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

Pull MIPS updates from Ralf Baechle:
 "This is the main pull request for MIPS:

   - a number of fixes that didn't make the 3.19 release.

   - a number of cleanups.

   - preliminary support for Cavium's Octeon 3 SOCs which feature up to
     48 MIPS64 R3 cores with FPU and hardware virtualization.

   - support for MIPS R6 processors.

     Revision 6 of the MIPS architecture is a major revision of the MIPS
     architecture which does away with many of original sins of the
     architecture such as branch delay slots.  This and other changes in
     R6 require major changes throughout the entire MIPS core
     architecture code and make up for the lion share of this pull
     request.

   - finally some preparatory work for eXtendend Physical Address
     support, which allows support of up to 40 bit of physical address
     space on 32 bit processors"

     [ Ahh, MIPS can't leave the PAE brain damage alone.  It's like
       every CPU architect has to make that mistake, but pee in the snow
       by changing the TLA.  But whether it's called PAE, LPAE or XPA,
       it's horrid crud   - Linus ]

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus: (114 commits)
  MIPS: sead3: Corrected get_c0_perfcount_int
  MIPS: mm: Remove dead macro definitions
  MIPS: OCTEON: irq: add CIB and other fixes
  MIPS: OCTEON: Don't do acknowledge operations for level triggered irqs.
  MIPS: OCTEON: More OCTEONIII support
  MIPS: OCTEON: Remove setting of processor specific CVMCTL icache bits.
  MIPS: OCTEON: Core-15169 Workaround and general CVMSEG cleanup.
  MIPS: OCTEON: Update octeon-model.h code for new SoCs.
  MIPS: OCTEON: Implement DCache errata workaround for all CN6XXX
  MIPS: OCTEON: Add little-endian support to asm/octeon/octeon.h
  MIPS: OCTEON: Implement the core-16057 workaround
  MIPS: OCTEON: Delete unused COP2 saving code
  MIPS: OCTEON: Use correct instruction to read 64-bit COP0 register
  MIPS: OCTEON: Save and restore CP2 SHA3 state
  MIPS: OCTEON: Fix FP context save.
  MIPS: OCTEON: Save/Restore wider multiply registers in OCTEON III CPUs
  MIPS: boot: Provide more uImage options
  MIPS: Remove unneeded #ifdef __KERNEL__ from asm/processor.h
  MIPS: ip22-gio: Remove legacy suspend/resume support
  mips: pci: Add ifdef around pci_proc_domain
  ...

5 files changed, 885 insertions, 282 deletions

diff --git a/arch/mips/cavium-octeon/csrc-octeon.c b/arch/mips/cavium-octeon/csrc-octeon.c
index b752c4ed0b79..1882e6475dd0 100644
--- a/arch/mips/cavium-octeon/csrc-octeon.c
+++ b/arch/mips/cavium-octeon/csrc-octeon.c
@@ -18,7 +18,7 @@
 #include <asm/octeon/octeon.h>
 #include <asm/octeon/cvmx-ipd-defs.h>
 #include <asm/octeon/cvmx-mio-defs.h>
-
+#include <asm/octeon/cvmx-rst-defs.h>
 
 static u64 f;
 static u64 rdiv;
@@ -39,11 +39,20 @@ void __init octeon_setup_delays(void)
 
 	if (current_cpu_type() == CPU_CAVIUM_OCTEON2) {
 		union cvmx_mio_rst_boot rst_boot;
+
 		rst_boot.u64 = cvmx_read_csr(CVMX_MIO_RST_BOOT);
 		rdiv = rst_boot.s.c_mul;	/* CPU clock */
 		sdiv = rst_boot.s.pnr_mul;	/* I/O clock */
 		f = (0x8000000000000000ull / sdiv) * 2;
+	} else if (current_cpu_type() == CPU_CAVIUM_OCTEON3) {
+		union cvmx_rst_boot rst_boot;
+
+		rst_boot.u64 = cvmx_read_csr(CVMX_RST_BOOT);
+		rdiv = rst_boot.s.c_mul;	/* CPU clock */
+		sdiv = rst_boot.s.pnr_mul;	/* I/O clock */
+		f = (0x8000000000000000ull / sdiv) * 2;
 	}
+
 }
 
 /*
diff --git a/arch/mips/cavium-octeon/dma-octeon.c b/arch/mips/cavium-octeon/dma-octeon.c
index 3778655c4a37..7d8987818ccf 100644
--- a/arch/mips/cavium-octeon/dma-octeon.c
+++ b/arch/mips/cavium-octeon/dma-octeon.c
@@ -276,7 +276,7 @@ void __init plat_swiotlb_setup(void)
 			continue;
 
 		/* These addresses map low for PCI. */
-		if (e->addr > 0x410000000ull && !OCTEON_IS_MODEL(OCTEON_CN6XXX))
+		if (e->addr > 0x410000000ull && !OCTEON_IS_OCTEON2())
 			continue;
 
 		addr_size += e->size;
@@ -308,7 +308,7 @@ void __init plat_swiotlb_setup(void)
 #endif
 #ifdef CONFIG_USB_OCTEON_OHCI
 	/* OCTEON II ohci is only 32-bit. */
-	if (OCTEON_IS_MODEL(OCTEON_CN6XXX) && max_addr >= 0x100000000ul)
+	if (OCTEON_IS_OCTEON2() && max_addr >= 0x100000000ul)
 		swiotlbsize = 64 * (1<<20);
 #endif
 	swiotlb_nslabs = swiotlbsize >> IO_TLB_SHIFT;
diff --git a/arch/mips/cavium-octeon/executive/cvmx-helper-board.c b/arch/mips/cavium-octeon/executive/cvmx-helper-board.c
index 5dfef84b9576..9eb0feef4417 100644
--- a/arch/mips/cavium-octeon/executive/cvmx-helper-board.c
+++ b/arch/mips/cavium-octeon/executive/cvmx-helper-board.c
@@ -767,7 +767,7 @@ enum cvmx_helper_board_usb_clock_types __cvmx_helper_board_usb_get_clock_type(vo
 		break;
 	}
 	/* Most boards except NIC10e use a 12MHz crystal */
-	if (OCTEON_IS_MODEL(OCTEON_FAM_2))
+	if (OCTEON_IS_OCTEON2())
 		return USB_CLOCK_TYPE_CRYSTAL_12;
 	return USB_CLOCK_TYPE_REF_48;
 }
diff --git a/arch/mips/cavium-octeon/octeon-irq.c b/arch/mips/cavium-octeon/octeon-irq.c
index 2bc4aa95944e..10f762557b92 100644
--- a/arch/mips/cavium-octeon/octeon-irq.c
+++ b/arch/mips/cavium-octeon/octeon-irq.c
@@ -3,12 +3,14 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2004-2012 Cavium, Inc.
+ * Copyright (C) 2004-2014 Cavium, Inc.
  */
 
+#include <linux/of_address.h>
 #include <linux/interrupt.h>
 #include <linux/irqdomain.h>
 #include <linux/bitops.h>
+#include <linux/of_irq.h>
 #include <linux/percpu.h>
 #include <linux/slab.h>
 #include <linux/irq.h>
@@ -22,16 +24,25 @@ static DEFINE_PER_CPU(unsigned long, octeon_irq_ciu0_en_mirror);
 static DEFINE_PER_CPU(unsigned long, octeon_irq_ciu1_en_mirror);
 static DEFINE_PER_CPU(raw_spinlock_t, octeon_irq_ciu_spinlock);
 
+struct octeon_irq_ciu_domain_data {
+	int num_sum;  /* number of sum registers (2 or 3). */
+};
+
 static __read_mostly u8 octeon_irq_ciu_to_irq[8][64];
 
-union octeon_ciu_chip_data {
-	void *p;
-	unsigned long l;
-	struct {
-		unsigned long line:6;
-		unsigned long bit:6;
-		unsigned long gpio_line:6;
-	} s;
+struct octeon_ciu_chip_data {
+	union {
+		struct {		/* only used for ciu3 */
+			u64 ciu3_addr;
+			unsigned int intsn;
+		};
+		struct {		/* only used for ciu/ciu2 */
+			u8 line;
+			u8 bit;
+			u8 gpio_line;
+		};
+	};
+	int current_cpu;	/* Next CPU expected to take this irq */
 };
 
 struct octeon_core_chip_data {
@@ -45,27 +56,40 @@ struct octeon_core_chip_data {
 
 static struct octeon_core_chip_data octeon_irq_core_chip_data[MIPS_CORE_IRQ_LINES];
 
-static void octeon_irq_set_ciu_mapping(int irq, int line, int bit, int gpio_line,
-				       struct irq_chip *chip,
-				       irq_flow_handler_t handler)
+static int octeon_irq_set_ciu_mapping(int irq, int line, int bit, int gpio_line,
+				      struct irq_chip *chip,
+				      irq_flow_handler_t handler)
 {
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
+
+	cd = kzalloc(sizeof(*cd), GFP_KERNEL);
+	if (!cd)
+		return -ENOMEM;
 
 	irq_set_chip_and_handler(irq, chip, handler);
 
-	cd.l = 0;
-	cd.s.line = line;
-	cd.s.bit = bit;
-	cd.s.gpio_line = gpio_line;
+	cd->line = line;
+	cd->bit = bit;
+	cd->gpio_line = gpio_line;
 
-	irq_set_chip_data(irq, cd.p);
+	irq_set_chip_data(irq, cd);
 	octeon_irq_ciu_to_irq[line][bit] = irq;
+	return 0;
 }
 
-static void octeon_irq_force_ciu_mapping(struct irq_domain *domain,
-					 int irq, int line, int bit)
+static void octeon_irq_free_cd(struct irq_domain *d, unsigned int irq)
 {
-	irq_domain_associate(domain, irq, line << 6 | bit);
+	struct irq_data *data = irq_get_irq_data(irq);
+	struct octeon_ciu_chip_data *cd = irq_data_get_irq_chip_data(data);
+
+	irq_set_chip_data(irq, NULL);
+	kfree(cd);
+}
+
+static int octeon_irq_force_ciu_mapping(struct irq_domain *domain,
+					int irq, int line, int bit)
+{
+	return irq_domain_associate(domain, irq, line << 6 | bit);
 }
 
 static int octeon_coreid_for_cpu(int cpu)
@@ -202,9 +226,10 @@ static int next_cpu_for_irq(struct irq_data *data)
 #ifdef CONFIG_SMP
 	int cpu;
 	int weight = cpumask_weight(data->affinity);
+	struct octeon_ciu_chip_data *cd = irq_data_get_irq_chip_data(data);
 
 	if (weight > 1) {
-		cpu = smp_processor_id();
+		cpu = cd->current_cpu;
 		for (;;) {
 			cpu = cpumask_next(cpu, data->affinity);
 			if (cpu >= nr_cpu_ids) {
@@ -219,6 +244,7 @@ static int next_cpu_for_irq(struct irq_data *data)
 	} else {
 		cpu = smp_processor_id();
 	}
+	cd->current_cpu = cpu;
 	return cpu;
 #else
 	return smp_processor_id();
@@ -231,15 +257,15 @@ static void octeon_irq_ciu_enable(struct irq_data *data)
 	int coreid = octeon_coreid_for_cpu(cpu);
 	unsigned long *pen;
 	unsigned long flags;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	raw_spinlock_t *lock = &per_cpu(octeon_irq_ciu_spinlock, cpu);
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	raw_spin_lock_irqsave(lock, flags);
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		pen = &per_cpu(octeon_irq_ciu0_en_mirror, cpu);
-		__set_bit(cd.s.bit, pen);
+		__set_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
@@ -248,7 +274,7 @@ static void octeon_irq_ciu_enable(struct irq_data *data)
 		cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), *pen);
 	} else {
 		pen = &per_cpu(octeon_irq_ciu1_en_mirror, cpu);
-		__set_bit(cd.s.bit, pen);
+		__set_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
@@ -263,15 +289,15 @@ static void octeon_irq_ciu_enable_local(struct irq_data *data)
 {
 	unsigned long *pen;
 	unsigned long flags;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	raw_spinlock_t *lock = this_cpu_ptr(&octeon_irq_ciu_spinlock);
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	raw_spin_lock_irqsave(lock, flags);
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		pen = this_cpu_ptr(&octeon_irq_ciu0_en_mirror);
-		__set_bit(cd.s.bit, pen);
+		__set_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
@@ -280,7 +306,7 @@ static void octeon_irq_ciu_enable_local(struct irq_data *data)
 		cvmx_write_csr(CVMX_CIU_INTX_EN0(cvmx_get_core_num() * 2), *pen);
 	} else {
 		pen = this_cpu_ptr(&octeon_irq_ciu1_en_mirror);
-		__set_bit(cd.s.bit, pen);
+		__set_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
@@ -295,15 +321,15 @@ static void octeon_irq_ciu_disable_local(struct irq_data *data)
 {
 	unsigned long *pen;
 	unsigned long flags;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	raw_spinlock_t *lock = this_cpu_ptr(&octeon_irq_ciu_spinlock);
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	raw_spin_lock_irqsave(lock, flags);
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		pen = this_cpu_ptr(&octeon_irq_ciu0_en_mirror);
-		__clear_bit(cd.s.bit, pen);
+		__clear_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
@@ -312,7 +338,7 @@ static void octeon_irq_ciu_disable_local(struct irq_data *data)
 		cvmx_write_csr(CVMX_CIU_INTX_EN0(cvmx_get_core_num() * 2), *pen);
 	} else {
 		pen = this_cpu_ptr(&octeon_irq_ciu1_en_mirror);
-		__clear_bit(cd.s.bit, pen);
+		__clear_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
@@ -328,27 +354,27 @@ static void octeon_irq_ciu_disable_all(struct irq_data *data)
 	unsigned long flags;
 	unsigned long *pen;
 	int cpu;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	raw_spinlock_t *lock;
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	for_each_online_cpu(cpu) {
 		int coreid = octeon_coreid_for_cpu(cpu);
 		lock = &per_cpu(octeon_irq_ciu_spinlock, cpu);
-		if (cd.s.line == 0)
+		if (cd->line == 0)
 			pen = &per_cpu(octeon_irq_ciu0_en_mirror, cpu);
 		else
 			pen = &per_cpu(octeon_irq_ciu1_en_mirror, cpu);
 
 		raw_spin_lock_irqsave(lock, flags);
-		__clear_bit(cd.s.bit, pen);
+		__clear_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
 		 */
 		wmb();
-		if (cd.s.line == 0)
+		if (cd->line == 0)
 			cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), *pen);
 		else
 			cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), *pen);
@@ -361,27 +387,27 @@ static void octeon_irq_ciu_enable_all(struct irq_data *data)
 	unsigned long flags;
 	unsigned long *pen;
 	int cpu;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	raw_spinlock_t *lock;
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	for_each_online_cpu(cpu) {
 		int coreid = octeon_coreid_for_cpu(cpu);
 		lock = &per_cpu(octeon_irq_ciu_spinlock, cpu);
-		if (cd.s.line == 0)
+		if (cd->line == 0)
 			pen = &per_cpu(octeon_irq_ciu0_en_mirror, cpu);
 		else
 			pen = &per_cpu(octeon_irq_ciu1_en_mirror, cpu);
 
 		raw_spin_lock_irqsave(lock, flags);
-		__set_bit(cd.s.bit, pen);
+		__set_bit(cd->bit, pen);
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
 		 * enabling the irq.
 		 */
 		wmb();
-		if (cd.s.line == 0)
+		if (cd->line == 0)
 			cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), *pen);
 		else
 			cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), *pen);
@@ -397,45 +423,106 @@ static void octeon_irq_ciu_enable_v2(struct irq_data *data)
 {
 	u64 mask;
 	int cpu = next_cpu_for_irq(data);
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.bit);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
 
 	/*
 	 * Called under the desc lock, so these should never get out
 	 * of sync.
 	 */
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		int index = octeon_coreid_for_cpu(cpu) * 2;
-		set_bit(cd.s.bit, &per_cpu(octeon_irq_ciu0_en_mirror, cpu));
+		set_bit(cd->bit, &per_cpu(octeon_irq_ciu0_en_mirror, cpu));
 		cvmx_write_csr(CVMX_CIU_INTX_EN0_W1S(index), mask);
 	} else {
 		int index = octeon_coreid_for_cpu(cpu) * 2 + 1;
-		set_bit(cd.s.bit, &per_cpu(octeon_irq_ciu1_en_mirror, cpu));
+		set_bit(cd->bit, &per_cpu(octeon_irq_ciu1_en_mirror, cpu));
 		cvmx_write_csr(CVMX_CIU_INTX_EN1_W1S(index), mask);
 	}
 }
 
 /*
+ * Enable the irq in the sum2 registers.
+ */
+static void octeon_irq_ciu_enable_sum2(struct irq_data *data)
+{
+	u64 mask;
+	int cpu = next_cpu_for_irq(data);
+	int index = octeon_coreid_for_cpu(cpu);
+	struct octeon_ciu_chip_data *cd;
+
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
+
+	cvmx_write_csr(CVMX_CIU_EN2_PPX_IP4_W1S(index), mask);
+}
+
+/*
+ * Disable the irq in the sum2 registers.
+ */
+static void octeon_irq_ciu_disable_local_sum2(struct irq_data *data)
+{
+	u64 mask;
+	int cpu = next_cpu_for_irq(data);
+	int index = octeon_coreid_for_cpu(cpu);
+	struct octeon_ciu_chip_data *cd;
+
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
+
+	cvmx_write_csr(CVMX_CIU_EN2_PPX_IP4_W1C(index), mask);
+}
+
+static void octeon_irq_ciu_ack_sum2(struct irq_data *data)
+{
+	u64 mask;
+	int cpu = next_cpu_for_irq(data);
+	int index = octeon_coreid_for_cpu(cpu);
+	struct octeon_ciu_chip_data *cd;
+
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
+
+	cvmx_write_csr(CVMX_CIU_SUM2_PPX_IP4(index), mask);
+}
+
+static void octeon_irq_ciu_disable_all_sum2(struct irq_data *data)
+{
+	int cpu;
+	struct octeon_ciu_chip_data *cd;
+	u64 mask;
+
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
+
+	for_each_online_cpu(cpu) {
+		int coreid = octeon_coreid_for_cpu(cpu);
+
+		cvmx_write_csr(CVMX_CIU_EN2_PPX_IP4_W1C(coreid), mask);
+	}
+}
+
+/*
  * Enable the irq on the current CPU for chips that
  * have the EN*_W1{S,C} registers.
  */
 static void octeon_irq_ciu_enable_local_v2(struct irq_data *data)
 {
 	u64 mask;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.bit);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
 
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		int index = cvmx_get_core_num() * 2;
-		set_bit(cd.s.bit, this_cpu_ptr(&octeon_irq_ciu0_en_mirror));
+		set_bit(cd->bit, this_cpu_ptr(&octeon_irq_ciu0_en_mirror));
 		cvmx_write_csr(CVMX_CIU_INTX_EN0_W1S(index), mask);
 	} else {
 		int index = cvmx_get_core_num() * 2 + 1;
-		set_bit(cd.s.bit, this_cpu_ptr(&octeon_irq_ciu1_en_mirror));
+		set_bit(cd->bit, this_cpu_ptr(&octeon_irq_ciu1_en_mirror));
 		cvmx_write_csr(CVMX_CIU_INTX_EN1_W1S(index), mask);
 	}
 }
@@ -443,18 +530,18 @@ static void octeon_irq_ciu_enable_local_v2(struct irq_data *data)
 static void octeon_irq_ciu_disable_local_v2(struct irq_data *data)
 {
 	u64 mask;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.bit);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
 
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		int index = cvmx_get_core_num() * 2;
-		clear_bit(cd.s.bit, this_cpu_ptr(&octeon_irq_ciu0_en_mirror));
+		clear_bit(cd->bit, this_cpu_ptr(&octeon_irq_ciu0_en_mirror));
 		cvmx_write_csr(CVMX_CIU_INTX_EN0_W1C(index), mask);
 	} else {
 		int index = cvmx_get_core_num() * 2 + 1;
-		clear_bit(cd.s.bit, this_cpu_ptr(&octeon_irq_ciu1_en_mirror));
+		clear_bit(cd->bit, this_cpu_ptr(&octeon_irq_ciu1_en_mirror));
 		cvmx_write_csr(CVMX_CIU_INTX_EN1_W1C(index), mask);
 	}
 }
@@ -465,12 +552,12 @@ static void octeon_irq_ciu_disable_local_v2(struct irq_data *data)
 static void octeon_irq_ciu_ack(struct irq_data *data)
 {
 	u64 mask;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.bit);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
 
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		int index = cvmx_get_core_num() * 2;
 		cvmx_write_csr(CVMX_CIU_INTX_SUM0(index), mask);
 	} else {
@@ -486,21 +573,23 @@ static void octeon_irq_ciu_disable_all_v2(struct irq_data *data)
 {
 	int cpu;
 	u64 mask;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.bit);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
 
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		for_each_online_cpu(cpu) {
 			int index = octeon_coreid_for_cpu(cpu) * 2;
-			clear_bit(cd.s.bit, &per_cpu(octeon_irq_ciu0_en_mirror, cpu));
+			clear_bit(cd->bit,
+				&per_cpu(octeon_irq_ciu0_en_mirror, cpu));
 			cvmx_write_csr(CVMX_CIU_INTX_EN0_W1C(index), mask);
 		}
 	} else {
 		for_each_online_cpu(cpu) {
 			int index = octeon_coreid_for_cpu(cpu) * 2 + 1;
-			clear_bit(cd.s.bit, &per_cpu(octeon_irq_ciu1_en_mirror, cpu));
+			clear_bit(cd->bit,
+				&per_cpu(octeon_irq_ciu1_en_mirror, cpu));
 			cvmx_write_csr(CVMX_CIU_INTX_EN1_W1C(index), mask);
 		}
 	}
@@ -514,21 +603,23 @@ static void octeon_irq_ciu_enable_all_v2(struct irq_data *data)
 {
 	int cpu;
 	u64 mask;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.bit);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->bit);
 
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		for_each_online_cpu(cpu) {
 			int index = octeon_coreid_for_cpu(cpu) * 2;
-			set_bit(cd.s.bit, &per_cpu(octeon_irq_ciu0_en_mirror, cpu));
+			set_bit(cd->bit,
+				&per_cpu(octeon_irq_ciu0_en_mirror, cpu));
 			cvmx_write_csr(CVMX_CIU_INTX_EN0_W1S(index), mask);
 		}
 	} else {
 		for_each_online_cpu(cpu) {
 			int index = octeon_coreid_for_cpu(cpu) * 2 + 1;
-			set_bit(cd.s.bit, &per_cpu(octeon_irq_ciu1_en_mirror, cpu));
+			set_bit(cd->bit,
+				&per_cpu(octeon_irq_ciu1_en_mirror, cpu));
 			cvmx_write_csr(CVMX_CIU_INTX_EN1_W1S(index), mask);
 		}
 	}
@@ -537,10 +628,10 @@ static void octeon_irq_ciu_enable_all_v2(struct irq_data *data)
 static void octeon_irq_gpio_setup(struct irq_data *data)
 {
 	union cvmx_gpio_bit_cfgx cfg;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	u32 t = irqd_get_trigger_type(data);
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	cfg.u64 = 0;
 	cfg.s.int_en = 1;
@@ -551,7 +642,7 @@ static void octeon_irq_gpio_setup(struct irq_data *data)
 	cfg.s.fil_cnt = 7;
 	cfg.s.fil_sel = 3;
 
-	cvmx_write_csr(CVMX_GPIO_BIT_CFGX(cd.s.gpio_line), cfg.u64);
+	cvmx_write_csr(CVMX_GPIO_BIT_CFGX(cd->gpio_line), cfg.u64);
 }
 
 static void octeon_irq_ciu_enable_gpio_v2(struct irq_data *data)
@@ -576,36 +667,36 @@ static int octeon_irq_ciu_gpio_set_type(struct irq_data *data, unsigned int t)
 
 static void octeon_irq_ciu_disable_gpio_v2(struct irq_data *data)
 {
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	cvmx_write_csr(CVMX_GPIO_BIT_CFGX(cd.s.gpio_line), 0);
+	cd = irq_data_get_irq_chip_data(data);
+	cvmx_write_csr(CVMX_GPIO_BIT_CFGX(cd->gpio_line), 0);
 
 	octeon_irq_ciu_disable_all_v2(data);
 }
 
 static void octeon_irq_ciu_disable_gpio(struct irq_data *data)
 {
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	cvmx_write_csr(CVMX_GPIO_BIT_CFGX(cd.s.gpio_line), 0);
+	cd = irq_data_get_irq_chip_data(data);
+	cvmx_write_csr(CVMX_GPIO_BIT_CFGX(cd->gpio_line), 0);
 
 	octeon_irq_ciu_disable_all(data);
 }
 
 static void octeon_irq_ciu_gpio_ack(struct irq_data *data)
 {
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	u64 mask;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << (cd.s.gpio_line);
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << (cd->gpio_line);
 
 	cvmx_write_csr(CVMX_GPIO_INT_CLR, mask);
 }
 
-static void octeon_irq_handle_gpio(unsigned int irq, struct irq_desc *desc)
+static void octeon_irq_handle_trigger(unsigned int irq, struct irq_desc *desc)
 {
 	if (irq_get_trigger_type(irq) & IRQ_TYPE_EDGE_BOTH)
 		handle_edge_irq(irq, desc);
@@ -644,11 +735,11 @@ static int octeon_irq_ciu_set_affinity(struct irq_data *data,
 	int cpu;
 	bool enable_one = !irqd_irq_disabled(data) && !irqd_irq_masked(data);
 	unsigned long flags;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 	unsigned long *pen;
 	raw_spinlock_t *lock;
 
-	cd.p = irq_data_get_irq_chip_data(data);
+	cd = irq_data_get_irq_chip_data(data);
 
 	/*
 	 * For non-v2 CIU, we will allow only single CPU affinity.
@@ -668,16 +759,16 @@ static int octeon_irq_ciu_set_affinity(struct irq_data *data,
 		lock = &per_cpu(octeon_irq_ciu_spinlock, cpu);
 		raw_spin_lock_irqsave(lock, flags);
 
-		if (cd.s.line == 0)
+		if (cd->line == 0)
 			pen = &per_cpu(octeon_irq_ciu0_en_mirror, cpu);
 		else
 			pen = &per_cpu(octeon_irq_ciu1_en_mirror, cpu);
 
 		if (cpumask_test_cpu(cpu, dest) && enable_one) {
 			enable_one = 0;
-			__set_bit(cd.s.bit, pen);
+			__set_bit(cd->bit, pen);
 		} else {
-			__clear_bit(cd.s.bit, pen);
+			__clear_bit(cd->bit, pen);
 		}
 		/*
 		 * Must be visible to octeon_irq_ip{2,3}_ciu() before
@@ -685,7 +776,7 @@ static int octeon_irq_ciu_set_affinity(struct irq_data *data,
 		 */
 		wmb();
 
-		if (cd.s.line == 0)
+		if (cd->line == 0)
 			cvmx_write_csr(CVMX_CIU_INTX_EN0(coreid * 2), *pen);
 		else
 			cvmx_write_csr(CVMX_CIU_INTX_EN1(coreid * 2 + 1), *pen);
@@ -706,24 +797,24 @@ static int octeon_irq_ciu_set_affinity_v2(struct irq_data *data,
 	int cpu;
 	bool enable_one = !irqd_irq_disabled(data) && !irqd_irq_masked(data);
 	u64 mask;
-	union octeon_ciu_chip_data cd;
+	struct octeon_ciu_chip_data *cd;
 
 	if (!enable_one)
 		return 0;
 
-	cd.p = irq_data_get_irq_chip_data(data);
-	mask = 1ull << cd.s.bit;
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << cd->bit;
 
-	if (cd.s.line == 0) {
+	if (cd->line == 0) {
 		for_each_online_cpu(cpu) {
 			unsigned long *pen = &per_cpu(octeon_irq_ciu0_en_mirror, cpu);
 			int index = octeon_coreid_for_cpu(cpu) * 2;
 			if (cpumask_test_cpu(cpu, dest) && enable_one) {
 				enable_one = false;
-				set_bit(cd.s.bit, pen);
+				set_bit(cd->bit, pen);
 				cvmx_write_csr(CVMX_CIU_INTX_EN0_W1S(index), mask);
 			} else {
-				clear_bit(cd.s.bit, pen);
+				clear_bit(cd->bit, pen);
 				cvmx_write_csr(CVMX_CIU_INTX_EN0_W1C(index), mask);
 			}
 		}
@@ -733,16 +824,44 @@ static int octeon_irq_ciu_set_affinity_v2(struct irq_data *data,
 			int index = octeon_coreid_for_cpu(cpu) * 2 + 1;
 			if (cpumask_test_cpu(cpu, dest) && enable_one) {
 				enable_one = false;
-				set_bit(cd.s.bit, pen);
+				set_bit(cd->bit, pen);
 				cvmx_write_csr(CVMX_CIU_INTX_EN1_W1S(index), mask);
 			} else {
-				clear_bit(cd.s.bit, pen);
+				clear_bit(cd->bit, pen);
 				cvmx_write_csr(CVMX_CIU_INTX_EN1_W1C(index), mask);
 			}
 		}
 	}
 	return 0;
 }
+
+static int octeon_irq_ciu_set_affinity_sum2(struct irq_data *data,
+					    const struct cpumask *dest,
+					    bool force)
+{
+	int cpu;
+	bool enable_one = !irqd_irq_disabled(data) && !irqd_irq_masked(data);
+	u64 mask;
+	struct octeon_ciu_chip_data *cd;
+
+	if (!enable_one)
+		return 0;
+
+	cd = irq_data_get_irq_chip_data(data);
+	mask = 1ull << cd->bit;
+
+	for_each_online_cpu(cpu) {
+		int index = octeon_coreid_for_cpu(cpu);
+
+		if (cpumask_test_cpu(cpu, dest) && enable_one) {
+			enable_one = false;
+			cvmx_write_csr(CVMX_CIU_EN2_PPX_IP4_W1S(index), mask);
+		} else {
+			cvmx_write_csr(CVMX_CIU_EN2_PPX_IP4_W1C(index), mask);
+		}
+	}
+	return 0;
+}
 #endif
 
 /*
@@ -752,6 +871,18 @@ static struct irq_chip octeon_irq_chip_ciu_v2 = {
 	.name = "CIU",
 	.irq_enable = octeon_irq_ciu_enable_v2,
 	.irq_disable = octeon_irq_ciu_disable_all_v2,
+	.irq_mask = octeon_irq_ciu_disable_local_v2,
+	.irq_unmask = octeon_irq_ciu_enable_v2,
+#ifdef CONFIG_SMP
+	.irq_set_affinity = octeon_irq_ciu_set_affinity_v2,
+	.irq_cpu_offline = octeon_irq_cpu_offline_ciu,
+#endif
+};
+
+static struct irq_chip octeon_irq_chip_ciu_v2_edge = {
+	.name = "CIU",
+	.irq_enable = octeon_irq_ciu_enable_v2,
+	.irq_disable = octeon_irq_ciu_disable_all_v2,
 	.irq_ack = octeon_irq_ciu_ack,
 	.irq_mask = octeon_irq_ciu_disable_local_v2,
 	.irq_unmask = octeon_irq_ciu_enable_v2,
@@ -761,10 +892,50 @@ static struct irq_chip octeon_irq_chip_ciu_v2 = {
 #endif
 };
 
+/*
+ * Newer octeon chips have support for lockless CIU operation.
+ */
+static struct irq_chip octeon_irq_chip_ciu_sum2 = {
+	.name = "CIU",
+	.irq_enable = octeon_irq_ciu_enable_sum2,
+	.irq_disable = octeon_irq_ciu_disable_all_sum2,
+	.irq_mask = octeon_irq_ciu_disable_local_sum2,
+	.irq_unmask = octeon_irq_ciu_enable_sum2,
+#ifdef CONFIG_SMP
+	.irq_set_affinity = octeon_irq_ciu_set_affinity_sum2,
+	.irq_cpu_offline = octeon_irq_cpu_offline_ciu,
+#endif
+};
+
+static struct irq_chip octeon_irq_chip_ciu_sum2_edge = {
+	.name = "CIU",
+	.irq_enable = octeon_irq_ciu_enable_sum2,
+	.irq_disable = octeon_irq_ciu_disable_all_sum2,
+	.irq_ack = octeon_irq_ciu_ack_sum2,
+	.irq_mask = octeon_irq_ciu_disable_local_sum2,
+	.irq_unmask = octeon_irq_ciu_enable_sum2,
+#ifdef CONFIG_SMP
+	.irq_set_affinity = octeon_irq_ciu_set_affinity_sum2,
+	.irq_cpu_offline = octeon_irq_cpu_offline_ciu,
+#endif
+};
+
 static struct irq_chip octeon_irq_chip_ciu = {
 	.name = "CIU",
 	.irq_enable = octeon_irq_ciu_enable,
 	.irq_disable = octeon_irq_ciu_disable_all,
+	.irq_mask = octeon_irq_ciu_disable_local,
+	.irq_unmask = octeon_irq_ciu_enable,
+#ifdef CONFIG_SMP
+	.irq_set_affinity = octeon_irq_ciu_set_affinity,
+	.irq_cpu_offline = octeon_irq_cpu_offline_ciu,
+#endif
+};
+
+static struct irq_chip octeon_irq_chip_ciu_edge = {
+	.name = "CIU",
+	.irq_enable = octeon_irq_ciu_enable,
+	.irq_disable = octeon_irq_ciu_disable_all,
 	.irq_ack = octeon_irq_ciu_ack,
 	.irq_mask = octeon_irq_ciu_disable_local,
 	.irq_unmask = octeon_irq_ciu_enable,
@@ -970,11 +1141,12 @@ static int octeon_irq_ciu_xlat(struct irq_domain *d,
 			       unsigned int *out_type)
 {
 	unsigned int ciu, bit;
+	struct octeon_irq_ciu_domain_data *dd = d->host_data;
 
 	ciu = intspec[0];
 	bit = intspec[1];
 
-	if (ciu > 1 || bit > 63)
+	if (ciu >= dd->num_sum || bit > 63)
 		return -EINVAL;
 
 	*out_hwirq = (ciu << 6) | bit;
@@ -984,6 +1156,7 @@ static int octeon_irq_ciu_xlat(struct irq_domain *d,
 }
 
 static struct irq_chip *octeon_irq_ciu_chip;
+static struct irq_chip *octeon_irq_ciu_chip_edge;
 static struct irq_chip *octeon_irq_gpio_chip;
 
 static bool octeon_irq_virq_in_range(unsigned int virq)
@@ -999,8 +1172,10 @@ static bool octeon_irq_virq_in_range(unsigned int virq)
 static int octeon_irq_ciu_map(struct irq_domain *d,
 			      unsigned int virq, irq_hw_number_t hw)
 {
+	int rv;
 	unsigned int line = hw >> 6;
 	unsigned int bit = hw & 63;
+	struct octeon_irq_ciu_domain_data *dd = d->host_data;
 
 	if (!octeon_irq_virq_in_range(virq))
 		return -EINVAL;
@@ -1009,54 +1184,61 @@ static int octeon_irq_ciu_map(struct irq_domain *d,
 	if (line == 0 && bit >= 16 && bit <32)
 		return 0;
 
-	if (line > 1 || octeon_irq_ciu_to_irq[line][bit] != 0)
+	if (line >= dd->num_sum || octeon_irq_ciu_to_irq[line][bit] != 0)
 		return -EINVAL;
 
-	if (octeon_irq_ciu_is_edge(line, bit))
-		octeon_irq_set_ciu_mapping(virq, line, bit, 0,
-					   octeon_irq_ciu_chip,
-					   handle_edge_irq);
-	else
-		octeon_irq_set_ciu_mapping(virq, line, bit, 0,
-					   octeon_irq_ciu_chip,
-					   handle_level_irq);
-
-	return 0;
+	if (line == 2) {
+		if (octeon_irq_ciu_is_edge(line, bit))
+			rv = octeon_irq_set_ciu_mapping(virq, line, bit, 0,
+				&octeon_irq_chip_ciu_sum2_edge,
+				handle_edge_irq);
+		else
+			rv = octeon_irq_set_ciu_mapping(virq, line, bit, 0,
+				&octeon_irq_chip_ciu_sum2,
+				handle_level_irq);
+	} else {
+		if (octeon_irq_ciu_is_edge(line, bit))
+			rv = octeon_irq_set_ciu_mapping(virq, line, bit, 0,
+				octeon_irq_ciu_chip_edge,
+				handle_edge_irq);
+		else
+			rv = octeon_irq_set_ciu_mapping(virq, line, bit, 0,
+				octeon_irq_ciu_chip,
+				handle_level_irq);
+	}
+	return rv;
 }
 
-static int octeon_irq_gpio_map_common(struct irq_domain *d,
-				      unsigned int virq, irq_hw_number_t hw,
-				      int line_limit, struct irq_chip *chip)
+static int octeon_irq_gpio_map(struct irq_domain *d,
+			       unsigned int virq, irq_hw_number_t hw)
 {
 	struct octeon_irq_gpio_domain_data *gpiod = d->host_data;
 	unsigned int line, bit;
+	int r;
 
 	if (!octeon_irq_virq_in_range(virq))
 		return -EINVAL;
 
 	line = (hw + gpiod->base_hwirq) >> 6;
 	bit = (hw + gpiod->base_hwirq) & 63;
-	if (line > line_limit || octeon_irq_ciu_to_irq[line][bit] != 0)
+	if (line > ARRAY_SIZE(octeon_irq_ciu_to_irq) ||
+		octeon_irq_ciu_to_irq[line][bit] != 0)
 		return -EINVAL;
 
-	octeon_irq_set_ciu_mapping(virq, line, bit, hw,
-				   chip, octeon_irq_handle_gpio);
-	return 0;
-}
-
-static int octeon_irq_gpio_map(struct irq_domain *d,
-			       unsigned int virq, irq_hw_number_t hw)
-{
-	return octeon_irq_gpio_map_common(d, virq, hw, 1, octeon_irq_gpio_chip);
+	r = octeon_irq_set_ciu_mapping(virq, line, bit, hw,
+		octeon_irq_gpio_chip, octeon_irq_handle_trigger);
+	return r;
 }
 
 static struct irq_domain_ops octeon_irq_domain_ciu_ops = {
 	.map = octeon_irq_ciu_map,
+	.unmap = octeon_irq_free_cd,
 	.xlate = octeon_irq_ciu_xlat,
 };
 
 static struct irq_domain_ops octeon_irq_domain_gpio_ops = {
 	.map = octeon_irq_gpio_map,
+	.unmap = octeon_irq_free_cd,
 	.xlate = octeon_irq_gpio_xlat,
 };
 
@@ -1095,6 +1277,26 @@ static void octeon_irq_ip3_ciu(void)
 	}
 }
 
+static void octeon_irq_ip4_ciu(void)
+{
+	int coreid = cvmx_get_core_num();
+	u64 ciu_sum = cvmx_read_csr(CVMX_CIU_SUM2_PPX_IP4(coreid));
+	u64 ciu_en = cvmx_read_csr(CVMX_CIU_EN2_PPX_IP4(coreid));
+
+	ciu_sum &= ciu_en;
+	if (likely(ciu_sum)) {
+		int bit = fls64(ciu_sum) - 1;
+		int irq = octeon_irq_ciu_to_irq[2][bit];
+
+		if (likely(irq))
+			do_IRQ(irq);
+		else
+			spurious_interrupt();
+	} else {
+		spurious_interrupt();
+	}
+}
+
 static bool octeon_irq_use_ip4;
 
 static void octeon_irq_local_enable_ip4(void *arg)
@@ -1176,7 +1378,10 @@ static void octeon_irq_setup_secondary_ciu(void)
 
 	/* Enable the CIU lines */
 	set_c0_status(STATUSF_IP3 | STATUSF_IP2);
-	clear_c0_status(STATUSF_IP4);
+	if (octeon_irq_use_ip4)
+		set_c0_status(STATUSF_IP4);
+	else
+		clear_c0_status(STATUSF_IP4);
 }
 
 static void octeon_irq_setup_secondary_ciu2(void)
@@ -1192,95 +1397,194 @@ static void octeon_irq_setup_secondary_ciu2(void)
 		clear_c0_status(STATUSF_IP4);
 }
 
-static void __init octeon_irq_init_ciu(void)
+static int __init octeon_irq_init_ciu(
+	struct device_node *ciu_node, struct device_node *parent)
 {
-	unsigned int i;
+	unsigned int i, r;
 	struct irq_chip *chip;
+	struct irq_chip *chip_edge;
 	struct irq_chip *chip_mbox;
 	struct irq_chip *chip_wd;
-	struct device_node *gpio_node;
-	struct device_node *ciu_node;
 	struct irq_domain *ciu_domain = NULL;
+	struct octeon_irq_ciu_domain_data *dd;
+
+	dd = kzalloc(sizeof(*dd), GFP_KERNEL);
+	if (!dd)
+		return -ENOMEM;
 </