metag: SMP support

Add SMP support for metag. This allows Linux to take control of multiple
hardware threads on a single Meta core, treating them as separate Linux
CPUs.

Signed-off-by: James Hogan <james.hogan@imgtec.com>

arch/metag/include/asm/cachepart.h

@@ -0,0 +1,42 @@
+/*
+ * Meta cache partition manipulation.
+ *
+ * Copyright 2010 Imagination Technologies Ltd.
+ */
+
+#ifndef _METAG_CACHEPART_H_
+#define _METAG_CACHEPART_H_
+
+/**
+ * get_dcache_size() - Get size of data cache.
+ */
+unsigned int get_dcache_size(void);
+
+/**
+ * get_icache_size() - Get size of code cache.
+ */
+unsigned int get_icache_size(void);
+
+/**
+ * get_global_dcache_size() - Get the thread's global dcache.
+ *
+ * Returns the size of the current thread's global dcache partition.
+ */
+unsigned int get_global_dcache_size(void);
+
+/**
+ * get_global_icache_size() - Get the thread's global icache.
+ *
+ * Returns the size of the current thread's global icache partition.
+ */
+unsigned int get_global_icache_size(void);
+
+/**
+ * check_for_dache_aliasing() - Ensure that the bootloader has configured the
+ * dache and icache properly to avoid aliasing
+ * @thread_id: Hardware thread ID
+ *
+ */
+void check_for_cache_aliasing(int thread_id);
+
+#endif

arch/metag/include/asm/core_reg.h

@@ -0,0 +1,35 @@
+#ifndef __ASM_METAG_CORE_REG_H_
+#define __ASM_METAG_CORE_REG_H_
+
+#include <asm/metag_regs.h>
+
+extern void core_reg_write(int unit, int reg, int thread, unsigned int val);
+extern unsigned int core_reg_read(int unit, int reg, int thread);
+
+/*
+ * These macros allow direct access from C to any register known to the
+ * assembler. Example candidates are TXTACTCYC, TXIDLECYC, and TXPRIVEXT.
+ */
+
+#define __core_reg_get(reg) ({						\
+	unsigned int __grvalue;						\
+	asm volatile("MOV	%0," #reg				\
+		     : "=r" (__grvalue));				\
+	__grvalue;							\
+})
+
+#define __core_reg_set(reg, value) do {					\
+	unsigned int __srvalue = (value);				\
+	asm volatile("MOV	" #reg ",%0"				\
+		     :							\
+		     : "r" (__srvalue));				\
+} while (0)
+
+#define __core_reg_swap(reg, value) do {				\
+	unsigned int __srvalue = (value);				\
+	asm volatile("SWAP	" #reg ",%0"				\
+		     : "+r" (__srvalue));				\
+	(value) = __srvalue;						\
+} while (0)
+
+#endif

arch/metag/include/asm/smp.h

@@ -0,0 +1,29 @@
+#ifndef __ASM_SMP_H
+#define __ASM_SMP_H
+
+#include <linux/cpumask.h>
+
+#define raw_smp_processor_id() (current_thread_info()->cpu)
+
+enum ipi_msg_type {
+	IPI_CALL_FUNC,
+	IPI_CALL_FUNC_SINGLE,
+	IPI_RESCHEDULE,
+};
+
+extern void arch_send_call_function_single_ipi(int cpu);
+extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
+#define arch_send_call_function_ipi_mask arch_send_call_function_ipi_mask
+
+asmlinkage void secondary_start_kernel(void);
+
+extern void secondary_startup(void);
+
+#ifdef CONFIG_HOTPLUG_CPU
+extern void __cpu_die(unsigned int cpu);
+extern int __cpu_disable(void);
+extern void cpu_die(void);
+#endif
+
+extern void smp_init_cpus(void);
+#endif /* __ASM_SMP_H */

arch/metag/include/asm/topology.h

@@ -0,0 +1,53 @@
+#ifndef _ASM_METAG_TOPOLOGY_H
+#define _ASM_METAG_TOPOLOGY_H
+
+#ifdef CONFIG_NUMA
+
+/* sched_domains SD_NODE_INIT for Meta machines */
+#define SD_NODE_INIT (struct sched_domain) {		\
+	.parent			= NULL,			\
+	.child			= NULL,			\
+	.groups			= NULL,			\
+	.min_interval		= 8,			\
+	.max_interval		= 32,			\
+	.busy_factor		= 32,			\
+	.imbalance_pct		= 125,			\
+	.cache_nice_tries	= 2,			\
+	.busy_idx		= 3,			\
+	.idle_idx		= 2,			\
+	.newidle_idx		= 0,			\
+	.wake_idx		= 0,			\
+	.forkexec_idx		= 0,			\
+	.flags			= SD_LOAD_BALANCE	\
+				| SD_BALANCE_FORK	\
+				| SD_BALANCE_EXEC	\
+				| SD_BALANCE_NEWIDLE	\
+				| SD_SERIALIZE,		\
+	.last_balance		= jiffies,		\
+	.balance_interval	= 1,			\
+	.nr_balance_failed	= 0,			\
+}
+
+#define cpu_to_node(cpu)	((void)(cpu), 0)
+#define parent_node(node)	((void)(node), 0)
+
+#define cpumask_of_node(node)	((void)node, cpu_online_mask)
+
+#define pcibus_to_node(bus)	((void)(bus), -1)
+#define cpumask_of_pcibus(bus)	(pcibus_to_node(bus) == -1 ? \
+					cpu_all_mask : \
+					cpumask_of_node(pcibus_to_node(bus)))
+
+#endif
+
+#define mc_capable()    (1)
+
+const struct cpumask *cpu_coregroup_mask(unsigned int cpu);
+
+extern cpumask_t cpu_core_map[NR_CPUS];
+
+#define topology_core_cpumask(cpu)	(&cpu_core_map[cpu])
+
+#include <asm-generic/topology.h>
+
+#endif /* _ASM_METAG_TOPOLOGY_H */

arch/metag/kernel/cachepart.c

@@ -0,0 +1,124 @@
+/*
+ * Meta cache partition manipulation.
+ *
+ * Copyright 2010 Imagination Technologies Ltd.
+ */
+
+#include <linux/kernel.h>
+#include <linux/io.h>
+#include <linux/errno.h>
+#include <asm/processor.h>
+#include <asm/cachepart.h>
+#include <asm/metag_isa.h>
+#include <asm/metag_mem.h>
+
+#define SYSC_DCPART(n)	(SYSC_DCPART0 + SYSC_xCPARTn_STRIDE * (n))
+#define SYSC_ICPART(n)	(SYSC_ICPART0 + SYSC_xCPARTn_STRIDE * (n))
+
+#define CACHE_ASSOCIATIVITY 4 /* 4 way set-assosiative */
+#define ICACHE 0
+#define DCACHE 1
+
+/* The CORE_CONFIG2 register is not available on Meta 1 */
+#ifdef CONFIG_METAG_META21
+unsigned int get_dcache_size(void)
+{
+	unsigned int config2 = metag_in32(METAC_CORE_CONFIG2);
+	return 0x1000 << ((config2 & METAC_CORECFG2_DCSZ_BITS)
+				>> METAC_CORECFG2_DCSZ_S);
+}
+
+unsigned int get_icache_size(void)
+{
+	unsigned int config2 = metag_in32(METAC_CORE_CONFIG2);
+	return 0x1000 << ((config2 & METAC_CORE_C2ICSZ_BITS)
+				>> METAC_CORE_C2ICSZ_S);
+}
+
+unsigned int get_global_dcache_size(void)
+{
+	unsigned int cpart = metag_in32(SYSC_DCPART(hard_processor_id()));
+	unsigned int temp = cpart & SYSC_xCPARTG_AND_BITS;
+	return (get_dcache_size() * ((temp >> SYSC_xCPARTG_AND_S) + 1)) >> 4;
+}
+
+unsigned int get_global_icache_size(void)
+{
+	unsigned int cpart = metag_in32(SYSC_ICPART(hard_processor_id()));
+	unsigned int temp = cpart & SYSC_xCPARTG_AND_BITS;
+	return (get_icache_size() * ((temp >> SYSC_xCPARTG_AND_S) + 1)) >> 4;
+}
+
+static unsigned int get_thread_cache_size(unsigned int cache, int thread_id)
+{
+	unsigned int cache_size;
+	unsigned int t_cache_part;
+	unsigned int isEnabled;
+	unsigned int offset = 0;
+	isEnabled = (cache == DCACHE ? metag_in32(MMCU_DCACHE_CTRL_ADDR) & 0x1 :
+		metag_in32(MMCU_ICACHE_CTRL_ADDR) & 0x1);
+	if (!isEnabled)
+		return 0;
+#if PAGE_OFFSET >= LINGLOBAL_BASE
+	/* Checking for global cache */
+	cache_size = (cache == DCACHE ? get_global_dache_size() :
+		get_global_icache_size());
+	offset = 8;
+#else
+	cache_size = (cache == DCACHE ? get_dcache_size() :
+		get_icache_size());
+#endif
+	t_cache_part = (cache == DCACHE ?
+		(metag_in32(SYSC_DCPART(thread_id)) >> offset) & 0xF :
+		(metag_in32(SYSC_ICPART(thread_id)) >> offset) & 0xF);
+	switch (t_cache_part) {
+	case 0xF:
+		return cache_size;
+	case 0x7:
+		return cache_size / 2;
+	case 0x3:
+		return cache_size / 4;
+	case 0x1:
+		return cache_size / 8;
+	case 0:
+		return cache_size / 16;
+	}
+	return -1;
+}
+
+void check_for_cache_aliasing(int thread_id)
+{
+	unsigned int thread_cache_size;
+	unsigned int cache_type;
+	for (cache_type = ICACHE; cache_type <= DCACHE; cache_type++) {
+		thread_cache_size =
+				get_thread_cache_size(cache_type, thread_id);
+		if (thread_cache_size < 0)
+			pr_emerg("Can't read %s cache size", \
+				 cache_type ? "DCACHE" : "ICACHE");
+		else if (thread_cache_size == 0)
+			/* Cache is off. No need to check for aliasing */
+			continue;
+		if (thread_cache_size / CACHE_ASSOCIATIVITY > PAGE_SIZE) {
+			pr_emerg("Cache aliasing detected in %s on Thread %d",
+				 cache_type ? "DCACHE" : "ICACHE", thread_id);
+			pr_warn("Total %s size: %u bytes",
+				cache_type ? "DCACHE" : "ICACHE ",
+				cache_type ? get_dcache_size()
+				: get_icache_size());
+			pr_warn("Thread %s size: %d bytes",
+				cache_type ? "CACHE" : "ICACHE",
+				thread_cache_size);
+			pr_warn("Page Size: %lu bytes", PAGE_SIZE);
+		}
+	}
+}
+
+#else
+
+void check_for_cache_aliasing(int thread_id)
+{
+	return;
+}
+
+#endif

arch/metag/kernel/core_reg.c

@@ -0,0 +1,117 @@
+/*
+ *  Support for reading and writing Meta core internal registers.
+ *
+ *  Copyright (C) 2011 Imagination Technologies Ltd.
+ *
+ */
+
+#include <linux/delay.h>
+#include <linux/export.h>
+
+#include <asm/core_reg.h>
+#include <asm/global_lock.h>
+#include <asm/hwthread.h>
+#include <asm/io.h>
+#include <asm/metag_mem.h>
+#include <asm/metag_regs.h>
+
+#define UNIT_BIT_MASK		TXUXXRXRQ_UXX_BITS
+#define REG_BIT_MASK		TXUXXRXRQ_RX_BITS
+#define THREAD_BIT_MASK		TXUXXRXRQ_TX_BITS
+
+#define UNIT_SHIFTS		TXUXXRXRQ_UXX_S
+#define REG_SHIFTS		TXUXXRXRQ_RX_S
+#define THREAD_SHIFTS		TXUXXRXRQ_TX_S
+
+#define UNIT_VAL(x)		(((x) << UNIT_SHIFTS) & UNIT_BIT_MASK)
+#define REG_VAL(x)		(((x) << REG_SHIFTS) & REG_BIT_MASK)
+#define THREAD_VAL(x)		(((x) << THREAD_SHIFTS) & THREAD_BIT_MASK)
+
+/*
+ * core_reg_write() - modify the content of a register in a core unit.
+ * @unit:	The unit to be modified.
+ * @reg:	Register number within the unit.
+ * @thread:	The thread we want to access.
+ * @val:	The new value to write.
+ *
+ * Check asm/metag_regs.h for a list/defines of supported units (ie: TXUPC_ID,
+ * TXUTR_ID, etc), and regnums within the units (ie: TXMASKI_REGNUM,
+ * TXPOLLI_REGNUM, etc).
+ */
+void core_reg_write(int unit, int reg, int thread, unsigned int val)
+{
+	unsigned long flags;
+
+	/* TXUCT_ID has its own memory mapped registers */
+	if (unit == TXUCT_ID) {
+		void __iomem *cu_reg = __CU_addr(thread, reg);
+		metag_out32(val, cu_reg);
+		return;
+	}
+
+	__global_lock2(flags);
+
+	/* wait for ready */
+	while (!(metag_in32(TXUXXRXRQ) & TXUXXRXRQ_DREADY_BIT))
+		udelay(10);
+
+	/* set the value to write */
+	metag_out32(val, TXUXXRXDT);
+
+	/* set the register to write */
+	val = UNIT_VAL(unit) | REG_VAL(reg) | THREAD_VAL(thread);
+	metag_out32(val, TXUXXRXRQ);
+
+	/* wait for finish */
+	while (!(metag_in32(TXUXXRXRQ) & TXUXXRXRQ_DREADY_BIT))
+		udelay(10);
+
+	__global_unlock2(flags);
+}
+EXPORT_SYMBOL(core_reg_write);
+
+/*
+ * core_reg_read() - read the content of a register in a core unit.
+ * @unit:	The unit to be modified.
+ * @reg:	Register number within the unit.
+ * @thread:	The thread we want to access.
+ *
+ * Check asm/metag_regs.h for a list/defines of supported units (ie: TXUPC_ID,
+ * TXUTR_ID, etc), and regnums within the units (ie: TXMASKI_REGNUM,
+ * TXPOLLI_REGNUM, etc).
+ */
+unsigned int core_reg_read(int unit, int reg, int thread)
+{
+	unsigned long flags;
+	unsigned int val;
+
+	/* TXUCT_ID has its own memory mapped registers */
+	if (unit == TXUCT_ID) {
+		void __iomem *cu_reg = __CU_addr(thread, reg);
+		val = metag_in32(cu_reg);
+		return val;
+	}
+
+	__global_lock2(flags);
+
+	/* wait for ready */
+	while (!(metag_in32(TXUXXRXRQ) & TXUXXRXRQ_DREADY_BIT))
+		udelay(10);
+
+	/* set the register to read */
+	val = (UNIT_VAL(unit) | REG_VAL(reg) | THREAD_VAL(thread) |
+							TXUXXRXRQ_RDnWR_BIT);
+	metag_out32(val, TXUXXRXRQ);
+
+	/* wait for finish */
+	while (!(metag_in32(TXUXXRXRQ) & TXUXXRXRQ_DREADY_BIT))
+		udelay(10);
+
+	/* read the register value */
+	val = metag_in32(TXUXXRXDT);
+
+	__global_unlock2(flags);
+
+	return val;
+}
+EXPORT_SYMBOL(core_reg_read);

arch/metag/kernel/head.S

@@ -43,3 +43,15 @@ __start:
 __exit:
 	XOR     TXENABLE,D0Re0,D0Re0
 	.size	__exit,.-__exit
+
+#ifdef CONFIG_SMP
+	.global _secondary_startup
+	.type _secondary_startup,function
+_secondary_startup:
+	MOVT	A0StP,#HI(_secondary_data_stack)
+	ADD	A0StP,A0StP,#LO(_secondary_data_stack)
+	GETD	A0StP,[A0StP]
+	ADD	A0StP,A0StP,#THREAD_INFO_SIZE
+	B	_secondary_start_kernel
+	.size	_secondary_startup,.-_secondary_startup
+#endif

arch/metag/kernel/smp.c

@@ -0,0 +1,575 @@
+/*
+ *  Copyright (C) 2009,2010,2011 Imagination Technologies Ltd.
+ *
+ *  Copyright (C) 2002 ARM Limited, 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 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/atomic.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/spinlock.h>
+#include <linux/sched.h>
+#include <linux/interrupt.h>
+#include <linux/cache.h>
+#include <linux/profile.h>
+#include <linux/errno.h>
+#include <linux/mm.h>
+#include <linux/err.h>
+#include <linux/cpu.h>
+#include <linux/smp.h>
+#include <linux/seq_file.h>
+#include <linux/irq.h>
+#include <linux/bootmem.h>
+
+#include <asm/cacheflush.h>
+#include <asm/cachepart.h>
+#include <asm/core_reg.h>
+#include <asm/cpu.h>
+#include <asm/mmu_context.h>
+#include <asm/pgtable.h>
+#include <asm/pgalloc.h>
+#include <asm/processor.h>
+#include <asm/setup.h>
+#include <asm/tlbflush.h>
+#include <asm/hwthread.h>
+#include <asm/traps.h>
+
+DECLARE_PER_CPU(PTBI, pTBI);
+
+void *secondary_data_stack;
+
+/*
+ * structures for inter-processor calls
+ * - A collection of single bit ipi messages.
+ */
+struct ipi_data {
+	spinlock_t lock;
+	unsigned long ipi_count;
+	unsigned long bits;
+};
+
+static DEFINE_PER_CPU(struct ipi_data, ipi_data) = {
+	.lock	= __SPIN_LOCK_UNLOCKED(ipi_data.lock),
+};
+
+static DEFINE_SPINLOCK(boot_lock);
+
+/*
+ * "thread" is assumed to be a valid Meta hardware thread ID.
+ */
+int __cpuinit boot_secondary(unsigned int thread, struct task_struct *idle)
+{
+	u32 val;
+
+	/*
+	 * set synchronisation state between this boot processor
+	 * and the secondary one
+	 */
+	spin_lock(&boot_lock);
+
+	core_reg_write(TXUPC_ID, 0, thread, (unsigned int)secondary_startup);
+	core_reg_write(TXUPC_ID, 1, thread, 0);
+
+	/*
+	 * Give the thread privilege (PSTAT) and clear potentially problematic
+	 * bits in the process (namely ISTAT, CBMarker, CBMarkerI, LSM_STEP).
+	 */
+	core_reg_write(TXUCT_ID, TXSTATUS_REGNUM, thread, TXSTATUS_PSTAT_BIT);
+
+	/* Clear the minim enable bit. */
+	val = core_reg_read(TXUCT_ID, TXPRIVEXT_REGNUM, thread);
+	core_reg_write(TXUCT_ID, TXPRIVEXT_REGNUM, thread, val & ~0x80);
+
+	/*
+	 * set the ThreadEnable bit (0x1) in the TXENABLE register
+	 * for the specified thread - off it goes!
+	 */
+	val = core_reg_read(TXUCT_ID, TXENABLE_REGNUM, thread);
+	core_reg_write(TXUCT_ID, TXENABLE_REGNUM, thread, val | 0x1);
+
+	/*
+	 * now the secondary core is starting up let it run its
+	 * calibrations, then wait for it to finish
+	 */
+	spin_unlock(&boot_lock);
+
+	return 0;
+}
+
+int __cpuinit __cpu_up(unsigned int cpu, struct task_struct *idle)
+{
+	unsigned int thread = cpu_2_hwthread_id[cpu];
+	int ret;
+
+	load_pgd(swapper_pg_dir, thread);
+
+	flush_tlb_all();
+
+	/*
+	 * Tell the secondary CPU where to find its idle thread's stack.
+	 */
+	secondary_data_stack = task_stack_page(idle);
+
+	wmb();
+
+	/*
+	 * Now bring the CPU into our world.
+	 */
+	ret = boot_secondary(thread, idle);
+	if (ret == 0) {
+		unsigned long timeout;
+
+		/*
+		 * CPU was successfully started, wait for it
+		 * to come online or time out.
+		 */
+		timeout = jiffies + HZ;
+		while (time_before(jiffies, timeout)) {
+			if (cpu_online(cpu))
+				break;
+
+			udelay(10);
+			barrier();
+		}
+
+		if (!cpu_online(cpu))
+			ret = -EIO;
+	}
+
+	secondary_data_stack = NULL;
+
+	if (ret) {
+		pr_crit("CPU%u: processor failed to boot\n", cpu);
+
+		/*
+		 * FIXME: We need to clean up the new idle thread. --rmk
+		 */
+	}
+
+	return ret;
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+static DECLARE_COMPLETION(cpu_killed);
+
+/*
+ * __cpu_disable runs on the processor to be shutdown.
+ */
+int __cpuexit __cpu_disable(void)
+{
+	unsigned int cpu = smp_processor_id();
+	struct task_struct *p;
+
+	/*
+	 * Take this CPU offline.  Once we clear this, we can't return,
+	 * and we must not schedule until we're ready to give up the cpu.
+	 */
+	set_cpu_online(cpu, false);
+
+	/*
+	 * OK - migrate IRQs away from this CPU
+	 */
+	migrate_irqs();
+
+	/*
+	 * Flush user cache and TLB mappings, and then remove this CPU
+	 * from the vm mask set of all processes.
+	 */
+	flush_cache_all();
+	local_flush_tlb_all();
+
+	read_lock(&tasklist_lock);
+	for_each_process(p) {
+		if (p->mm)
+			cpumask_clear_cpu(cpu, mm_cpumask(p->mm));
+	}
+	read_unlock(&tasklist_lock);
+
+	return 0;
+}
+
+/*
+ * called on the thread which is asking for a CPU to be shutdown -
+ * waits until shutdown has completed, or it is timed out.
+ */
+void __cpuexit __cpu_die(unsigned int cpu)
+{
+	if (!wait_for_completion_timeout(&cpu_killed, msecs_to_jiffies(1)))
+		pr_err("CPU%u: unable to kill\n", cpu);
+}
+
+/*
+ * Called from the idle thread for the CPU which has been shutdown.
+ *
+ * Note that we do not return from this function. If this cpu is
+ * brought online again it will need to run secondary_startup().
+ */
+void __cpuexit cpu_die(void)
+{
+	local_irq_disable();
+	idle_task_exit();
+
+	complete(&cpu_killed);
+
+	asm ("XOR	TXENABLE, D0Re0,D0Re0\n");
+}
+#endif /* CONFIG_HOTPLUG_CPU */
+
+/*
+ * Called by both boot and secondaries to move global data into
+ * per-processor storage.
+ */
+void __cpuinit smp_store_cpu_info(unsigned int cpuid)
+{
+	struct cpuinfo_metag *cpu_info = &per_cpu(cpu_data, cpuid);
+
+	cpu_info->loops_per_jiffy = loops_per_jiffy;
+}
+
+/*
+ * This is the secondary CPU boot entry.  We're using this CPUs
+ * idle thread stack and the global page tables.
+ */
+asmlinkage void secondary_start_kernel(void)
+{
+	struct mm_struct *mm = &init_mm;
+	unsigned int cpu = smp_processor_id();
+
+	/*
+	 * All kernel threads share the same mm context; grab a
+	 * reference and switch to it.
+	 */
+	atomic_inc(&mm->mm_users);
+	atomic_inc(&mm->mm_count);
+	current->active_mm = mm;
+	cpumask_set_cpu(cpu, mm_cpumask(mm));
+	enter_lazy_tlb(mm, current);
+	local_flush_tlb_all();
+
+	/*
+	 * TODO: Some day it might be useful for each Linux CPU to
+	 * have its own TBI structure. That would allow each Linux CPU
+	 * to run different interrupt handlers for the same IRQ
+	 * number.
+	 *
+	 * For now, simply copying the pointer to the boot CPU's TBI
+	 * structure is sufficient because we always want to run the
+	 * same interrupt handler whatever CPU takes the interrupt.
+	 */
+	per_cpu(pTBI, cpu) = __TBI(TBID_ISTAT_BIT);
+
+	if (!per_cpu(pTBI, cpu))
+		panic("No TBI found!");
+
+	per_cpu_trap_init(cpu);
+
+	preempt_disable();
+
+	setup_txprivext();
+
+	/*
+	 * Enable local interrupts.
+	 */
+	tbi_startup_interrupt(TBID_SIGNUM_TRT);
+	notify_cpu_starting(cpu);
+	local_irq_enable();
+
+	pr_info("CPU%u (thread %u): Booted secondary processor\n",
+		cpu, cpu_2_hwthread_id[cpu]);
+
+	calibrate_delay();
+	smp_store_cpu_info(cpu);
+
+	/*
+	 * OK, now it's safe to let the boot CPU continue
+	 */
+	set_cpu_online(cpu, true);
+
+	/*
+	 * Check for cache aliasing.
+	 * Preemption is disabled
+	 */
+	check_for_cache_aliasing(cpu);
+
+	/*
+	 * OK, it's off to the idle thread for us
+	 */
+	cpu_idle();
+}
+
+void __init smp_cpus_done(unsigned int max_cpus)
+{
+	int cpu;
+	unsigned long bogosum = 0;
+
+	for_each_online_cpu(cpu)
+		bogosum += per_cpu(cpu_data, cpu).loops_per_jiffy;
+
+	pr_info("SMP: Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
+		num_online_cpus(),
+		bogosum / (500000/HZ),
+		(bogosum / (5000/HZ)) % 100);
+}
+
+void __init smp_prepare_cpus(unsigned int max_cpus)
+{
+	unsigned int cpu = smp_processor_id();
+
+	init_new_context(current, &init_mm);
+	current_thread_info()->cpu = cpu;
+
+	smp_store_cpu_info(cpu);
+	init_cpu_present(cpu_possible_mask);
+}
+
+void __init smp_prepare_boot_cpu(void)
+{
+	unsigned int cpu = smp_processor_id();
+
+	per_cpu(pTBI, cpu) = __TBI(TBID_ISTAT_BIT);
+
+	if (!per_cpu(pTBI, cpu))
+		panic("No TBI found!");
+}
+
+static void smp_cross_call(cpumask_t callmap, enum ipi_msg_type msg);
+
+static void send_ipi_message(const struct cpumask *mask, enum ipi_msg_type msg)
+{
+	unsigned long flags;
+	unsigned int cpu;
+	cpumask_t map;
+
+	cpumask_clear(&map);
+	local_irq_save(flags);
+
+	for_each_cpu(cpu, mask) {
+		struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
+
+		spin_lock(&ipi->lock);
+
+		/*
+		 * KICK interrupts are queued in hardware so we'll get
+		 * multiple interrupts if we call smp_cross_call()
+		 * multiple times for one msg. The problem is that we
+		 * only have one bit for each message - we can't queue
+		 * them in software.
+		 *
+		 * The first time through ipi_handler() we'll clear
+		 * the msg bit, having done all the work. But when we
+		 * return we'll get _another_ interrupt (and another,
+		 * and another until we've handled all the queued
+		 * KICKs). Running ipi_handler() when there's no work
+		 * to do is bad because that's how kick handler
+		 * chaining detects who the KICK was intended for.
+		 * See arch/metag/kernel/kick.c for more details.
+		 *
+		 * So only add 'cpu' to 'map' if we haven't already
+		 * queued a KICK interrupt for 'msg'.
+		 */
+		if (!(ipi->bits & (1 << msg))) {
+			ipi->bits |= 1 << msg;
+			cpumask_set_cpu(cpu, &map);
+		}
+
+		spin_unlock(&ipi->lock);
+	}
+
+	/*
+	 * Call the platform specific cross-CPU call function.
+	 */
+	smp_cross_call(map, msg);
+
+	local_irq_restore(flags);
+}
+
+void arch_send_call_function_ipi_mask(const struct cpumask *mask)
+{
+	send_ipi_message(mask, IPI_CALL_FUNC);
+}
+
+void arch_send_call_function_single_ipi(int cpu)
+{
+	send_ipi_message(cpumask_of(cpu), IPI_CALL_FUNC_SINGLE);
+}
+
+void show_ipi_list(struct seq_file *p)
+{
+	unsigned int cpu;
+
+	seq_puts(p, "IPI:");
+
+	for_each_present_cpu(cpu)
+		seq_printf(p, " %10lu", per_cpu(ipi_data, cpu).ipi_count);
+
+	seq_putc(p, '\n');
+}
+
+static DEFINE_SPINLOCK(stop_lock);
+
+/*
+ * Main handler for inter-processor interrupts
+ *
+ * For Meta, the ipimask now only identifies a single
+ * category of IPI (Bit 1 IPIs have been replaced by a
+ * different mechanism):
+ *
+ *  Bit 0 - Inter-processor function call
+ */
+static int do_IPI(struct pt_regs *regs)
+{
+	unsigned int cpu = smp_processor_id();
+	struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
+	struct pt_regs *old_regs = set_irq_regs(regs);
+	unsigned long msgs, nextmsg;
+	int handled = 0;
+
+	ipi->ipi_count++;
+
+	spin_lock(&ipi->lock);
+	msgs = ipi->bits;
+	nextmsg = msgs & -msgs;
+	ipi->bits &= ~nextmsg;
+	spin_unlock(&ipi->lock);
+
+	if (nextmsg) {
+		handled = 1;
+
+		nextmsg = ffz(~nextmsg);
+		switch (nextmsg) {
+		case IPI_RESCHEDULE:
+			scheduler_ipi();
+			break;
+
+		case IPI_CALL_FUNC:
+			generic_smp_call_function_interrupt();
+			break;
+
+		case IPI_CALL_FUNC_SINGLE:
+			generic_smp_call_function_single_interrupt();
+			break;
+
+		default:
+			pr_crit("CPU%u: Unknown IPI message 0x%lx\n",
+				cpu, nextmsg);
+			break;
+		}
+	}
+
+	set_irq_regs(old_regs);
+
+	return handled;
+}
+
+void smp_send_reschedule(int cpu)
+{
+	send_ipi_message(cpumask_of(cpu), IPI_RESCHEDULE);
+}
+
+static void stop_this_cpu(void *data)
+{
+	unsigned int cpu = smp_processor_id();
+
+	if (system_state == SYSTEM_BOOTING ||
+	    system_state == SYSTEM_RUNNING) {
+		spin_lock(&stop_lock);
+		pr_crit("CPU%u: stopping\n", cpu);
+		dump_stack();
+		spin_unlock(&stop_lock);
+	}
+
+	set_cpu_online(cpu, false);
+
+	local_irq_disable();
+
+	hard_processor_halt(HALT_OK);
+}
+
+void smp_send_stop(void)
+{
+	smp_call_function(stop_this_cpu, NULL, 0);
+}
+
+/*
+ * not supported here
+ */
+int setup_profiling_timer(unsigned int multiplier)
+{
+	return -EINVAL;
+}
+
+/*
+ * We use KICKs for inter-processor interrupts.
+ *
+ * For every CPU in "callmap" the IPI data must already have been
+ * stored in that CPU's "ipi_data" member prior to calling this
+ * function.
+ */
+static void kick_raise_softirq(cpumask_t callmap, unsigned int irq)
+{
+	int cpu;
+
+	for_each_cpu(cpu, &callmap) {
+		unsigned int thread;
+
+		thread = cpu_2_hwthread_id[cpu];
+
+		BUG_ON(thread == BAD_HWTHREAD_ID);
+
+		metag_out32(1, T0KICKI + (thread * TnXKICK_STRIDE));
+	}
+}
+
+static TBIRES ipi_handler(TBIRES State, int SigNum, int Triggers,
+		   int Inst, PTBI pTBI, int *handled)
+{
+	*handled = do_IPI((struct pt_regs *)State.Sig.pCtx);
+
+	return State;
+}
+
+static struct kick_irq_handler ipi_irq = {
+	.func = ipi_handler,
+};
+
+static void smp_cross_call(cpumask_t callmap, enum ipi_msg_type msg)
+{
+	kick_raise_softirq(callmap, 1);
+}
+
+static inline unsigned int get_core_count(void)
+{
+	int i;
+	unsigned int ret = 0;
+
+	for (i = 0; i < CONFIG_NR_CPUS; i++) {
+		if (core_reg_read(TXUCT_ID, TXENABLE_REGNUM, i))
+			ret++;
+	}
+
+	return ret;
+}
+
+/*
+ * Initialise the CPU possible map early - this describes the CPUs
+ * which may be present or become present in the system.
+ */
+void __init smp_init_cpus(void)
+{
+	unsigned int i, ncores = get_core_count();
+
+	/* If no hwthread_map early param was set use default mapping */
+	for (i = 0; i < NR_CPUS; i++)
+		if (cpu_2_hwthread_id[i] == BAD_HWTHREAD_ID) {
+			cpu_2_hwthread_id[i] = i;
+			hwthread_id_2_cpu[i] = i;
+		}
+
+	for (i = 0; i < ncores; i++)
+		set_cpu_possible(i, true);
+
+	kick_register_func(&ipi_irq);
+}

arch/metag/kernel/topology.c

@@ -0,0 +1,77 @@
+/*
+ *  Copyright (C) 2007  Paul Mundt
+ *  Copyright (C) 2010  Imagination Technolohies Ltd.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+#include <linux/cpu.h>
+#include <linux/cpumask.h>
+#include <linux/init.h>
+#include <linux/percpu.h>
+#include <linux/node.h>
+#include <linux/nodemask.h>
+#include <linux/topology.h>
+
+#include <asm/cpu.h>
+
+DEFINE_PER_CPU(struct cpuinfo_metag, cpu_data);
+
+cpumask_t cpu_core_map[NR_CPUS];
+
+static cpumask_t cpu_coregroup_map(unsigned int cpu)
+{
+	return *cpu_possible_mask;
+}
+
+const struct cpumask *cpu_coregroup_mask(unsigned int cpu)
+{
+	return &cpu_core_map[cpu];
+}
+
+int arch_update_cpu_topology(void)
+{
+	unsigned int cpu;
+
+	for_each_possible_cpu(cpu)
+		cpu_core_map[cpu] = cpu_coregroup_map(cpu);
+
+	return 0;
+}
+
+static int __init topology_init(void)
+{
+	int i, ret;
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+	for_each_online_node(i)
+		register_one_node(i);
+#endif
+
+	for_each_present_cpu(i) {
+		struct cpuinfo_metag *cpuinfo = &per_cpu(cpu_data, i);
+#ifdef CONFIG_HOTPLUG_CPU
+		cpuinfo->cpu.hotpluggable = 1;
+#endif
+		ret = register_cpu(&cpuinfo->cpu, i);
+		if (unlikely(ret))
+			pr_warn("%s: register_cpu %d failed (%d)\n",
+				__func__, i, ret);
+	}
+
+#if defined(CONFIG_NUMA) && !defined(CONFIG_SMP)
+	/*
+	 * In the UP case, make sure the CPU association is still
+	 * registered under each node. Without this, sysfs fails
+	 * to make the connection between nodes other than node0
+	 * and cpu0.
+	 */
+	for_each_online_node(i)
+		if (i != numa_node_id())
+			register_cpu_under_node(raw_smp_processor_id(), i);
+#endif
+
+	return 0;
+}
+subsys_initcall(topology_init);