+        * spun any cpus up but lets be paranoid.
+        */
+       BUG_ON(boot_cpuid != smp_processor_id());
+
+       /* Fixup boot cpu */
+       smp_store_cpu_info(boot_cpuid);
+       cpu_callin_map[boot_cpuid] = 1;
+
+       max_cpus = smp_ops->probe();
+       smp_space_timers(max_cpus);
+
+       for_each_possible_cpu(cpu)
+               if (cpu != boot_cpuid)
+                       smp_create_idle(cpu);
+}
+
+void __devinit smp_prepare_boot_cpu(void)
+{
+       BUG_ON(smp_processor_id() != boot_cpuid);
+
+       cpu_set(boot_cpuid, cpu_online_map);
+#ifdef CONFIG_PPC64
+       paca[boot_cpuid].__current = current;
+#endif
+       current_set[boot_cpuid] = task_thread_info(current);
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+/* State of each CPU during hotplug phases */
+DEFINE_PER_CPU(int, cpu_state) = { 0 };
+
+int generic_cpu_disable(void)
+{
+       unsigned int cpu = smp_processor_id();
+
+       if (cpu == boot_cpuid)
+               return -EBUSY;
+
+       cpu_clear(cpu, cpu_online_map);
+#ifdef CONFIG_PPC64
+       vdso_data->processorCount--;
+       fixup_irqs(cpu_online_map);
+#endif
+       return 0;
+}
+
+int generic_cpu_enable(unsigned int cpu)
+{
+       /* Do the normal bootup if we haven't
+        * already bootstrapped. */
+       if (system_state != SYSTEM_RUNNING)
+               return -ENOSYS;
+
+       /* get the target out of it's holding state */
+       per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
+       smp_wmb();
+
+       while (!cpu_online(cpu))
+               cpu_relax();
+
+#ifdef CONFIG_PPC64
+       fixup_irqs(cpu_online_map);
+       /* counter the irq disable in fixup_irqs */
+       local_irq_enable();
+#endif
+       return 0;
+}
+
+void generic_cpu_die(unsigned int cpu)
+{
+       int i;
+
+       for (i = 0; i < 100; i++) {
+               smp_rmb();
+               if (per_cpu(cpu_state, cpu) == CPU_DEAD)
+                       return;
+               msleep(100);
+       }
+       printk(KERN_ERR "CPU%d didn't die...\n", cpu);
+}
+
+void generic_mach_cpu_die(void)
+{
+       unsigned int cpu;
+
+       local_irq_disable();
+       cpu = smp_processor_id();
+       printk(KERN_DEBUG "CPU%d offline\n", cpu);
+       __get_cpu_var(cpu_state) = CPU_DEAD;
+       smp_wmb();
+       while (__get_cpu_var(cpu_state) != CPU_UP_PREPARE)
+               cpu_relax();
+
+#ifdef CONFIG_PPC64
+       flush_tlb_pending();
+#endif
+       cpu_set(cpu, cpu_online_map);
+       local_irq_enable();
+}
+#endif
+
+static int __devinit cpu_enable(unsigned int cpu)
+{
+       if (smp_ops->cpu_enable)
+               return smp_ops->cpu_enable(cpu);
+
+       return -ENOSYS;
+}
+
+int __devinit __cpu_up(unsigned int cpu)
+{
+       int c;
+
+       secondary_ti = current_set[cpu];
+       if (!cpu_enable(cpu))
+               return 0;
+
+       if (smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu))
+               return -EINVAL;
+
+       /* Make sure callin-map entry is 0 (can be leftover a CPU
+        * hotplug
+        */
+       cpu_callin_map[cpu] = 0;
+
+       /* The information for processor bringup must
+        * be written out to main store before we release
+        * the processor.
+        */
+       smp_mb();
+
+       /* wake up cpus */
+       DBG("smp: kicking cpu %d\n", cpu);
+       smp_ops->kick_cpu(cpu);
+
+       /*
+        * wait to see if the cpu made a callin (is actually up).
+        * use this value that I found through experimentation.
+        * -- Cort
+        */
+       if (system_state < SYSTEM_RUNNING)
+               for (c = 5000; c && !cpu_callin_map[cpu]; c--)
+                       udelay(100);
+#ifdef CONFIG_HOTPLUG_CPU
+       else
+               /*
+                * CPUs can take much longer to come up in the
+                * hotplug case.  Wait five seconds.
+                */
+               for (c = 25; c && !cpu_callin_map[cpu]; c--) {
+                       msleep(200);
+               }
+#endif
+
+       if (!cpu_callin_map[cpu]) {
+               printk("Processor %u is stuck.\n", cpu);
+               return -ENOENT;
+       }
+
+       printk("Processor %u found.\n", cpu);
+
+       if (smp_ops->give_timebase)
+               smp_ops->give_timebase();
+
+       /* Wait until cpu puts itself in the online map */
+       while (!cpu_online(cpu))
+               cpu_relax();
+
+       return 0;
+}
+
+
+/* Activate a secondary processor. */
+int __devinit start_secondary(void *unused)
+{
+       unsigned int cpu = smp_processor_id();
+
+       atomic_inc(&init_mm.mm_count);
+       current->active_mm = &init_mm;
+
+       smp_store_cpu_info(cpu);
+       set_dec(tb_ticks_per_jiffy);
+       preempt_disable();
+       cpu_callin_map[cpu] = 1;
+
+       smp_ops->setup_cpu(cpu);
+       if (smp_ops->take_timebase)
+               smp_ops->take_timebase();
+
+       if (system_state > SYSTEM_BOOTING)
+               snapshot_timebase();
+
+       spin_lock(&call_lock);
+       cpu_set(cpu, cpu_online_map);
+       spin_unlock(&call_lock);
+
+       local_irq_enable();
+
+       cpu_idle();
+       return 0;
+}
+
+int setup_profiling_timer(unsigned int multiplier)
+{
+       return 0;
+}
+
+void __init smp_cpus_done(unsigned int max_cpus)
+{
+       cpumask_t old_mask;
+
+       /* We want the setup_cpu() here to be called from CPU 0, but our
+        * init thread may have been "borrowed" by another CPU in the meantime
+        * se we pin us down to CPU 0 for a short while
+        */
+       old_mask = current->cpus_allowed;
+       set_cpus_allowed(current, cpumask_of_cpu(boot_cpuid));
+       
+       smp_ops->setup_cpu(boot_cpuid);
+
+       set_cpus_allowed(current, old_mask);
+
+       snapshot_timebases();
+
+       dump_numa_cpu_topology();
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+int __cpu_disable(void)
+{
+       if (smp_ops->cpu_disable)
+               return smp_ops->cpu_disable();
+
+       return -ENOSYS;
+}
+
+void __cpu_die(unsigned int cpu)
+{
+       if (smp_ops->cpu_die)
+               smp_ops->cpu_die(cpu);
+}
+#endif