Hexagon: Add SMP support
authorRichard Kuo <rkuo@codeaurora.org>
Mon, 31 Oct 2011 23:46:34 +0000 (18:46 -0500)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 1 Nov 2011 14:34:20 +0000 (07:34 -0700)
Signed-off-by: Richard Kuo <rkuo@codeaurora.org>
Signed-off-by: Linas Vepstas <linas@codeaurora.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/hexagon/include/asm/smp.h [new file with mode: 0644]
arch/hexagon/kernel/smp.c [new file with mode: 0644]
arch/hexagon/kernel/topology.c [new file with mode: 0644]

diff --git a/arch/hexagon/include/asm/smp.h b/arch/hexagon/include/asm/smp.h
new file mode 100644 (file)
index 0000000..87c869a
--- /dev/null
@@ -0,0 +1,44 @@
+/*
+ * SMP definitions for the Hexagon architecture
+ *
+ * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#ifndef __ASM_SMP_H
+#define __ASM_SMP_H
+
+#include <linux/cpumask.h>
+
+#define raw_smp_processor_id() (current_thread_info()->cpu)
+
+enum ipi_message_type {
+       IPI_NOP = 0,
+       IPI_RESCHEDULE = 1,
+       IPI_CALL_FUNC,
+       IPI_CALL_FUNC_SINGLE,
+       IPI_CPU_STOP,
+       IPI_TIMER,
+};
+
+extern void send_ipi(const struct cpumask *cpumask, enum ipi_message_type msg);
+extern void smp_start_cpus(void);
+extern void arch_send_call_function_single_ipi(int cpu);
+extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
+
+extern void smp_vm_unmask_irq(void *info);
+
+#endif
diff --git a/arch/hexagon/kernel/smp.c b/arch/hexagon/kernel/smp.c
new file mode 100644 (file)
index 0000000..c871a2c
--- /dev/null
@@ -0,0 +1,276 @@
+/*
+ * SMP support for Hexagon
+ *
+ * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#include <linux/err.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/percpu.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/spinlock.h>
+
+#include <asm/system.h>  /*  xchg  */
+#include <asm/time.h>    /*  timer_interrupt  */
+#include <asm/hexagon_vm.h>
+
+#define BASE_IPI_IRQ 26
+
+/*
+ * cpu_possible_map needs to be filled out prior to setup_per_cpu_areas
+ * (which is prior to any of our smp_prepare_cpu crap), in order to set
+ * up the...  per_cpu areas.
+ */
+
+struct ipi_data {
+       unsigned long bits;
+};
+
+static DEFINE_PER_CPU(struct ipi_data, ipi_data);
+
+static inline void __handle_ipi(unsigned long *ops, struct ipi_data *ipi,
+                               int cpu)
+{
+       unsigned long msg = 0;
+       do {
+               msg = find_next_bit(ops, BITS_PER_LONG, msg+1);
+
+               switch (msg) {
+
+               case IPI_TIMER:
+                       ipi_timer();
+                       break;
+
+               case IPI_CALL_FUNC:
+                       generic_smp_call_function_interrupt();
+                       break;
+
+               case IPI_CALL_FUNC_SINGLE:
+                       generic_smp_call_function_single_interrupt();
+                       break;
+
+               case IPI_CPU_STOP:
+                       /*
+                        * call vmstop()
+                        */
+                       __vmstop();
+                       break;
+
+               case IPI_RESCHEDULE:
+                       scheduler_ipi();
+                       break;
+               }
+       } while (msg < BITS_PER_LONG);
+}
+
+/*  Used for IPI call from other CPU's to unmask int  */
+void smp_vm_unmask_irq(void *info)
+{
+       __vmintop_locen((long) info);
+}
+
+
+/*
+ * This is based on Alpha's IPI stuff.
+ * Supposed to take (int, void*) as args now.
+ * Specifically, first arg is irq, second is the irq_desc.
+ */
+
+irqreturn_t handle_ipi(int irq, void *desc)
+{
+       int cpu = smp_processor_id();
+       struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
+       unsigned long ops;
+
+       while ((ops = xchg(&ipi->bits, 0)) != 0)
+               __handle_ipi(&ops, ipi, cpu);
+       return IRQ_HANDLED;
+}
+
+void send_ipi(const struct cpumask *cpumask, enum ipi_message_type msg)
+{
+       unsigned long flags;
+       unsigned long cpu;
+       unsigned long retval;
+
+       local_irq_save(flags);
+
+       for_each_cpu(cpu, cpumask) {
+               struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
+
+               set_bit(msg, &ipi->bits);
+               /*  Possible barrier here  */
+               retval = __vmintop_post(BASE_IPI_IRQ+cpu);
+
+               if (retval != 0) {
+                       printk(KERN_ERR "interrupt %ld not configured?\n",
+                               BASE_IPI_IRQ+cpu);
+               }
+       }
+
+       local_irq_restore(flags);
+}
+
+static struct irqaction ipi_intdesc = {
+       .handler = handle_ipi,
+       .flags = IRQF_TRIGGER_RISING,
+       .name = "ipi_handler"
+};
+
+void __init smp_prepare_boot_cpu(void)
+{
+}
+
+/*
+ * interrupts should already be disabled from the VM
+ * SP should already be correct; need to set THREADINFO_REG
+ * to point to current thread info
+ */
+
+void __cpuinit start_secondary(void)
+{
+       unsigned int cpu;
+       unsigned long thread_ptr;
+
+       /*  Calculate thread_info pointer from stack pointer  */
+       __asm__ __volatile__(
+               "%0 = SP;\n"
+               : "=r" (thread_ptr)
+       );
+
+       thread_ptr = thread_ptr & ~(THREAD_SIZE-1);
+
+       __asm__ __volatile__(
+               QUOTED_THREADINFO_REG " = %0;\n"
+               :
+               : "r" (thread_ptr)
+       );
+
+       /*  Set the memory struct  */
+       atomic_inc(&init_mm.mm_count);
+       current->active_mm = &init_mm;
+
+       cpu = smp_processor_id();
+
+       setup_irq(BASE_IPI_IRQ + cpu, &ipi_intdesc);
+
+       /*  Register the clock_event dummy  */
+       setup_percpu_clockdev();
+
+       printk(KERN_INFO "%s cpu %d\n", __func__, current_thread_info()->cpu);
+
+       set_cpu_online(cpu, true);
+       while (!cpumask_test_cpu(cpu, cpu_active_mask))
+               cpu_relax();
+       local_irq_enable();
+
+       cpu_idle();
+}
+
+
+/*
+ * called once for each present cpu
+ * apparently starts up the CPU and then
+ * maintains control until "cpu_online(cpu)" is set.
+ */
+
+int __cpuinit __cpu_up(unsigned int cpu)
+{
+       struct task_struct *idle;
+       struct thread_info *thread;
+       void *stack_start;
+
+       /*  Create new init task for the CPU  */
+       idle = fork_idle(cpu);
+       if (IS_ERR(idle))
+               panic(KERN_ERR "fork_idle failed\n");
+
+       thread = (struct thread_info *)idle->stack;
+       thread->cpu = cpu;
+
+       /*  Boot to the head.  */
+       stack_start =  ((void *) thread) + THREAD_SIZE;
+       __vmstart(start_secondary, stack_start);
+
+       while (!cpu_isset(cpu, cpu_online_map))
+               barrier();
+
+       return 0;
+}
+
+void __init smp_cpus_done(unsigned int max_cpus)
+{
+}
+
+void __init smp_prepare_cpus(unsigned int max_cpus)
+{
+       int i;
+
+       /*
+        * should eventually have some sort of machine
+        * descriptor that has this stuff
+        */
+
+       /*  Right now, let's just fake it. */
+       for (i = 0; i < max_cpus; i++)
+               cpu_set(i, cpu_present_map);
+
+       /*  Also need to register the interrupts for IPI  */
+       if (max_cpus > 1)
+               setup_irq(BASE_IPI_IRQ, &ipi_intdesc);
+}
+
+void smp_send_reschedule(int cpu)
+{
+       send_ipi(cpumask_of(cpu), IPI_RESCHEDULE);
+}
+
+void smp_send_stop(void)
+{
+       struct cpumask targets;
+       cpumask_copy(&targets, cpu_online_mask);
+       cpumask_clear_cpu(smp_processor_id(), &targets);
+       send_ipi(&targets, IPI_CPU_STOP);
+}
+
+void arch_send_call_function_single_ipi(int cpu)
+{
+       send_ipi(cpumask_of(cpu), IPI_CALL_FUNC_SINGLE);
+}
+
+void arch_send_call_function_ipi_mask(const struct cpumask *mask)
+{
+       send_ipi(mask, IPI_CALL_FUNC);
+}
+
+int setup_profiling_timer(unsigned int multiplier)
+{
+       return -EINVAL;
+}
+
+void smp_start_cpus(void)
+{
+       int i;
+
+       for (i = 0; i < NR_CPUS; i++)
+               cpu_set(i, cpu_possible_map);
+}
diff --git a/arch/hexagon/kernel/topology.c b/arch/hexagon/kernel/topology.c
new file mode 100644 (file)
index 0000000..ba44751
--- /dev/null
@@ -0,0 +1,52 @@
+/*
+ * CPU topology for Hexagon
+ *
+ * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#include <linux/cpu.h>
+#include <linux/cpumask.h>
+#include <linux/init.h>
+#include <linux/node.h>
+#include <linux/nodemask.h>
+#include <linux/percpu.h>
+
+/*  Swiped from MIPS.  */
+
+static DEFINE_PER_CPU(struct cpu, cpu_devices);
+
+static int __init topology_init(void)
+{
+       int i, ret;
+
+       for_each_present_cpu(i) {
+
+               /*
+                * register_cpu takes a per_cpu pointer and
+                * just points it at another per_cpu struct...
+                */
+
+               ret = register_cpu(&per_cpu(cpu_devices, i), i);
+               if (ret)
+                       printk(KERN_WARNING "topology_init: register_cpu %d "
+                              "failed (%d)\n", i, ret);
+       }
+
+       return 0;
+}
+
+subsys_initcall(topology_init);