MIPS: Add Cavium OCTEON processor support files to arch/mips/cavium-octeon.

These are the rest of the new files needed to add OCTEON processor
support to the Linux kernel.  Other than Makefile and Kconfig which
should be obvious, we have:

csrc-octeon.c   -- Clock source driver for OCTEON.
dma-octeon.c    -- Helper functions for mapping DMA memory.
flash_setup.c   -- Register on-board flash with the MTD subsystem.
octeon-irq.c    -- OCTEON interrupt controller managment.
octeon-memcpy.S -- Optimized memcpy() implementation.
serial.c        -- Register 8250 platform driver and early console.
setup.c         -- Early architecture initialization.
smp.c           -- OCTEON SMP support.
octeon_switch.S -- Scheduler context switch for OCTEON.
c-octeon.c      -- OCTEON cache controller support.
cex-oct.S       -- OCTEON cache exception handler.

asm/mach-cavium-octeon/*.h -- Architecture include files.

Signed-off-by: Tomaso Paoletti <tpaoletti@caviumnetworks.com>
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

 create mode 100644 arch/mips/cavium-octeon/Kconfig
 create mode 100644 arch/mips/cavium-octeon/Makefile
 create mode 100644 arch/mips/cavium-octeon/csrc-octeon.c
 create mode 100644 arch/mips/cavium-octeon/dma-octeon.c
 create mode 100644 arch/mips/cavium-octeon/flash_setup.c
 create mode 100644 arch/mips/cavium-octeon/octeon-irq.c
 create mode 100644 arch/mips/cavium-octeon/octeon-memcpy.S
 create mode 100644 arch/mips/cavium-octeon/serial.c
 create mode 100644 arch/mips/cavium-octeon/setup.c
 create mode 100644 arch/mips/cavium-octeon/smp.c
 create mode 100644 arch/mips/include/asm/mach-cavium-octeon/cpu-feature-overrides.h
 create mode 100644 arch/mips/include/asm/mach-cavium-octeon/dma-coherence.h
 create mode 100644 arch/mips/include/asm/mach-cavium-octeon/irq.h
 create mode 100644 arch/mips/include/asm/mach-cavium-octeon/kernel-entry-init.h
 create mode 100644 arch/mips/include/asm/mach-cavium-octeon/war.h
 create mode 100644 arch/mips/include/asm/octeon/octeon.h
 create mode 100644 arch/mips/kernel/octeon_switch.S
 create mode 100644 arch/mips/mm/c-octeon.c
 create mode 100644 arch/mips/mm/cex-oct.S

19 files changed, 4243 insertions, 0 deletions

diff --git a/arch/mips/cavium-octeon/Kconfig b/arch/mips/cavium-octeon/Kconfig
new file mode 100644
index 00000000000..094c17e38e1
--- /dev/null
+++ b/arch/mips/cavium-octeon/Kconfig
@@ -0,0 +1,85 @@
+config CAVIUM_OCTEON_SPECIFIC_OPTIONS
+	bool "Enable Octeon specific options"
+	depends on CPU_CAVIUM_OCTEON
+	default "y"
+
+config CAVIUM_OCTEON_2ND_KERNEL
+	bool "Build the kernel to be used as a 2nd kernel on the same chip"
+	depends on CAVIUM_OCTEON_SPECIFIC_OPTIONS
+	default "n"
+	help
+	  This option configures this kernel to be linked at a different
+	  address and use the 2nd uart for output. This allows a kernel built
+	  with this option to be run at the same time as one built without this
+	  option.
+
+config CAVIUM_OCTEON_HW_FIX_UNALIGNED
+	bool "Enable hardware fixups of unaligned loads and stores"
+	depends on CAVIUM_OCTEON_SPECIFIC_OPTIONS
+	default "y"
+	help
+	  Configure the Octeon hardware to automatically fix unaligned loads
+	  and stores. Normally unaligned accesses are fixed using a kernel
+	  exception handler. This option enables the hardware automatic fixups,
+	  which requires only an extra 3 cycles. Disable this option if you
+	  are running code that relies on address exceptions on unaligned
+	  accesses.
+
+config CAVIUM_OCTEON_CVMSEG_SIZE
+	int "Number of L1 cache lines reserved for CVMSEG memory"
+	depends on CAVIUM_OCTEON_SPECIFIC_OPTIONS
+	range 0 54
+	default 1
+	help
+	  CVMSEG LM is a segment that accesses portions of the dcache as a
+	  local memory; the larger CVMSEG is, the smaller the cache is.
+	  This selects the size of CVMSEG LM, which is in cache blocks. The
+	  legally range is from zero to 54 cache blocks (i.e. CVMSEG LM is
+	  between zero and 6192 bytes).
+
+config CAVIUM_OCTEON_LOCK_L2
+	bool "Lock often used kernel code in the L2"
+	depends on CAVIUM_OCTEON_SPECIFIC_OPTIONS
+	default "y"
+	help
+	  Enable locking parts of the kernel into the L2 cache.
+
+config CAVIUM_OCTEON_LOCK_L2_TLB
+	bool "Lock the TLB handler in L2"
+	depends on CAVIUM_OCTEON_LOCK_L2
+	default "y"
+	help
+	  Lock the low level TLB fast path into L2.
+
+config CAVIUM_OCTEON_LOCK_L2_EXCEPTION
+	bool "Lock the exception handler in L2"
+	depends on CAVIUM_OCTEON_LOCK_L2
+	default "y"
+	help
+	  Lock the low level exception handler into L2.
+
+config CAVIUM_OCTEON_LOCK_L2_LOW_LEVEL_INTERRUPT
+	bool "Lock the interrupt handler in L2"
+	depends on CAVIUM_OCTEON_LOCK_L2
+	default "y"
+	help
+	  Lock the low level interrupt handler into L2.
+
+config CAVIUM_OCTEON_LOCK_L2_INTERRUPT
+	bool "Lock the 2nd level interrupt handler in L2"
+	depends on CAVIUM_OCTEON_LOCK_L2
+	default "y"
+	help
+	  Lock the 2nd level interrupt handler in L2.
+
+config CAVIUM_OCTEON_LOCK_L2_MEMCPY
+	bool "Lock memcpy() in L2"
+	depends on CAVIUM_OCTEON_LOCK_L2
+	default "y"
+	help
+	  Lock the kernel's implementation of memcpy() into L2.
+
+config ARCH_SPARSEMEM_ENABLE
+	def_bool y
+	select SPARSEMEM_STATIC
+	depends on CPU_CAVIUM_OCTEON
diff --git a/arch/mips/cavium-octeon/Makefile b/arch/mips/cavium-octeon/Makefile
new file mode 100644
index 00000000000..1c2a7faf588
--- /dev/null
+++ b/arch/mips/cavium-octeon/Makefile
@@ -0,0 +1,16 @@
+#
+# Makefile for the Cavium Octeon 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) 2005-2008 Cavium Networks
+#
+
+obj-y := setup.o serial.o octeon-irq.o csrc-octeon.o
+obj-y += dma-octeon.o flash_setup.o
+obj-y += octeon-memcpy.o
+
+obj-$(CONFIG_SMP)                     += smp.o
diff --git a/arch/mips/cavium-octeon/csrc-octeon.c b/arch/mips/cavium-octeon/csrc-octeon.c
new file mode 100644
index 00000000000..70fd92c3165
--- /dev/null
+++ b/arch/mips/cavium-octeon/csrc-octeon.c
@@ -0,0 +1,58 @@
+/*
+ * 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) 2007 by Ralf Baechle
+ */
+#include <linux/clocksource.h>
+#include <linux/init.h>
+
+#include <asm/time.h>
+
+#include <asm/octeon/octeon.h>
+#include <asm/octeon/cvmx-ipd-defs.h>
+
+/*
+ * Set the current core's cvmcount counter to the value of the
+ * IPD_CLK_COUNT.  We do this on all cores as they are brought
+ * on-line.  This allows for a read from a local cpu register to
+ * access a synchronized counter.
+ *
+ */
+void octeon_init_cvmcount(void)
+{
+	unsigned long flags;
+	unsigned loops = 2;
+
+	/* Clobber loops so GCC will not unroll the following while loop. */
+	asm("" : "+r" (loops));
+
+	local_irq_save(flags);
+	/*
+	 * Loop several times so we are executing from the cache,
+	 * which should give more deterministic timing.
+	 */
+	while (loops--)
+		write_c0_cvmcount(cvmx_read_csr(CVMX_IPD_CLK_COUNT));
+	local_irq_restore(flags);
+}
+
+static cycle_t octeon_cvmcount_read(void)
+{
+	return read_c0_cvmcount();
+}
+
+static struct clocksource clocksource_mips = {
+	.name		= "OCTEON_CVMCOUNT",
+	.read		= octeon_cvmcount_read,
+	.mask		= CLOCKSOURCE_MASK(64),
+	.flags		= CLOCK_SOURCE_IS_CONTINUOUS,
+};
+
+void __init plat_time_init(void)
+{
+	clocksource_mips.rating = 300;
+	clocksource_set_clock(&clocksource_mips, mips_hpt_frequency);
+	clocksource_register(&clocksource_mips);
+}
diff --git a/arch/mips/cavium-octeon/dma-octeon.c b/arch/mips/cavium-octeon/dma-octeon.c
new file mode 100644
index 00000000000..01b1ef94b36
--- /dev/null
+++ b/arch/mips/cavium-octeon/dma-octeon.c
@@ -0,0 +1,32 @@
+/*
+ * 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) 2000  Ani Joshi <ajoshi@unixbox.com>
+ * Copyright (C) 2000, 2001  Ralf Baechle <ralf@gnu.org>
+ * Copyright (C) 2005 Ilya A. Volynets-Evenbakh <ilya@total-knowledge.com>
+ * swiped from i386, and cloned for MIPS by Geert, polished by Ralf.
+ * IP32 changes by Ilya.
+ * Cavium Networks: Create new dma setup for Cavium Networks Octeon based on
+ * the kernels original.
+ */
+#include <linux/types.h>
+#include <linux/mm.h>
+
+#include <dma-coherence.h>
+
+dma_addr_t octeon_map_dma_mem(struct device *dev, void *ptr, size_t size)
+{
+	/* Without PCI/PCIe this function can be called for Octeon internal
+	   devices such as USB. These devices all support 64bit addressing */
+	mb();
+	return virt_to_phys(ptr);
+}
+
+void octeon_unmap_dma_mem(struct device *dev, dma_addr_t dma_addr)
+{
+	/* Without PCI/PCIe this function can be called for Octeon internal
+	 * devices such as USB. These devices all support 64bit addressing */
+	return;
+}
diff --git a/arch/mips/cavium-octeon/flash_setup.c b/arch/mips/cavium-octeon/flash_setup.c
new file mode 100644
index 00000000000..553d36cbcc4
--- /dev/null
+++ b/arch/mips/cavium-octeon/flash_setup.c
@@ -0,0 +1,84 @@
+/*
+ *   Octeon Bootbus flash setup
+ *
+ * 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) 2007, 2008 Cavium Networks
+ */
+#include <linux/kernel.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/map.h>
+#include <linux/mtd/partitions.h>
+
+#include <asm/octeon/octeon.h>
+
+static struct map_info flash_map;
+static struct mtd_info *mymtd;
+#ifdef CONFIG_MTD_PARTITIONS
+static int nr_parts;
+static struct mtd_partition *parts;
+static const char *part_probe_types[] = {
+	"cmdlinepart",
+#ifdef CONFIG_MTD_REDBOOT_PARTS
+	"RedBoot",
+#endif
+	NULL
+};
+#endif
+
+/**
+ * Module/ driver initialization.
+ *
+ * Returns Zero on success
+ */
+static int __init flash_init(void)
+{
+	/*
+	 * Read the bootbus region 0 setup to determine the base
+	 * address of the flash.
+	 */
+	union cvmx_mio_boot_reg_cfgx region_cfg;
+	region_cfg.u64 = cvmx_read_csr(CVMX_MIO_BOOT_REG_CFGX(0));
+	if (region_cfg.s.en) {
+		/*
+		 * The bootloader always takes the flash and sets its
+		 * address so the entire flash fits below
+		 * 0x1fc00000. This way the flash aliases to
+		 * 0x1fc00000 for booting. Software can access the
+		 * full flash at the true address, while core boot can
+		 * access 4MB.
+		 */
+		/* Use this name so old part lines work */
+		flash_map.name = "phys_mapped_flash";
+		flash_map.phys = region_cfg.s.base << 16;
+		flash_map.size = 0x1fc00000 - flash_map.phys;
+		flash_map.bankwidth = 1;
+		flash_map.virt = ioremap(flash_map.phys, flash_map.size);
+		pr_notice("Bootbus flash: Setting flash for %luMB flash at "
+			  "0x%08lx\n", flash_map.size >> 20, flash_map.phys);
+		simple_map_init(&flash_map);
+		mymtd = do_map_probe("cfi_probe", &flash_map);
+		if (mymtd) {
+			mymtd->owner = THIS_MODULE;
+
+#ifdef CONFIG_MTD_PARTITIONS
+			nr_parts = parse_mtd_partitions(mymtd,
+							part_probe_types,
+							&parts, 0);
+			if (nr_parts > 0)
+				add_mtd_partitions(mymtd, parts, nr_parts);
+			else
+				add_mtd_device(mymtd);
+#else
+			add_mtd_device(mymtd);
+#endif
+		} else {
+			pr_err("Failed to register MTD device for flash\n");
+		}
+	}
+	return 0;
+}
+
+late_initcall(flash_init);
diff --git a/arch/mips/cavium-octeon/octeon-irq.c b/arch/mips/cavium-octeon/octeon-irq.c
new file mode 100644
index 00000000000..fc72984a5da
--- /dev/null
+++ b/arch/mips/cavium-octeon/octeon-irq.c
@@ -0,0 +1,497 @@
+/*
+ * 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) 2004-2008 Cavium Networks
+ */
+#include <linux/irq.h>
+#include <linux/interrupt.h>
+#include <linux/hardirq.h>
+
+#include <asm/octeon/octeon.h>
+
+DEFINE_RWLOCK(octeon_irq_ciu0_rwlock);
+DEFINE_RWLOCK(octeon_irq_ciu1_rwlock);
+DEFINE_SPINLOCK(octeon_irq_msi_lock);
+
+static void octeon_irq_core_ack(unsigned int irq)
+{
+	unsigned int bit = irq - OCTEON_IRQ_SW0;
+	/*
+	 * 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 octeon_irq_core_eoi(unsigned int irq)
+{
+	irq_desc_t *desc = irq_desc + irq;
+	unsigned int bit = irq - OCTEON_IRQ_SW0;
+	/*
+	 * If an IRQ is being processed while we are disabling it the
+	 * handler will attempt to unmask the interrupt after it has
+	 * been disabled.
+	 */
+	if (desc->status & IRQ_DISABLED)
+		return;
+
+	/* There is a race here.  We should fix it.  */
+
+	/*
+	 * 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 << bit);
+}
+
+static void octeon_irq_core_enable(unsigned int irq)
+{
+	unsigned long flags;
+	unsigned int bit = irq - OCTEON_IRQ_SW0;
+
+	/*
+	 * We need to disable interrupts to make sure our updates are
+	 * atomic.
+	 */
+	local_irq_save(flags);
+	set_c0_status(0x100 << bit);
+	local_irq_restore(flags);
+}
+
+static void octeon_irq_core_disable_local(unsigned int irq)
+{
+	unsigned long flags;
+	unsigned int bit = irq - OCTEON_IRQ_SW0;
+	/*
+	 * We need to disable interrupts to make sure our updates are
+	 * atomic.
+	 */
+	local_irq_save(flags);
+	clear_c0_status(0x100 << bit);
+	local_irq_restore(flags);
+}
+
+static void octeon_irq_core_disable(unsigned int irq)
+{
+#ifdef CONFIG_SMP
+	on_each_cpu((void (*)(void *)) octeon_irq_core_disable_local,
+		    (void *) (long) irq, 1);
+#else
+	octeon_irq_core_disable_local(irq);
+#endif
+}
+
+static struct irq_chip octeon_irq_chip_core = {
+	.name = "Core",
+	.enable = octeon_irq_core_enable,
+	.disable = octeon_irq_core_disable,
+	.ack = octeon_irq_core_ack,
+	.eoi = octeon_irq_core_eoi,
+};
+
+
+static void octeon_irq_ciu0_ack(unsigned int irq)
+{
+	/*
+	 * In order to avoid any locking accessing the CIU, we
+	 * acknowledge CIU interrupts by disabling all of them.  This
+	 * way we can use a per core register and avoid any out of
+	 * core locking requirements.  This has the side affect that
+	 * CIU interrupts can't be processed recursively.
+	 *
+	 * 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 << 2);
+}
+
+static void octeon_irq_ciu0_eoi(unsigned int irq)
+{
+	/*
+	 * Enable all CIU interrupts again.  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 << 2);
+}
+
+static void octeon_irq_ciu0_enable(unsigned int irq)
+{
+	int coreid = cvmx_get_core_num();
+	unsigned long flags;
+	uint64_t en0;
+	int bit = irq - OCTEON_IRQ_WORKQ0;	/* Bit 0-63 of EN0 */
+
+	/*
+	 * A read lock is used here to make sure only one core is ever
+	 * updating the CIU enable bits at a time. During an enable
+	 * the cores don't interfere with each other. During a disable
+	 * the write lock stops any enables that might cause a
+	 * problem.
+	 */
+	read_lock_irqsave(&octeon_irq_ciu0_rwlock, flags);
+	en0 = cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
+	en0 |= 1ull << bit;
+	cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), en0);
+	cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
+	read_unlock_irqrestore(&octeon_irq_ciu0_rwlock, flags);
+}
+
+static void octeon_irq_ciu0_disable(unsigned int irq)
+{
+	int bit = irq - OCTEON_IRQ_WORKQ0;	/* Bit 0-63 of EN0 */
+	unsigned long flags;
+	uint64_t en0;
+#ifdef CONFIG_SMP
+	int cpu;
+	write_lock_irqsave(&octeon_irq_ciu0_rwlock, flags);
+	for_each_online_cpu(cpu) {
+		int coreid = cpu_logical_map(cpu);
+		en0 = cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
+		en0 &= ~(1ull << bit);
+		cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), en0);
+	}
+	/*
+	 * We need to do a read after the last update to make sure all
+	 * of them are done.
+	 */
+	cvmx_read_csr(CVMX_CIU_INTX_EN0(cvmx_get_core_num() * 2));
+	write_unlock_irqrestore(&octeon_irq_ciu0_rwlock, flags);
+#else
+	int coreid = cvmx_get_core_num();
+	local_irq_save(flags);
+	en0 = cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
+	en0 &= ~(1ull << bit);
+	cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), en0);
+	cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
+	local_irq_restore(flags);
+#endif
+}
+
+#ifdef CONFIG_SMP
+static void octeon_irq_ciu0_set_affinity(unsigned int irq, const struct cpumask *dest)
+{
+	int cpu;
+	int bit = irq - OCTEON_IRQ_WORKQ0;	/* Bit 0-63 of EN0 */
+
+	write_lock(&octeon_irq_ciu0_rwlock);
+	for_each_online_cpu(cpu) {
+		int coreid = cpu_logical_map(cpu);
+		uint64_t en0 =
+			cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
+		if (cpumask_test_cpu(cpu, dest))
+			en0 |= 1ull << bit;
+		else
+			en0 &= ~(1ull << bit);
+		cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), en0);
+	}
+	/*
+	 * We need to do a read after the last update to make sure all
+	 * of them are done.
+	 */
+	cvmx_read_csr(CVMX_CIU_INTX_EN0(cvmx_get_core_num() * 2));
+	write_unlock(&octeon_irq_ciu0_rwlock);
+}
+#endif
+
+static struct irq_chip octeon_irq_chip_ciu0 = {
+	.name = "CIU0",
+	.enable = octeon_irq_ciu0_enable,
+	.disable = octeon_irq_ciu0_disable,
+	.ack = octeon_irq_ciu0_ack,
+	.eoi = octeon_irq_ciu0_eoi,
+#ifdef CONFIG_SMP
+	.set_affinity = octeon_irq_ciu0_set_affinity,
+#endif
+};
+
+
+static void octeon_irq_ciu1_ack(unsigned int irq)
+{
+	/*
+	 * In order to avoid any locking accessing the CIU, we
+	 * acknowledge CIU interrupts by disabling all of them.  This
+	 * way we can use a per core register and avoid any out of
+	 * core locking requirements.  This has the side affect that
+	 * CIU interrupts can't be processed recursively.  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 << 3);
+}
+
+static void octeon_irq_ciu1_eoi(unsigned int irq)
+{
+	/*
+	 * Enable all CIU interrupts again.  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 << 3);
+}
+
+static void octeon_irq_ciu1_enable(unsigned int irq)
+{
+	int coreid = cvmx_get_core_num();
+	unsigned long flags;
+	uint64_t en1;
+	int bit = irq - OCTEON_IRQ_WDOG0;	/* Bit 0-63 of EN1 */
+
+	/*
+	 * A read lock is used here to make sure only one core is ever
+	 * updating the CIU enable bits at a time.  During an enable
+	 * the cores don't interfere with each other.  During a disable
+	 * the write lock stops any enables that might cause a
+	 * problem.
+	 */
+	read_lock_irqsave(&octeon_irq_ciu1_rwlock, flags);
+	en1 = cvmx_read_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1));
+	en1 |= 1ull << bit;
+	cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), en1);
+	cvmx_read_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1));
+	read_unlock_irqrestore(&octeon_irq_ciu1_rwlock, flags);
+}
+
+static void octeon_irq_ciu1_disable(unsigned int irq)
+{
+	int bit = irq - OCTEON_IRQ_WDOG0;	/* Bit 0-63 of EN1 */
+	unsigned long flags;
+	uint64_t en1;
+#ifdef CONFIG_SMP
+	int cpu;
+	write_lock_irqsave(&octeon_irq_ciu1_rwlock, flags);
+	for_each_online_cpu(cpu) {
+		int coreid = cpu_logical_map(cpu);
+		en1 = cvmx_read_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1));
+		en1 &= ~(1ull << bit);
+		cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), en1);
+	}
+	/*
+	 * We need to do a read after the last update to make sure all
+	 * of them are done.
+	 */
+	cvmx_read_csr(CVMX_CIU_INTX_EN1(cvmx_get_core_num() * 2 + 1));
+	write_unlock_irqrestore(&octeon_irq_ciu1_rwlock, flags);
+#else
+	int coreid = cvmx_get_core_num();
+	local_irq_save(flags);
+	en1 = cvmx_read_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1));
+	en1 &= ~(1ull << bit);
+	cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), en1);
+	cvmx_read_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1));
+	local_irq_restore(flags);
+#endif
+}
+
+#ifdef CONFIG_SMP
+static void octeon_irq_ciu1_set_affinity(unsigned int irq, const struct cpumask *dest)
+{
+	int cpu;
+	int bit = irq - OCTEON_IRQ_WDOG0;	/* Bit 0-63 of EN1 */
+
+	write_lock(&octeon_irq_ciu1_rwlock);
+	for_each_online_cpu(cpu) {
+		int coreid = cpu_logical_map(cpu);
+		uint64_t en1 =
+			cvmx_read_csr(CVMX_CIU_INTX_EN1
+				(coreid * 2 + 1));
+		if (cpumask_test_cpu(cpu, dest))
+			en1 |= 1ull << bit;
+		else
+			en1 &= ~(1ull << bit);
+		cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), en1);
+	}
+	/*
+	 * We need to do a read after the last update to make sure all
+	 * of them are done.
+	 */
+	cvmx_read_csr(CVMX_CIU_INTX_EN1(cvmx_get_core_num() * 2 + 1));
+	write_unlock(&octeon_irq_ciu1_rwlock);
+}
+#endif
+
+static struct irq_chip octeon_irq_chip_ciu1 = {
+	.name = "CIU1",
+	.enable = octeon_irq_ciu1_enable,
+	.disable = octeon_irq_ciu1_disable,
+	.ack = octeon_irq_ciu1_ack,
+	.eoi = octeon_irq_ciu1_eoi,
+#ifdef CONFIG_SMP
+	.set_affinity = octeon_irq_ciu1_set_affinity,
+#endif
+};
+
+#ifdef CONFIG_PCI_MSI
+
+static void octeon_irq_msi_ack(unsigned int irq)
+{
+	if (!octeon_has_feature(OCTEON_FEATURE_PCIE)) {
+		/* These chips have PCI */
+		cvmx_write_csr(CVMX_NPI_NPI_MSI_RCV,
+			       1ull << (irq - OCTEON_IRQ_MSI_BIT0));
+	} else {
+		/*
+		 * These chips have PCIe. Thankfully the ACK doesn't
+		 * need any locking.
+		 */
+		cvmx_write_csr(CVMX_PEXP_NPEI_MSI_RCV0,
+			       1ull << (irq - OCTEON_IRQ_MSI_BIT0));
+	}
+}
+
+static void octeon_irq_msi_eoi(unsigned int irq)
+{
+	/* Nothing needed */
+}
+
+static void octeon_irq_msi_enable(unsigned int irq)
+{
+	if (!octeon_has_feature(OCTEON_FEATURE_PCIE)) {
+		/*
+		 * Octeon PCI doesn't have the ability to mask/unmask
+		 * MSI interrupts individually.  Instead of
+		 * masking/unmasking them in groups of 16, we simple
+		 * assume MSI devices are well behaved.  MSI
+		 * interrupts are always enable and the ACK is assumed
+		 * to be enough.
+		 */
+	} else {
+		/* These chips have PCIe.  Note that we only support
+		 * the first 64 MSI interrupts.  Unfortunately all the
+		 * MSI enables are in the same register.  We use
+		 * MSI0's lock to control access to them all.
+		 */
+		uint64_t en;
+		unsigned long flags;
+		spin_lock_irqsave(&octeon_irq_msi_lock, flags);
+		en = cvmx_read_csr(CVMX_PEXP_NPEI_MSI_ENB0);
+		en |= 1ull << (irq - OCTEON_IRQ_MSI_BIT0);
+		cvmx_write_csr(CVMX_PEXP_NPEI_MSI_ENB0, en);
+		cvmx_read_csr(CVMX_PEXP_NPEI_MSI_ENB0);
+		spin_unlock_irqrestore(&octeon_irq_msi_lock, flags);
+	}
+}
+
+static void octeon_irq_msi_disable(unsigned int irq)
+{
+	if (!octeon_has_feature(OCTEON_FEATURE_PCIE)) {
+		/* See comment in enable */
+	} else {
+		/*
+		 * These chips have PCIe.  Note that we only support
+		 * the first 64 MSI interrupts.  Unfortunately all the
+		 * MSI enables are in the same register.  We use
+		 * MSI0's lock to control access to them all.
+		 */
+		uint64_t en;
+		unsigned long flags;
+		spin_lock_irqsave(&octeon_irq_msi_lock, flags);
+		en = cvmx_read_csr(CVMX_PEXP_NPEI_MSI_ENB0);
+		en &= ~(1ull << (irq - OCTEON_IRQ_MSI_BIT0));
+		cvmx_write_csr(CVMX_PEXP_NPEI_MSI_ENB0, en);
+		cvmx_read_csr(CVMX_PEXP_NPEI_MSI_ENB0);
+		spin_unlock_irqrestore(&octeon_irq_msi_lock, flags);
+	}
+}
+
+static struct irq_chip octeon_irq_chip_msi = {
+	.name = "MSI",
+	.enable = octeon_irq_msi_enable,
+	.disable = octeon_irq_msi_disable,
+	.ack = octeon_irq_msi_ack,
+	.eoi = octeon_irq_msi_eoi,
+};
+#endif
+
+void __init arch_init_irq(void)
+{
+	int irq;
+
+#ifdef CONFIG_SMP
+	/* Set the default affinity to the boot cpu. */
+	cpumask_clear(irq_default_affinity);
+	cpumask_set_cpu(smp_processor_id(), irq_default_affinity);
+#endif
+
+	if (NR_IRQS < OCTEON_IRQ_LAST)
+		pr_err("octeon_irq_init: NR_IRQS is set too low\n");
+
+	/* 0 - 15 reserved for i8259 master and slave controller. */
+
+	/* 17 - 23 Mips internal */
+	for (irq = OCTEON_IRQ_SW0; irq <= OCTEON_IRQ_TIMER; irq++) {
+		set_irq_chip_and_handler(irq, &octeon_irq_chip_core,
+					 handle_percpu_irq);
+	}
+
+	/* 24 - 87 CIU_INT_SUM0 */
+	for (irq = OCTEON_IRQ_WORKQ0; irq <= OCTEON_IRQ_BOOTDMA; irq++) {
+		set_irq_chip_and_handler(irq, &octeon_irq_chip_ciu0,
+					 handle_percpu_irq);
+	}
+
+	/* 88 - 151 CIU_INT_SUM1 */
+	for (irq = OCTEON_IRQ_WDOG0; irq <= OCTEON_IRQ_RESERVED151; irq++) {
+		set_irq_chip_and_handler(irq, &octeon_irq_chip_ciu1,
+					 handle_percpu_irq);
+	}
+
+#ifdef CONFIG_PCI_MSI
+	/* 152 - 215 PCI/PCIe MSI interrupts */
+	for (irq = OCTEON_IRQ_MSI_BIT0; irq <= OCTEON_IRQ_MSI_BIT63; irq++) {
+		set_irq_chip_and_handler(irq, &octeon_irq_chip_msi,
+					 handle_percpu_irq);
+	}
+#endif
+	set_c0_status(0x300 << 2);
+}
+
+asmlinkage void plat_irq_dispatch(void)
+{
+	const unsigned long core_id = cvmx_get_core_num();
+	const uint64_t ciu_sum0_address = CVMX_CIU_INTX_SUM0(core_id * 2);
+	const uint64_t ciu_en0_address = CVMX_CIU_INTX_EN0(core_id * 2);
+	const uint64_t ciu_sum1_address = CVMX_CIU_INT_SUM1;
+	const uint64_t ciu_en1_address = CVMX_CIU_INTX_EN1(core_id * 2 + 1);
+	unsigned long cop0_cause;
+	unsigned long cop0_status;
+	uint64_t ciu_en;
+	uint64_t ciu_sum;
+
+	while (1) {
+		cop0_cause = read_c0_cause();
+		cop0_status = read_c0_status();
+		cop0_cause &= cop0_status;
+		cop0_cause &= ST0_IM;
+
+		if (unlikely(cop0_cause & STATUSF_IP2)) {
+			ciu_sum = cvmx_read_csr(ciu_sum0_address);
+			ciu_en = cvmx_read_csr(ciu_en0_address);
+			ciu_sum &= ciu_en;
+			if (likely(ciu_sum))
+				do_IRQ(fls64(ciu_sum) + OCTEON_IRQ_WORKQ0 - 1);
+			else
+				spurious_interrupt();
+		} else if (unlikely(cop0_cause & STATUSF_IP3)) {
+			ciu_sum = cvmx_read_csr(ciu_sum1_address);
+			ciu_en = cvmx_read_csr(ciu_en1_address);
+			ciu_sum &= ciu_en;
+			if (likely(ciu_sum))
+				do_IRQ(fls64(ciu_sum) + OCTEON_IRQ_WDOG0 - 1);
+			else
+				spurious_interrupt();
+		} else if (likely(cop0_cause)) {
+			do_IRQ(fls(cop0_cause) - 9 + MIPS_CPU_IRQ_BASE);
+		} else {
+			break;
+		}
+	}
+}
diff --git a/arch/mips/cavium-octeon/octeon-memcpy.S b/arch/mips/cavium-octeon/octeon-memcpy.S
new file mode 100644
index 00000000000..88e0cddca20
--- /dev/null
+++ b/arch/mips/cavium-octeon/octeon-memcpy.S
@@ -0,0 +1,521 @@
+/*
+ * 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.
+ *
+ * Unified implementation of memcpy, memmove and the __copy_user backend.
+ *
+ * Copyright (C) 1998, 99, 2000, 01, 2002 Ralf Baechle (ralf@gnu.org)
+ * Copyright (C) 1999, 2000, 01, 2002 Silicon Graphics, Inc.
+ * Copyright (C) 2002 Broadcom, Inc.
+ *   memcpy/copy_user author: Mark Vandevoorde
+ *
+ * Mnemonic names for arguments to memcpy/__copy_user
+ */
+
+#include <asm/asm.h>
+#include <asm/asm-offsets.h>
+#include <asm/regdef.h>
+
+#define dst a0
+#define src a1
+#define len a2
+
+/*
+ * Spec
+ *
+ * memcpy copies len bytes from src to dst and sets v0 to dst.
+ * It assumes that
+ *   - src and dst don't overlap
+ *   - src is readable
+ *   - dst is writable
+ * memcpy uses the standard calling convention
+ *
+ * __copy_user copies up to len bytes from src to dst and sets a2 (len) to
+ * the number of uncopied bytes due to an exception caused by a read or write.
+ * __copy_user assumes that src and dst don't overlap, and that the call is
+ * implementing one of the following:
+ *   copy_to_user
+ *     - src is readable  (no exceptions when reading src)
+ *   copy_from_user
+ *     - dst is writable  (no exceptions when writing dst)
+ * __copy_user uses a non-standard calling convention; see
+ * arch/mips/include/asm/uaccess.h
+ *
+ * When an exception happens on a load, the handler must
+ # ensure that all of the destination buffer is overwritten to prevent
+ * leaking information to user mode programs.
+ */
+
+/*
+ * Implementation
+ */
+
+/*
+ * The exception handler for loads requires that:
+ *  1- AT contain the address of the byte just past the end of the source
+ *     of the copy,
+ *  2- src_entry <= src < AT, and
+ *  3- (dst - src) == (dst_entry - src_entry),
+ * The _entry suffix denotes values when __copy_user was called.
+ *
+ * (1) is set up up by uaccess.h and maintained by not writing AT in copy_user
+ * (2) is met by incrementing src by the number of bytes copied
+ * (3) is met by not doing loads between a pair of increments of dst and src
+ *
+ * The exception handlers for stores adjust len (if necessary) and return.
+ * These handlers do not need to overwrite any data.
+ *
+ * For __rmemcpy and memmove an exception is always a kernel bug, therefore
+ * they're not protected.
+ */
+
+#define EXC(inst_reg,addr,handler)		\
+9:	inst_reg, addr;				\
+	.section __ex_table,"a";		\
+	PTR	9b, handler;			\
+	.previous
+
+/*
+ * Only on the 64-bit kernel we can made use of 64-bit registers.
+ */
+#ifdef CONFIG_64BIT
+#define USE_DOUBLE
+#endif
+
+#ifdef USE_DOUBLE
+
+#define LOAD   ld
+#define LOADL  ldl
+#define LOADR  ldr
+#define STOREL sdl
+#define STORER sdr
+#define STORE  sd
+#define ADD    daddu
+#define SUB    dsubu
+#define SRL    dsrl
+#define SRA    dsra
+#define SLL    dsll
+#define SLLV   dsllv
+#define SRLV   dsrlv
+#define NBYTES 8
+#define LOG_NBYTES 3
+
+/*
+ * As we are sharing code base with the mips32 tree (which use the o32 ABI
+ * register definitions). We need to redefine the register definitions from
+ * the n64 ABI register naming to the o32 ABI register naming.
+ */
+#undef t0
+#undef t1
+#undef t2
+#undef t3
+#define t0	$8
+#define t1	$9
+#define t2	$10