Merge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (38 commits)
  MIPS: O32: Fix ppoll
  MIPS: Oprofile: Rename cpu_type from godson2 to loongson2
  MIPS: Alchemy: Fix hang with high-frequency edge interrupts
  MIPS: TXx9: Fix spi-baseclk value
  MIPS: bcm63xx: Set the correct BCM3302 CPU name
  MIPS: Loongson 2: Set cpu_has_dc_aliases and cpu_icache_snoops_remote_store
  MIPS: Avoid potential hazard on Context register
  MIPS: Octeon: Use lockless interrupt controller operations when possible.
  MIPS: Octeon: Use write_{un,}lock_irq{restore,save} to set irq affinity
  MIPS: Set S-cache linesize to 64-bytes for MTI's S-cache
  MIPS: SMTC: Avoid queing multiple reschedule IPIs
  MIPS: GCMP: Avoid accessing registers when they are not present
  MIPS: GIC: Random fixes and enhancements.
  MIPS: CMP: Fix memory barriers for correct operation of amon_cpu_start
  MIPS: Fix abs.[sd] and neg.[sd] emulation for NaN operands
  MIPS: SPRAM: Clean up support code a little
  MIPS: 1004K: Enable SPRAM support.
  MIPS: Malta: Enable PCI 2.1 compatibility in PIIX4
  MIPS: Kconfig: Fix duplicate default value for MIPS_L1_CACHE_SHIFT.
  MIPS: MTI: Fix accesses to device registers on MIPS boards
  ...

54 files changed, 544 insertions, 440 deletions

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 705a7a9170f..03bd56a2fb6 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -1012,9 +1012,9 @@ config BOOT_ELF32
 
 config MIPS_L1_CACHE_SHIFT
 	int
-	default "4" if MACH_DECSTATION || MIKROTIK_RB532
+	default "4" if MACH_DECSTATION || MIKROTIK_RB532 || PMC_MSP4200_EVAL
+	default "6" if MIPS_CPU_SCACHE
 	default "7" if SGI_IP22 || SGI_IP27 || SGI_IP28 || SNI_RM || CPU_CAVIUM_OCTEON
-	default "4" if PMC_MSP4200_EVAL
 	default "5"
 
 config HAVE_STD_PC_SERIAL_PORT
diff --git a/arch/mips/alchemy/common/irq.c b/arch/mips/alchemy/common/irq.c
index c88c821b4c3..d670928afcf 100644
--- a/arch/mips/alchemy/common/irq.c
+++ b/arch/mips/alchemy/common/irq.c
@@ -354,6 +354,28 @@ static void au1x_ic1_ack(unsigned int irq_nr)
 	au_sync();
 }
 
+static void au1x_ic0_maskack(unsigned int irq_nr)
+{
+	unsigned int bit = irq_nr - AU1000_INTC0_INT_BASE;
+
+	au_writel(1 << bit, IC0_WAKECLR);
+	au_writel(1 << bit, IC0_MASKCLR);
+	au_writel(1 << bit, IC0_RISINGCLR);
+	au_writel(1 << bit, IC0_FALLINGCLR);
+	au_sync();
+}
+
+static void au1x_ic1_maskack(unsigned int irq_nr)
+{
+	unsigned int bit = irq_nr - AU1000_INTC1_INT_BASE;
+
+	au_writel(1 << bit, IC1_WAKECLR);
+	au_writel(1 << bit, IC1_MASKCLR);
+	au_writel(1 << bit, IC1_RISINGCLR);
+	au_writel(1 << bit, IC1_FALLINGCLR);
+	au_sync();
+}
+
 static int au1x_ic1_setwake(unsigned int irq, unsigned int on)
 {
 	unsigned int bit = irq - AU1000_INTC1_INT_BASE;
@@ -379,25 +401,21 @@ static int au1x_ic1_setwake(unsigned int irq, unsigned int on)
 /*
  * irq_chips for both ICs; this way the mask handlers can be
  * as short as possible.
- *
- * NOTE: the ->ack() callback is used by the handle_edge_irq
- *	 flowhandler only, the ->mask_ack() one by handle_level_irq,
- *	 so no need for an irq_chip for each type of irq (level/edge).
  */
 static struct irq_chip au1x_ic0_chip = {
 	.name		= "Alchemy-IC0",
-	.ack		= au1x_ic0_ack,		/* edge */
+	.ack		= au1x_ic0_ack,
 	.mask		= au1x_ic0_mask,
-	.mask_ack	= au1x_ic0_mask,	/* level */
+	.mask_ack	= au1x_ic0_maskack,
 	.unmask		= au1x_ic0_unmask,
 	.set_type	= au1x_ic_settype,
 };
 
 static struct irq_chip au1x_ic1_chip = {
 	.name		= "Alchemy-IC1",
-	.ack		= au1x_ic1_ack,		/* edge */
+	.ack		= au1x_ic1_ack,
 	.mask		= au1x_ic1_mask,
-	.mask_ack	= au1x_ic1_mask,	/* level */
+	.mask_ack	= au1x_ic1_maskack,
 	.unmask		= au1x_ic1_unmask,
 	.set_type	= au1x_ic_settype,
 	.set_wake	= au1x_ic1_setwake,
diff --git a/arch/mips/alchemy/mtx-1/board_setup.c b/arch/mips/alchemy/mtx-1/board_setup.c
index cc32c69a74a..45b61c9b82b 100644
--- a/arch/mips/alchemy/mtx-1/board_setup.c
+++ b/arch/mips/alchemy/mtx-1/board_setup.c
@@ -69,6 +69,7 @@ void __init board_setup(void)
 #else
 	au_writel(0xf, Au1500_PCI_CFG);
 #endif
+	board_pci_idsel = mtx1_pci_idsel;
 #endif
 
 	/* Initialize sys_pinfunc */
@@ -85,8 +86,6 @@ void __init board_setup(void)
 	alchemy_gpio_direction_output(211, 1);	/* green on */
 	alchemy_gpio_direction_output(212, 0);	/* red off */
 
-	board_pci_idsel = mtx1_pci_idsel;
-
 	printk(KERN_INFO "4G Systems MTX-1 Board\n");
 }
 
diff --git a/arch/mips/ar7/platform.c b/arch/mips/ar7/platform.c
index e2278c04459..835f3f0319c 100644
--- a/arch/mips/ar7/platform.c
+++ b/arch/mips/ar7/platform.c
@@ -503,6 +503,7 @@ static int __init ar7_register_devices(void)
 {
 	u16 chip_id;
 	int res;
+	u32 *bootcr, val;
 #ifdef CONFIG_SERIAL_8250
 	static struct uart_port uart_port[2];
 
@@ -595,7 +596,13 @@ static int __init ar7_register_devices(void)
 
 	ar7_wdt_res.end = ar7_wdt_res.start + 0x20;
 
-	res = platform_device_register(&ar7_wdt);
+	bootcr = (u32 *)ioremap_nocache(AR7_REGS_DCL, 4);
+	val = *bootcr;
+	iounmap(bootcr);
+
+	/* Register watchdog only if enabled in hardware */
+	if (val & AR7_WDT_HW_ENA)
+		res = platform_device_register(&ar7_wdt);
 
 	return res;
 }
diff --git a/arch/mips/bcm63xx/Makefile b/arch/mips/bcm63xx/Makefile
index c146d1edede..00064b66080 100644
--- a/arch/mips/bcm63xx/Makefile
+++ b/arch/mips/bcm63xx/Makefile
@@ -1,5 +1,5 @@
 obj-y		+= clk.o cpu.o cs.o gpio.o irq.o prom.o setup.o timer.o \
-		   dev-dsp.o dev-enet.o dev-pcmcia.o dev-uart.o
+		   dev-dsp.o dev-enet.o dev-pcmcia.o dev-uart.o dev-wdt.o
 obj-$(CONFIG_EARLY_PRINTK)	+= early_printk.o
 
 obj-y		+= boards/
diff --git a/arch/mips/bcm63xx/boards/board_bcm963xx.c b/arch/mips/bcm63xx/boards/board_bcm963xx.c
index 78e155d21be..05a35cf5963 100644
--- a/arch/mips/bcm63xx/boards/board_bcm963xx.c
+++ b/arch/mips/bcm63xx/boards/board_bcm963xx.c
@@ -24,7 +24,6 @@
 #include <bcm63xx_dev_enet.h>
 #include <bcm63xx_dev_dsp.h>
 #include <bcm63xx_dev_pcmcia.h>
-#include <bcm63xx_dev_uart.h>
 #include <board_bcm963xx.h>
 
 #define PFX	"board_bcm963xx: "
@@ -794,8 +793,6 @@ int __init board_register_devices(void)
 {
 	u32 val;
 
-	bcm63xx_uart_register();
-
 	if (board.has_pccard)
 		bcm63xx_pcmcia_register();
 
diff --git a/arch/mips/bcm63xx/cpu.c b/arch/mips/bcm63xx/cpu.c
index 6dc43f0483e..70378bb5e3f 100644
--- a/arch/mips/bcm63xx/cpu.c
+++ b/arch/mips/bcm63xx/cpu.c
@@ -10,6 +10,7 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/cpu.h>
+#include <asm/cpu-info.h>
 #include <bcm63xx_cpu.h>
 #include <bcm63xx_regs.h>
 #include <bcm63xx_io.h>
@@ -284,6 +285,7 @@ void __init bcm63xx_cpu_init(void)
 {
 	unsigned int tmp, expected_cpu_id;
 	struct cpuinfo_mips *c = &current_cpu_data;
+	unsigned int cpu = smp_processor_id();
 
 	/* soc registers location depends on cpu type */
 	expected_cpu_id = 0;
@@ -293,6 +295,7 @@ void __init bcm63xx_cpu_init(void)
 	 * BCM6338 as the same PrId as BCM3302 see arch/mips/kernel/cpu-probe.c
 	 */
 	case CPU_BCM3302:
+		__cpu_name[cpu] = "Broadcom BCM6338";
 		expected_cpu_id = BCM6338_CPU_ID;
 		bcm63xx_regs_base = bcm96338_regs_base;
 		bcm63xx_irqs = bcm96338_irqs;
diff --git a/arch/mips/bcm63xx/dev-uart.c b/arch/mips/bcm63xx/dev-uart.c
index 5f3d89c4a98..b0519461ad9 100644
--- a/arch/mips/bcm63xx/dev-uart.c
+++ b/arch/mips/bcm63xx/dev-uart.c
@@ -10,7 +10,6 @@
 #include <linux/kernel.h>
 #include <linux/platform_device.h>
 #include <bcm63xx_cpu.h>
-#include <bcm63xx_dev_uart.h>
 
 static struct resource uart_resources[] = {
 	{
@@ -39,3 +38,4 @@ int __init bcm63xx_uart_register(void)
 	uart_resources[1].start = bcm63xx_get_irq_number(IRQ_UART0);
 	return platform_device_register(&bcm63xx_uart_device);
 }
+arch_initcall(bcm63xx_uart_register);
diff --git a/arch/mips/bcm63xx/dev-wdt.c b/arch/mips/bcm63xx/dev-wdt.c
new file mode 100644
index 00000000000..3e6c716a4c1
--- /dev/null
+++ b/arch/mips/bcm63xx/dev-wdt.c
@@ -0,0 +1,37 @@
+/*
+ * 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) 2008 Florian Fainelli <florian@openwrt.org>
+ */
+
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/platform_device.h>
+#include <bcm63xx_cpu.h>
+
+static struct resource wdt_resources[] = {
+	{
+		.start		= -1, /* filled at runtime */
+		.end		= -1, /* filled at runtime */
+		.flags		= IORESOURCE_MEM,
+	},
+};
+
+static struct platform_device bcm63xx_wdt_device = {
+	.name		= "bcm63xx-wdt",
+	.id		= 0,
+	.num_resources	= ARRAY_SIZE(wdt_resources),
+	.resource	= wdt_resources,
+};
+
+int __init bcm63xx_wdt_register(void)
+{
+	wdt_resources[0].start = bcm63xx_regset_address(RSET_WDT);
+	wdt_resources[0].end = wdt_resources[0].start;
+	wdt_resources[0].end += RSET_WDT_SIZE - 1;
+
+	return platform_device_register(&bcm63xx_wdt_device);
+}
+arch_initcall(bcm63xx_wdt_register);
diff --git a/arch/mips/bcm63xx/setup.c b/arch/mips/bcm63xx/setup.c
index b18a0ca926f..d0056598fbf 100644
--- a/arch/mips/bcm63xx/setup.c
+++ b/arch/mips/bcm63xx/setup.c
@@ -75,7 +75,9 @@ void bcm63xx_machine_reboot(void)
 		bcm6348_a1_reboot();
 
 	printk(KERN_INFO "triggering watchdog soft-reset...\n");
-	bcm_perf_writel(SYS_PLL_SOFT_RESET, PERF_SYS_PLL_CTL_REG);
+	reg = bcm_perf_readl(PERF_SYS_PLL_CTL_REG);
+	reg |= SYS_PLL_SOFT_RESET;
+	bcm_perf_writel(reg, PERF_SYS_PLL_CTL_REG);
 	while (1)
 		;
 }
diff --git a/arch/mips/cavium-octeon/octeon-irq.c b/arch/mips/cavium-octeon/octeon-irq.c
index 384f1842bfb..6f2acf09328 100644
--- a/arch/mips/cavium-octeon/octeon-irq.c
+++ b/arch/mips/cavium-octeon/octeon-irq.c
@@ -17,6 +17,15 @@ DEFINE_RWLOCK(octeon_irq_ciu0_rwlock);
 DEFINE_RWLOCK(octeon_irq_ciu1_rwlock);
 DEFINE_SPINLOCK(octeon_irq_msi_lock);
 
+static int octeon_coreid_for_cpu(int cpu)
+{
+#ifdef CONFIG_SMP
+	return cpu_logical_map(cpu);
+#else
+	return cvmx_get_core_num();
+#endif
+}
+
 static void octeon_irq_core_ack(unsigned int irq)
 {
 	unsigned int bit = irq - OCTEON_IRQ_SW0;
@@ -152,11 +161,10 @@ 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);
+		int coreid = octeon_coreid_for_cpu(cpu);
 		en0 = cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
 		en0 &= ~(1ull << bit);
 		cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), en0);
@@ -167,26 +175,57 @@ static void octeon_irq_ciu0_disable(unsigned int irq)
 	 */
 	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
+}
+
+/*
+ * Enable the irq on the current core for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static void octeon_irq_ciu0_enable_v2(unsigned int irq)
+{
+	int index = cvmx_get_core_num() * 2;
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WORKQ0);
+
+	cvmx_write_csr(CVMX_CIU_INTX_EN0_W1S(index), mask);
+}
+
+/*
+ * Disable the irq on the current core for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static void octeon_irq_ciu0_disable_v2(unsigned int irq)
+{
+	int index = cvmx_get_core_num() * 2;
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WORKQ0);
+
+	cvmx_write_csr(CVMX_CIU_INTX_EN0_W1C(index), mask);
+}
+
+/*
+ * Disable the irq on the all cores for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static void octeon_irq_ciu0_disable_all_v2(unsigned int irq)
+{
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WORKQ0);
+	int index;
+	int cpu;
+	for_each_online_cpu(cpu) {
+		index = octeon_coreid_for_cpu(cpu) * 2;
+		cvmx_write_csr(CVMX_CIU_INTX_EN0_W1C(index), mask);
+	}
 }
 
 #ifdef CONFIG_SMP
 static int octeon_irq_ciu0_set_affinity(unsigned int irq, const struct cpumask *dest)
 {
 	int cpu;
+	unsigned long flags;
 	int bit = irq - OCTEON_IRQ_WORKQ0;	/* Bit 0-63 of EN0 */
 
-	write_lock(&octeon_irq_ciu0_rwlock);
+	write_lock_irqsave(&octeon_irq_ciu0_rwlock, flags);
 	for_each_online_cpu(cpu) {
-		int coreid = cpu_logical_map(cpu);
+		int coreid = octeon_coreid_for_cpu(cpu);
 		uint64_t en0 =
 			cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2));
 		if (cpumask_test_cpu(cpu, dest))
@@ -200,11 +239,45 @@ static int octeon_irq_ciu0_set_affinity(unsigned int irq, const struct cpumask *
 	 * of them are done.
 	 */
 	cvmx_read_csr(CVMX_CIU_INTX_EN0(cvmx_get_core_num() * 2));
-	write_unlock(&octeon_irq_ciu0_rwlock);
+	write_unlock_irqrestore(&octeon_irq_ciu0_rwlock, flags);
 
 	return 0;
 }
+
+/*
+ * Set affinity for the irq for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static int octeon_irq_ciu0_set_affinity_v2(unsigned int irq,
+					   const struct cpumask *dest)
+{
+	int cpu;
+	int index;
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WORKQ0);
+	for_each_online_cpu(cpu) {
+		index = octeon_coreid_for_cpu(cpu) * 2;
+		if (cpumask_test_cpu(cpu, dest))
+			cvmx_write_csr(CVMX_CIU_INTX_EN0_W1S(index), mask);
+		else
+			cvmx_write_csr(CVMX_CIU_INTX_EN0_W1C(index), mask);
+	}
+	return 0;
+}
+#endif
+
+/*
+ * Newer octeon chips have support for lockless CIU operation.
+ */
+static struct irq_chip octeon_irq_chip_ciu0_v2 = {
+	.name = "CIU0",
+	.enable = octeon_irq_ciu0_enable_v2,
+	.disable = octeon_irq_ciu0_disable_all_v2,
+	.ack = octeon_irq_ciu0_disable_v2,
+	.eoi = octeon_irq_ciu0_enable_v2,
+#ifdef CONFIG_SMP
+	.set_affinity = octeon_irq_ciu0_set_affinity_v2,
 #endif
+};
 
 static struct irq_chip octeon_irq_chip_ciu0 = {
 	.name = "CIU0",
@@ -269,11 +342,10 @@ 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);
+		int coreid = octeon_coreid_for_cpu(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);
@@ -284,26 +356,58 @@ static void octeon_irq_ciu1_disable(unsigned int irq)
 	 */
 	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
+}
+
+/*
+ * Enable the irq on the current core for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static void octeon_irq_ciu1_enable_v2(unsigned int irq)
+{
+	int index = cvmx_get_core_num() * 2 + 1;
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WDOG0);
+
+	cvmx_write_csr(CVMX_CIU_INTX_EN1_W1S(index), mask);
+}
+
+/*
+ * Disable the irq on the current core for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static void octeon_irq_ciu1_disable_v2(unsigned int irq)
+{
+	int index = cvmx_get_core_num() * 2 + 1;
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WDOG0);
+
+	cvmx_write_csr(CVMX_CIU_INTX_EN1_W1C(index), mask);
+}
+
+/*
+ * Disable the irq on the all cores for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static void octeon_irq_ciu1_disable_all_v2(unsigned int irq)
+{
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WDOG0);
+	int index;
+	int cpu;
+	for_each_online_cpu(cpu) {
+		index = octeon_coreid_for_cpu(cpu) * 2 + 1;
+		cvmx_write_csr(CVMX_CIU_INTX_EN1_W1C(index), mask);
+	}
 }
 
 #ifdef CONFIG_SMP
-static int octeon_irq_ciu1_set_affinity(unsigned int irq, const struct cpumask *dest)
+static int octeon_irq_ciu1_set_affinity(unsigned int irq,
+					const struct cpumask *dest)
 {
 	int cpu;
+	unsigned long flags;
 	int bit = irq - OCTEON_IRQ_WDOG0;	/* Bit 0-63 of EN1 */
 
-	write_lock(&octeon_irq_ciu1_rwlock);
+	write_lock_irqsave(&octeon_irq_ciu1_rwlock, flags);
 	for_each_online_cpu(cpu) {
-		int coreid = cpu_logical_map(cpu);
+		int coreid = octeon_coreid_for_cpu(cpu);
 		uint64_t en1 =
 			cvmx_read_csr(CVMX_CIU_INTX_EN1
 				(coreid * 2 + 1));
@@ -318,12 +422,46 @@ static int octeon_irq_ciu1_set_affinity(unsigned int irq, const struct cpumask *
 	 * of them are done.
 	 */
 	cvmx_read_csr(CVMX_CIU_INTX_EN1(cvmx_get_core_num() * 2 + 1));
-	write_unlock(&octeon_irq_ciu1_rwlock);
+	write_unlock_irqrestore(&octeon_irq_ciu1_rwlock, flags);
+
+	return 0;
+}
 
+/*
+ * Set affinity for the irq for chips that have the EN*_W1{S,C}
+ * registers.
+ */
+static int octeon_irq_ciu1_set_affinity_v2(unsigned int irq,
+					   const struct cpumask *dest)
+{
+	int cpu;
+	int index;
+	u64 mask = 1ull << (irq - OCTEON_IRQ_WDOG0);
+	for_each_online_cpu(cpu) {
+		index = octeon_coreid_for_cpu(cpu) * 2 + 1;
+		if (cpumask_test_cpu(cpu, dest))
+			cvmx_write_csr(CVMX_CIU_INTX_EN1_W1S(index), mask);
+		else
+			cvmx_write_csr(CVMX_CIU_INTX_EN1_W1C(index), mask);
+	}
 	return 0;
 }
 #endif
 
+/*
+ * Newer octeon chips have support for lockless CIU operation.
+ */
+static struct irq_chip octeon_irq_chip_ciu1_v2 = {
+	.name = "CIU0",
+	.enable = octeon_irq_ciu1_enable_v2,
+	.disable = octeon_irq_ciu1_disable_all_v2,
+	.ack = octeon_irq_ciu1_disable_v2,
+	.eoi = octeon_irq_ciu1_enable_v2,
+#ifdef CONFIG_SMP
+	.set_affinity = octeon_irq_ciu1_set_affinity_v2,
+#endif
+};
+
 static struct irq_chip octeon_irq_chip_ciu1 = {
 	.name = "CIU1",
 	.enable = octeon_irq_ciu1_enable,
@@ -420,6 +558,8 @@ static struct irq_chip octeon_irq_chip_msi = {
 void __init arch_init_irq(void)
 {
 	int irq;
+	struct irq_chip *chip0;
+	struct irq_chip *chip1;
 
 #ifdef CONFIG_SMP
 	/* Set the default affinity to the boot cpu. */
@@ -430,6 +570,16 @@ void __init arch_init_irq(void)
 	if (NR_IRQS < OCTEON_IRQ_LAST)
 		pr_err("octeon_irq_init: NR_IRQS is set too low\n");
 
+	if (OCTEON_IS_MODEL(OCTEON_CN58XX_PASS2_X) ||
+	    OCTEON_IS_MODEL(OCTEON_CN56XX_PASS2_X) ||
+	    OCTEON_IS_MODEL(OCTEON_CN52XX_PASS2_X)) {
+		chip0 = &octeon_irq_chip_ciu0_v2;
+		chip1 = &octeon_irq_chip_ciu1_v2;
+	} else {
+		chip0 = &octeon_irq_chip_ciu0;
+		chip1 = &octeon_irq_chip_ciu1;
+	}
+
 	/* 0 - 15 reserved for i8259 master and slave controller. */
 
 	/* 17 - 23 Mips internal */
@@ -440,14 +590,12 @@ void __init arch_init_irq(void)
 
 	/* 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);
+		set_irq_chip_and_handler(irq, chip0, 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);
+		set_irq_chip_and_handler(irq, chip1, handle_percpu_irq);
 	}
 
 #ifdef CONFIG_PCI_MSI
@@ -505,14 +653,10 @@ asmlinkage void plat_irq_dispatch(void)
 #ifdef CONFIG_HOTPLUG_CPU
 static int is_irq_enabled_on_cpu(unsigned int irq, unsigned int cpu)
 {
-       unsigned int isset;
-#ifdef CONFIG_SMP
-       int coreid = cpu_logical_map(cpu);
-#else
-	int coreid = cvmx_get_core_num();
-#endif
+	unsigned int isset;
+	int coreid = octeon_coreid_for_cpu(cpu);
 	int bit = (irq < OCTEON_IRQ_WDOG0) ?
-		irq - OCTEON_IRQ_WORKQ0 : irq - OCTEON_IRQ_WDOG0;
+		   irq - OCTEON_IRQ_WORKQ0 : irq - OCTEON_IRQ_WDOG0;
        if (irq < 64) {
 		isset = (cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2)) &
 			(1ull << bit)) >> bit;
diff --git a/arch/mips/cavium-octeon/smp.c b/arch/mips/cavium-octeon/smp.c
index 32d51a31dc4..c198efdf583 100644
--- a/arch/mips/cavium-octeon/smp.c
+++ b/arch/mips/cavium-octeon/smp.c
@@ -65,11 +65,12 @@ void octeon_send_ipi_single(int cpu, unsigned int action)
 	cvmx_write_csr(CVMX_CIU_MBOX_SETX(coreid), action);
 }
 
-static inline void octeon_send_ipi_mask(cpumask_t mask, unsigned int action)
+static inline void octeon_send_ipi_mask(const struct cpumask *mask,
+					unsigned int action)
 {
 	unsigne