MIPS: Add code for new system 'paravirt'

For para-virtualized guests running under KVM or other equivalent
hypervisor.

Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: Andreas Herrmann <andreas.herrmann@caviumnetworks.com>
Cc: linux-mips@linux-mips.org
Cc: James Hogan <james.hogan@imgtec.com>
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/7004/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

6 files changed, 645 insertions, 0 deletions

diff --git a/arch/mips/paravirt/Makefile b/arch/mips/paravirt/Makefile
new file mode 100644
index 000000000000..5023af733a35
--- /dev/null
+++ b/arch/mips/paravirt/Makefile
@@ -0,0 +1,14 @@
+#
+# Makefile for MIPS para-virtualized specific kernel interface routines
+# under Linux.
+#
+# 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.
+#
+# Copyright (C) 2013 Cavium, Inc.
+#
+
+obj-y := setup.o serial.o paravirt-irq.o
+
+obj-$(CONFIG_SMP)		+= paravirt-smp.o
diff --git a/arch/mips/paravirt/Platform b/arch/mips/paravirt/Platform
new file mode 100644
index 000000000000..7e76ef25ea17
--- /dev/null
+++ b/arch/mips/paravirt/Platform
@@ -0,0 +1,8 @@
+#
+# Generic para-virtualized guest.
+#
+platform-$(CONFIG_MIPS_PARAVIRT)	+= paravirt/
+cflags-$(CONFIG_MIPS_PARAVIRT)		+=				\
+		-I$(srctree)/arch/mips/include/asm/mach-paravirt
+
+load-$(CONFIG_MIPS_PARAVIRT)	= 0xffffffff80010000
diff --git a/arch/mips/paravirt/paravirt-irq.c b/arch/mips/paravirt/paravirt-irq.c
new file mode 100644
index 000000000000..8987b06c9de9
--- /dev/null
+++ b/arch/mips/paravirt/paravirt-irq.c
@@ -0,0 +1,368 @@
+/*
+ * 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.
+ *
+ * Copyright (C) 2013 Cavium, Inc.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/cpumask.h>
+#include <linux/kernel.h>
+#include <linux/mutex.h>
+
+#include <asm/io.h>
+
+#define MBOX_BITS_PER_CPU 2
+
+static int cpunum_for_cpu(int cpu)
+{
+#ifdef CONFIG_SMP
+	return cpu_logical_map(cpu);
+#else
+	return get_ebase_cpunum();
+#endif
+}
+
+struct core_chip_data {
+	struct mutex core_irq_mutex;
+	bool current_en;
+	bool desired_en;
+	u8 bit;
+};
+
+static struct core_chip_data irq_core_chip_data[8];
+
+static void irq_core_ack(struct irq_data *data)
+{
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+	unsigned int bit = cd->bit;
+
+	/*
+	 * We don't need to disable IRQs to make these atomic since
+	 * they are already disabled earlier in the low level
+	 * interrupt code.
+	 */
+	clear_c0_status(0x100 << bit);
+	/* The two user interrupts must be cleared manually. */
+	if (bit < 2)
+		clear_c0_cause(0x100 << bit);
+}
+
+static void irq_core_eoi(struct irq_data *data)
+{
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+
+	/*
+	 * We don't need to disable IRQs to make these atomic since
+	 * they are already disabled earlier in the low level
+	 * interrupt code.
+	 */
+	set_c0_status(0x100 << cd->bit);
+}
+
+static void irq_core_set_enable_local(void *arg)
+{
+	struct irq_data *data = arg;
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+	unsigned int mask = 0x100 << cd->bit;
+
+	/*
+	 * Interrupts are already disabled, so these are atomic.
+	 */
+	if (cd->desired_en)
+		set_c0_status(mask);
+	else
+		clear_c0_status(mask);
+
+}
+
+static void irq_core_disable(struct irq_data *data)
+{
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+	cd->desired_en = false;
+}
+
+static void irq_core_enable(struct irq_data *data)
+{
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+	cd->desired_en = true;
+}
+
+static void irq_core_bus_lock(struct irq_data *data)
+{
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+
+	mutex_lock(&cd->core_irq_mutex);
+}
+
+static void irq_core_bus_sync_unlock(struct irq_data *data)
+{
+	struct core_chip_data *cd = irq_data_get_irq_chip_data(data);
+
+	if (cd->desired_en != cd->current_en) {
+		on_each_cpu(irq_core_set_enable_local, data, 1);
+		cd->current_en = cd->desired_en;
+	}
+
+	mutex_unlock(&cd->core_irq_mutex);
+}
+
+static struct irq_chip irq_chip_core = {
+	.name = "Core",
+	.irq_enable = irq_core_enable,
+	.irq_disable = irq_core_disable,
+	.irq_ack = irq_core_ack,
+	.irq_eoi = irq_core_eoi,
+	.irq_bus_lock = irq_core_bus_lock,
+	.irq_bus_sync_unlock = irq_core_bus_sync_unlock,
+
+	.irq_cpu_online = irq_core_eoi,
+	.irq_cpu_offline = irq_core_ack,
+	.flags = IRQCHIP_ONOFFLINE_ENABLED,
+};
+
+static void __init irq_init_core(void)
+{
+	int i;
+	int irq;
+	struct core_chip_data *cd;
+
+	/* Start with a clean slate */
+	clear_c0_status(ST0_IM);
+	clear_c0_cause(CAUSEF_IP0 | CAUSEF_IP1);
+
+	for (i = 0; i < ARRAY_SIZE(irq_core_chip_data); i++) {
+		cd = irq_core_chip_data + i;
+		cd->current_en = false;
+		cd->desired_en = false;
+		cd->bit = i;
+		mutex_init(&cd->core_irq_mutex);
+
+		irq = MIPS_CPU_IRQ_BASE + i;
+
+		switch (i) {
+		case 0: /* SW0 */
+		case 1: /* SW1 */
+		case 5: /* IP5 */
+		case 6: /* IP6 */
+		case 7: /* IP7 */
+			irq_set_chip_data(irq, cd);
+			irq_set_chip_and_handler(irq, &irq_chip_core,
+						 handle_percpu_irq);
+			break;
+		default:
+			break;
+		}
+	}
+}
+
+static void __iomem *mips_irq_chip;
+#define MIPS_IRQ_CHIP_NUM_BITS 0
+#define MIPS_IRQ_CHIP_REGS 8
+
+static int mips_irq_cpu_stride;
+static int mips_irq_chip_reg_raw;
+static int mips_irq_chip_reg_src;
+static int mips_irq_chip_reg_en;
+static int mips_irq_chip_reg_raw_w1s;
+static int mips_irq_chip_reg_raw_w1c;
+static int mips_irq_chip_reg_en_w1s;
+static int mips_irq_chip_reg_en_w1c;
+
+static void irq_pci_enable(struct irq_data *data)
+{
+	u32 mask = 1u << data->irq;
+
+	__raw_writel(mask, mips_irq_chip + mips_irq_chip_reg_en_w1s);
+}
+
+static void irq_pci_disable(struct irq_data *data)
+{
+	u32 mask = 1u << data->irq;
+
+	__raw_writel(mask, mips_irq_chip + mips_irq_chip_reg_en_w1c);
+}
+
+static void irq_pci_ack(struct irq_data *data)
+{
+}
+
+static void irq_pci_mask(struct irq_data *data)
+{
+	u32 mask = 1u << data->irq;
+
+	__raw_writel(mask, mips_irq_chip + mips_irq_chip_reg_en_w1c);
+}
+
+static void irq_pci_unmask(struct irq_data *data)
+{
+	u32 mask = 1u << data->irq;
+
+	__raw_writel(mask, mips_irq_chip + mips_irq_chip_reg_en_w1s);
+}
+
+static struct irq_chip irq_chip_pci = {
+	.name = "PCI",
+	.irq_enable = irq_pci_enable,
+	.irq_disable = irq_pci_disable,
+	.irq_ack = irq_pci_ack,
+	.irq_mask = irq_pci_mask,
+	.irq_unmask = irq_pci_unmask,
+};
+
+static void irq_mbox_all(struct irq_data *data,  void __iomem *base)
+{
+	int cpu;
+	unsigned int mbox = data->irq - MIPS_IRQ_MBOX0;
+	u32 mask;
+
+	WARN_ON(mbox >= MBOX_BITS_PER_CPU);
+
+	for_each_online_cpu(cpu) {
+		unsigned int cpuid = cpunum_for_cpu(cpu);
+		mask = 1 << (cpuid * MBOX_BITS_PER_CPU + mbox);
+		__raw_writel(mask, base + (cpuid * mips_irq_cpu_stride));
+	}
+}
+
+static void irq_mbox_enable(struct irq_data *data)
+{
+	irq_mbox_all(data, mips_irq_chip + mips_irq_chip_reg_en_w1s + sizeof(u32));
+}
+
+static void irq_mbox_disable(struct irq_data *data)
+{
+	irq_mbox_all(data, mips_irq_chip + mips_irq_chip_reg_en_w1c + sizeof(u32));
+}
+
+static void irq_mbox_ack(struct irq_data *data)
+{
+	u32 mask;
+	unsigned int mbox = data->irq - MIPS_IRQ_MBOX0;
+
+	WARN_ON(mbox >= MBOX_BITS_PER_CPU);
+
+	mask = 1 << (get_ebase_cpunum() * MBOX_BITS_PER_CPU + mbox);
+	__raw_writel(mask, mips_irq_chip + mips_irq_chip_reg_raw_w1c + sizeof(u32));
+}
+
+void irq_mbox_ipi(int cpu, unsigned int actions)
+{
+	unsigned int cpuid = cpunum_for_cpu(cpu);
+	u32 mask;
+
+	WARN_ON(actions >= (1 << MBOX_BITS_PER_CPU));
+
+	mask = actions << (cpuid * MBOX_BITS_PER_CPU);
+	__raw_writel(mask, mips_irq_chip + mips_irq_chip_reg_raw_w1s + sizeof(u32));
+}
+
+static void irq_mbox_cpu_onoffline(struct irq_data *data,  void __iomem *base)
+{
+	unsigned int mbox = data->irq - MIPS_IRQ_MBOX0;
+	unsigned int cpuid = get_ebase_cpunum();
+	u32 mask;
+
+	WARN_ON(mbox >= MBOX_BITS_PER_CPU);
+
+	mask = 1 << (cpuid * MBOX_BITS_PER_CPU + mbox);
+	__raw_writel(mask, base + (cpuid * mips_irq_cpu_stride));
+
+}
+
+static void irq_mbox_cpu_online(struct irq_data *data)
+{
+	irq_mbox_cpu_onoffline(data, mips_irq_chip + mips_irq_chip_reg_en_w1s + sizeof(u32));
+}
+
+static void irq_mbox_cpu_offline(struct irq_data *data)
+{
+	irq_mbox_cpu_onoffline(data, mips_irq_chip + mips_irq_chip_reg_en_w1c + sizeof(u32));
+}
+
+static struct irq_chip irq_chip_mbox = {
+	.name = "MBOX",
+	.irq_enable = irq_mbox_enable,
+	.irq_disable = irq_mbox_disable,
+	.irq_ack = irq_mbox_ack,
+	.irq_cpu_online = irq_mbox_cpu_online,
+	.irq_cpu_offline = irq_mbox_cpu_offline,
+	.flags = IRQCHIP_ONOFFLINE_ENABLED,
+};
+
+static void __init irq_pci_init(void)
+{
+	int i, stride;
+	u32 num_bits;
+
+	mips_irq_chip = ioremap(0x1e010000, 4096);
+
+	num_bits = __raw_readl(mips_irq_chip + MIPS_IRQ_CHIP_NUM_BITS);
+	stride = 8 * (1 + ((num_bits - 1) / 64));
+
+
+	pr_notice("mips_irq_chip: %u bits, reg stride: %d\n", num_bits, stride);
+	mips_irq_chip_reg_raw		= MIPS_IRQ_CHIP_REGS + 0 * stride;
+	mips_irq_chip_reg_raw_w1s	= MIPS_IRQ_CHIP_REGS + 1 * stride;
+	mips_irq_chip_reg_raw_w1c	= MIPS_IRQ_CHIP_REGS + 2 * stride;
+	mips_irq_chip_reg_src		= MIPS_IRQ_CHIP_REGS + 3 * stride;
+	mips_irq_chip_reg_en		= MIPS_IRQ_CHIP_REGS + 4 * stride;
+	mips_irq_chip_reg_en_w1s	= MIPS_IRQ_CHIP_REGS + 5 * stride;
+	mips_irq_chip_reg_en_w1c	= MIPS_IRQ_CHIP_REGS + 6 * stride;
+	mips_irq_cpu_stride		= stride * 4;
+
+	for (i = 0; i < 4; i++)
+		irq_set_chip_and_handler(i + MIPS_IRQ_PCIA, &irq_chip_pci, handle_level_irq);
+
+	for (i = 0; i < 2; i++)
+		irq_set_chip_and_handler(i + MIPS_IRQ_MBOX0, &irq_chip_mbox, handle_percpu_irq);
+
+
+	set_c0_status(STATUSF_IP2);
+}
+
+static void irq_pci_dispatch(void)
+{
+	unsigned int cpuid = get_ebase_cpunum();
+	u32 en;
+
+	en = __raw_readl(mips_irq_chip + mips_irq_chip_reg_src +
+			(cpuid * mips_irq_cpu_stride));
+
+	if (!en) {
+		en = __raw_readl(mips_irq_chip + mips_irq_chip_reg_src + (cpuid * mips_irq_cpu_stride) + sizeof(u32));
+		en = (en >> (2 * cpuid)) & 3;
+
+		if (!en)
+			spurious_interrupt();
+		else
+			do_IRQ(__ffs(en) + MIPS_IRQ_MBOX0);	/* MBOX type */
+	} else {
+		do_IRQ(__ffs(en));
+	}
+}
+
+
+void __init arch_init_irq(void)
+{
+	irq_init_core();
+	irq_pci_init();
+}
+
+asmlinkage void plat_irq_dispatch(void)
+{
+	unsigned int pending = read_c0_cause() & read_c0_status() & ST0_IM;
+	int ip;
+
+	if (unlikely(!pending)) {
+		spurious_interrupt();
+		return;
+	}
+
+	ip = ffs(pending) - 1 - STATUSB_IP0;
+	if (ip == 2)
+		irq_pci_dispatch();
+	else
+		do_IRQ(MIPS_CPU_IRQ_BASE + ip);
+}
diff --git a/arch/mips/paravirt/paravirt-smp.c b/arch/mips/paravirt/paravirt-smp.c
new file mode 100644
index 000000000000..73a123e08c28
--- /dev/null
+++ b/arch/mips/paravirt/paravirt-smp.c
@@ -0,0 +1,148 @@
+/*
+ * 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.
+ *
+ * Copyright (C) 2013 Cavium, Inc.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/cpumask.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+
+#include <asm/mipsregs.h>
+#include <asm/setup.h>
+#include <asm/time.h>
+#include <asm/smp.h>
+
+/*
+ * Writing the sp releases the CPU, so writes must be ordered, gp
+ * first, then sp.
+ */
+unsigned long paravirt_smp_sp[NR_CPUS];
+unsigned long paravirt_smp_gp[NR_CPUS];
+
+static int numcpus = 1;
+
+static int __init set_numcpus(char *str)
+{
+	int newval;
+
+	if (get_option(&str, &newval)) {
+		if (newval < 1 || newval >= NR_CPUS)
+			goto bad;
+		numcpus = newval;
+		return 0;
+	}
+bad:
+	return -EINVAL;
+}
+early_param("numcpus", set_numcpus);
+
+
+static void paravirt_smp_setup(void)
+{
+	int id;
+	unsigned int cpunum = get_ebase_cpunum();
+
+	if (WARN_ON(cpunum >= NR_CPUS))
+		return;
+
+	/* The present CPUs are initially just the boot cpu (CPU 0). */
+	for (id = 0; id < NR_CPUS; id++) {
+		set_cpu_possible(id, id == 0);
+		set_cpu_present(id, id == 0);
+	}
+	__cpu_number_map[cpunum] = 0;
+	__cpu_logical_map[0] = cpunum;
+
+	for (id = 0; id < numcpus; id++) {
+		set_cpu_possible(id, true);
+		set_cpu_present(id, true);
+		__cpu_number_map[id] = id;
+		__cpu_logical_map[id] = id;
+	}
+}
+
+void irq_mbox_ipi(int cpu, unsigned int actions);
+static void paravirt_send_ipi_single(int cpu, unsigned int action)
+{
+	irq_mbox_ipi(cpu, action);
+}
+
+static void paravirt_send_ipi_mask(const struct cpumask *mask, unsigned int action)
+{
+	unsigned int cpu;
+
+	for_each_cpu_mask(cpu, *mask)
+		paravirt_send_ipi_single(cpu, action);
+}
+
+static void paravirt_init_secondary(void)
+{
+	unsigned int sr;
+
+	sr = set_c0_status(ST0_BEV);
+	write_c0_ebase((u32)ebase);
+
+	sr |= STATUSF_IP2; /* Interrupt controller on IP2 */
+	write_c0_status(sr);
+
+	irq_cpu_online();
+}
+
+static void paravirt_smp_finish(void)
+{
+	/* to generate the first CPU timer interrupt */
+	write_c0_compare(read_c0_count() + mips_hpt_frequency / HZ);
+	local_irq_enable();
+}
+
+static void paravirt_cpus_done(void)
+{
+}
+
+static void paravirt_boot_secondary(int cpu, struct task_struct *idle)
+{
+	paravirt_smp_gp[cpu] = (unsigned long)task_thread_info(idle);
+	smp_wmb();
+	paravirt_smp_sp[cpu] = __KSTK_TOS(idle);
+}
+
+static irqreturn_t paravirt_reched_interrupt(int irq, void *dev_id)
+{
+	scheduler_ipi();
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t paravirt_function_interrupt(int irq, void *dev_id)
+{
+	smp_call_function_interrupt();
+	return IRQ_HANDLED;
+}
+
+static void paravirt_prepare_cpus(unsigned int max_cpus)
+{
+	if (request_irq(MIPS_IRQ_MBOX0, paravirt_reched_interrupt,
+			IRQF_PERCPU | IRQF_NO_THREAD, "Scheduler",
+			paravirt_reched_interrupt)) {
+		panic("Cannot request_irq for SchedulerIPI");
+	}
+	if (request_irq(MIPS_IRQ_MBOX1, paravirt_function_interrupt,
+			IRQF_PERCPU | IRQF_NO_THREAD, "SMP-Call",
+			paravirt_function_interrupt)) {
+		panic("Cannot request_irq for SMP-Call");
+	}
+}
+
+struct plat_smp_ops paravirt_smp_ops = {
+	.send_ipi_single	= paravirt_send_ipi_single,
+	.send_ipi_mask		= paravirt_send_ipi_mask,
+	.init_secondary		= paravirt_init_secondary,
+	.smp_finish		= paravirt_smp_finish,
+	.cpus_done		= paravirt_cpus_done,
+	.boot_secondary		= paravirt_boot_secondary,
+	.smp_setup		= paravirt_smp_setup,
+	.prepare_cpus		= paravirt_prepare_cpus,
+};
diff --git a/arch/mips/paravirt/serial.c b/arch/mips/paravirt/serial.c
new file mode 100644
index 000000000000..02b665c02272
--- /dev/null
+++ b/arch/mips/paravirt/serial.c
@@ -0,0 +1,40 @@
+/*
+ * 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.
+ *
+ * Copyright (C) 2013 Cavium, Inc.
+ */
+
+#include <linux/kernel.h>
+#include <linux/virtio_console.h>
+#include <linux/kvm_para.h>
+
+/*
+ * Emit one character to the boot console.
+ */
+int prom_putchar(char c)
+{
+	kvm_hypercall3(KVM_HC_MIPS_CONSOLE_OUTPUT, 0 /*  port 0 */,
+		(unsigned long)&c, 1 /* len == 1 */);
+
+	return 1;
+}
+
+#ifdef CONFIG_VIRTIO_CONSOLE
+static int paravirt_put_chars(u32 vtermno, const char *buf, int count)
+{
+	kvm_hypercall3(KVM_HC_MIPS_CONSOLE_OUTPUT, vtermno,
+		(unsigned long)buf, count);
+
+	return count;
+}
+
+static int __init paravirt_cons_init(void)
+{
+	virtio_cons_early_init(paravirt_put_chars);
+	return 0;
+}
+core_initcall(paravirt_cons_init);
+
+#endif
diff --git a/arch/mips/paravirt/setup.c b/arch/mips/paravirt/setup.c
new file mode 100644
index 000000000000..cb8448b373a7
--- /dev/null
+++ b/arch/mips/paravirt/setup.c
@@ -0,0 +1,67 @@
+/*
+ * 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.
+ *
+ * Copyright (C) 2013 Cavium, Inc.
+ */
+
+#include <linux/kernel.h>
+#include <linux/kvm_para.h>
+
+#include <asm/reboot.h>
+#include <asm/bootinfo.h>
+#include <asm/smp-ops.h>
+#include <asm/time.h>
+
+extern struct plat_smp_ops paravirt_smp_ops;
+
+const char *get_system_type(void)
+{
+	return "MIPS Para-Virtualized Guest";
+}
+
+void __init plat_time_init(void)
+{
+	mips_hpt_frequency = kvm_hypercall0(KVM_HC_MIPS_GET_CLOCK_FREQ);
+
+	preset_lpj = mips_hpt_frequency / (2 * HZ);
+}
+
+static void pv_machine_halt(void)
+{
+	kvm_hypercall0(KVM_HC_MIPS_EXIT_VM);
+}
+
+/*
+ * Early entry point for arch setup
+ */
+void __init prom_init(void)
+{
+	int i;
+	int argc = fw_arg0;
+	char **argv = (char **)fw_arg1;
+
+#ifdef CONFIG_32BIT
+	set_io_port_base(KSEG1ADDR(0x1e000000));
+#else /* CONFIG_64BIT */
+	set_io_port_base(PHYS_TO_XKSEG_UNCACHED(0x1e000000));
+#endif
+
+	for (i = 0; i < argc; i++) {
+		strlcat(arcs_cmdline, argv[i], COMMAND_LINE_SIZE);
+		if (i < argc - 1)
+			strlcat(arcs_cmdline, " ", COMMAND_LINE_SIZE);
+	}
+	_machine_halt = pv_machine_halt;
+	register_smp_ops(&paravirt_smp_ops);
+}
+
+void __init plat_mem_setup(void)
+{
+	/* Do nothing, the "mem=???" parser handles our memory. */
+}
+
+void __init prom_free_prom_memory(void)
+{
+}