[PATCH 01/15] powerpc/a2: Add some #defines for A2 specific instructions

[PATCH 01/15] powerpc/a2: Add some #defines for A2 specific instructions

[Michael Ellerman]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/ppc-opcode.h |   25 +++++++++++++++++++++++++
 1 files changed, 25 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/include/asm/ppc-opcode.h b/arch/powerpc/include/asm/ppc-opcode.h
index 384a96d..3e25b25 100644
--- a/arch/powerpc/include/asm/ppc-opcode.h
+++ b/arch/powerpc/include/asm/ppc-opcode.h
@@ -59,6 +59,14 @@
 #define PPC_INST_NAP			0x4c000364
 #define PPC_INST_SLEEP			0x4c0003a4
 
+/* A2 specific instructions */
+#define PPC_INST_ERATWE			0x7c0001a6
+#define PPC_INST_ERATRE			0x7c000166
+#define PPC_INST_ERATILX		0x7c000066
+#define PPC_INST_ERATIVAX		0x7c000666
+#define PPC_INST_ERATSX			0x7c000126
+#define PPC_INST_ERATSX_DOT		0x7c000127
+
 /* macros to insert fields into opcodes */
 #define __PPC_RA(a)	(((a) & 0x1f) << 16)
 #define __PPC_RB(b)	(((b) & 0x1f) << 11)
@@ -70,6 +78,8 @@
 #define __PPC_XT(s)	__PPC_XS(s)
 #define __PPC_T_TLB(t)	(((t) & 0x3) << 21)
 #define __PPC_WC(w)	(((w) & 0x3) << 21)
+#define __PPC_WS(w)	(((w) & 0x1f) << 11)
+
 /*
  * Only use the larx hint bit on 64bit CPUs. e500v1/v2 based CPUs will treat a
  * larx with EH set as an illegal instruction.
@@ -116,6 +126,21 @@
 #define PPC_TLBIVAX(a,b)	stringify_in_c(.long PPC_INST_TLBIVAX | \
 					__PPC_RA(a) | __PPC_RB(b))
 
+#define PPC_ERATWE(s, a, w)	stringify_in_c(.long PPC_INST_ERATWE | \
+					__PPC_RS(s) | __PPC_RA(a) | __PPC_WS(w))
+#define PPC_ERATRE(s, a, w)	stringify_in_c(.long PPC_INST_ERATRE | \
+					__PPC_RS(s) | __PPC_RA(a) | __PPC_WS(w))
+#define PPC_ERATILX(t, a, b)	stringify_in_c(.long PPC_INST_ERATILX | \
+					__PPC_T_TLB(t) | __PPC_RA(a) | \
+					__PPC_RB(b))
+#define PPC_ERATIVAX(s, a, b)	stringify_in_c(.long PPC_INST_ERATIVAX | \
+					__PPC_RS(s) | __PPC_RA(a) | __PPC_RB(b))
+#define PPC_ERATSX(t, a, w)	stringify_in_c(.long PPC_INST_ERATSX | \
+					__PPC_RS(t) | __PPC_RA(a) | __PPC_RB(b))
+#define PPC_ERATSX_DOT(t, a, w)	stringify_in_c(.long PPC_INST_ERATSX_DOT | \
+					__PPC_RS(t) | __PPC_RA(a) | __PPC_RB(b))
+
+
 /*
  * Define what the VSX XX1 form instructions will look like, then add
  * the 128 bit load store instructions based on that.
-- 
1.7.1

[PATCH 02/15] of: Export of_irq_find_parent()

[Michael Ellerman]

From: Michael Ellerman <michael@ellerman.id.au>

We have platform code that needs to find a node's interrupt parent, so
export of_irq_find_parent() so we can use it.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 drivers/of/irq.c       |    2 +-
 include/linux/of_irq.h |    1 +
 2 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/drivers/of/irq.c b/drivers/of/irq.c
index 75b0d3c..9f689f1 100644
--- a/drivers/of/irq.c
+++ b/drivers/of/irq.c
@@ -56,7 +56,7 @@ EXPORT_SYMBOL_GPL(irq_of_parse_and_map);
  * Returns a pointer to the interrupt parent node, or NULL if the interrupt
  * parent could not be determined.
  */
-static struct device_node *of_irq_find_parent(struct device_node *child)
+struct device_node *of_irq_find_parent(struct device_node *child)
 {
 	struct device_node *p;
 	const __be32 *parp;
diff --git a/include/linux/of_irq.h b/include/linux/of_irq.h
index 109e013..e6955f5 100644
--- a/include/linux/of_irq.h
+++ b/include/linux/of_irq.h
@@ -68,6 +68,7 @@ extern int of_irq_to_resource(struct device_node *dev, int index,
 extern int of_irq_count(struct device_node *dev);
 extern int of_irq_to_resource_table(struct device_node *dev,
 		struct resource *res, int nr_irqs);
+extern struct device_node *of_irq_find_parent(struct device_node *child);
 
 #endif /* CONFIG_OF_IRQ */
 #endif /* CONFIG_OF */
-- 
1.7.1

[PATCH 03/15] powerpc: Add SCOM infrastructure

[Michael Ellerman]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

SCOM is a side-band configuration bus implemented on some processors.
This code provides a way for code to map and operate on devices via
SCOM, while the details of how that is implemented is left up to a
SCOM "controller" in the platform code.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/scom.h |  156 +++++++++++++++++++++++++++++++
 arch/powerpc/sysdev/Kconfig     |    7 ++
 arch/powerpc/sysdev/Makefile    |    2 +
 arch/powerpc/sysdev/scom.c      |  192 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 357 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/include/asm/scom.h b/arch/powerpc/include/asm/scom.h
new file mode 100644
index 0000000..0cabfd7
--- /dev/null
+++ b/arch/powerpc/include/asm/scom.h
@@ -0,0 +1,156 @@
+/*
+ * Copyright 2010 Benjamin Herrenschmidt, IBM Corp
+ *                <benh@kernel.crashing.org>
+ *     and        David Gibson, IBM Corporation.
+ *
+ *   This program is free software;  you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation; either version 2 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ *   the GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program;  if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#ifndef _ASM_POWERPC_SCOM_H
+#define _ASM_POWERPC_SCOM_H
+
+#ifdef __KERNEL__
+#ifndef __ASSEMBLY__
+#ifdef CONFIG_PPC_SCOM
+
+/*
+ * The SCOM bus is a sideband bus used for accessing various internal
+ * registers of the processor or the chipset. The implementation details
+ * differ between processors and platforms, and the access method as
+ * well.
+ *
+ * This API allows to "map" ranges of SCOM register numbers associated
+ * with a given SCOM controller. The later must be represented by a
+ * device node, though some implementations might support NULL if there
+ * is no possible ambiguity
+ *
+ * Then, scom_read/scom_write can be used to accesses registers inside
+ * that range. The argument passed is a register number relative to
+ * the beginning of the range mapped.
+ */
+
+typedef void *scom_map_t;
+
+/* Value for an invalid SCOM map */
+#define SCOM_MAP_INVALID	(NULL)
+
+/* The scom_controller data structure is what the platform passes
+ * to the core code in scom_init, it provides the actual implementation
+ * of all the SCOM functions
+ */
+struct scom_controller {
+	scom_map_t (*map)(struct device_node *ctrl_dev, u64 reg, u64 count);
+	void (*unmap)(scom_map_t map);
+
+	u64 (*read)(scom_map_t map, u32 reg);
+	void (*write)(scom_map_t map, u32 reg, u64 value);
+};
+
+extern const struct scom_controller *scom_controller;
+
+/**
+ * scom_init - Initialize the SCOM backend, called by the platform
+ * @controller: The platform SCOM controller
+ */
+static inline void scom_init(const struct scom_controller *controller)
+{
+	scom_controller = controller;
+}
+
+/**
+ * scom_map_ok - Test is a SCOM mapping is successful
+ * @map: The result of scom_map to test
+ */
+static inline int scom_map_ok(scom_map_t map)
+{
+	return map != SCOM_MAP_INVALID;
+}
+
+/**
+ * scom_map - Map a block of SCOM registers
+ * @ctrl_dev: Device node of the SCOM controller
+ *            some implementations allow NULL here
+ * @reg: first SCOM register to map
+ * @count: Number of SCOM registers to map
+ */
+
+static inline scom_map_t scom_map(struct device_node *ctrl_dev,
+				  u64 reg, u64 count)
+{
+	return scom_controller->map(ctrl_dev, reg, count);
+}
+
+/**
+ * scom_find_parent - Find the SCOM controller for a device
+ * @dev: OF node of the device
+ *
+ * This is not meant for general usage, but in combination with
+ * scom_map() allows to map registers not represented by the
+ * device own scom-reg property. Useful for applying HW workarounds
+ * on things not properly represented in the device-tree for example.
+ */
+struct device_node *scom_find_parent(struct device_node *dev);
+
+
+/**
+ * scom_map_device - Map a device's block of SCOM registers
+ * @dev: OF node of the device
+ * @index: Register bank index (index in "scom-reg" property)
+ *
+ * This function will use the device-tree binding for SCOM which
+ * is to follow "scom-parent" properties until it finds a node with
+ * a "scom-controller" property to find the controller. It will then
+ * use the "scom-reg" property which is made of reg/count pairs,
+ * each of them having a size defined by the controller's #scom-cells
+ * property
+ */
+extern scom_map_t scom_map_device(struct device_node *dev, int index);
+
+
+/**
+ * scom_unmap - Unmap a block of SCOM registers
+ * @map: Result of scom_map is to be unmapped
+ */
+static inline void scom_unmap(scom_map_t map)
+{
+	if (scom_map_ok(map))
+		scom_controller->unmap(map);
+}
+
+/**
+ * scom_read - Read a SCOM register
+ * @map: Result of scom_map
+ * @reg: Register index within that map
+ */
+static inline u64 scom_read(scom_map_t map, u32 reg)
+{
+	return scom_controller->read(map, reg);
+}
+
+/**
+ * scom_write - Write to a SCOM register
+ * @map: Result of scom_map
+ * @reg: Register index within that map
+ * @value: Value to write
+ */
+static inline void scom_write(scom_map_t map, u32 reg, u64 value)
+{
+	scom_controller->write(map, reg, value);
+}
+
+#endif /* CONFIG_PPC_SCOM */
+#endif /* __ASSEMBLY__ */
+#endif /* __KERNEL__ */
+#endif /* _ASM_POWERPC_SCOM_H */
diff --git a/arch/powerpc/sysdev/Kconfig b/arch/powerpc/sysdev/Kconfig
index cfc1877..d775fd1 100644
--- a/arch/powerpc/sysdev/Kconfig
+++ b/arch/powerpc/sysdev/Kconfig
@@ -15,3 +15,10 @@ config PPC_MSI_BITMAP
 
 source "arch/powerpc/sysdev/xics/Kconfig"
 
+config PPC_SCOM
+	bool
+
+config SCOM_DEBUGFS
+	bool "Expose SCOM controllers via debugfs"
+	depends on PPC_SCOM
+	default n
diff --git a/arch/powerpc/sysdev/Makefile b/arch/powerpc/sysdev/Makefile
index 9516e75..6076e00 100644
--- a/arch/powerpc/sysdev/Makefile
+++ b/arch/powerpc/sysdev/Makefile
@@ -58,6 +58,8 @@ ifeq ($(CONFIG_SUSPEND),y)
 obj-$(CONFIG_6xx)		+= 6xx-suspend.o
 endif
 
+obj-$(CONFIG_PPC_SCOM)		+= scom.o
+
 subdir-ccflags-$(CONFIG_PPC_WERROR) := -Werror
 
 obj-$(CONFIG_PPC_XICS)		+= xics/
diff --git a/arch/powerpc/sysdev/scom.c b/arch/powerpc/sysdev/scom.c
new file mode 100644
index 0000000..b2593ce
--- /dev/null
+++ b/arch/powerpc/sysdev/scom.c
@@ -0,0 +1,192 @@
+/*
+ * Copyright 2010 Benjamin Herrenschmidt, IBM Corp
+ *                <benh@kernel.crashing.org>
+ *     and        David Gibson, IBM Corporation.
+ *
+ *   This program is free software;  you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation; either version 2 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ *   the GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program;  if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/debugfs.h>
+#include <linux/slab.h>
+#include <asm/prom.h>
+#include <asm/scom.h>
+
+const struct scom_controller *scom_controller;
+EXPORT_SYMBOL_GPL(scom_controller);
+
+struct device_node *scom_find_parent(struct device_node *node)
+{
+	struct device_node *par, *tmp;
+	const u32 *p;
+
+	for (par = of_node_get(node); par;) {
+		if (of_get_property(par, "scom-controller", NULL))
+			break;
+		p = of_get_property(par, "scom-parent", NULL);
+		tmp = par;
+		if (p == NULL)
+			par = of_get_parent(par);
+		else
+			par = of_find_node_by_phandle(*p);
+		of_node_put(tmp);
+	}
+	return par;
+}
+EXPORT_SYMBOL_GPL(scom_find_parent);
+
+scom_map_t scom_map_device(struct device_node *dev, int index)
+{
+	struct device_node *parent;
+	unsigned int cells, size;
+	const u32 *prop;
+	u64 reg, cnt;
+	scom_map_t ret;
+
+	parent = scom_find_parent(dev);
+
+	if (parent == NULL)
+		return 0;
+
+	prop = of_get_property(parent, "#scom-cells", NULL);
+	cells = prop ? *prop : 1;
+
+	prop = of_get_property(dev, "scom-reg", &size);
+	if (!prop)
+		return 0;
+	size >>= 2;
+
+	if (index >= (size / (2*cells)))
+		return 0;
+
+	reg = of_read_number(&prop[index * cells * 2], cells);
+	cnt = of_read_number(&prop[index * cells * 2 + cells], cells);
+
+	ret = scom_map(parent, reg, cnt);
+	of_node_put(parent);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(scom_map_device);
+
+#ifdef CONFIG_SCOM_DEBUGFS
+struct scom_debug_entry {
+	struct device_node *dn;
+	unsigned long addr;
+	scom_map_t map;
+	spinlock_t lock;
+	char name[8];
+	struct debugfs_blob_wrapper blob;
+};
+
+static int scom_addr_set(void *data, u64 val)
+{
+	struct scom_debug_entry *ent = data;
+
+	ent->addr = 0;
+	scom_unmap(ent->map);
+
+	ent->map = scom_map(ent->dn, val, 1);
+	if (scom_map_ok(ent->map))
+		ent->addr = val;
+	else
+		return -EFAULT;
+
+	return 0;
+}
+
+static int scom_addr_get(void *data, u64 *val)
+{
+	struct scom_debug_entry *ent = data;
+	*val = ent->addr;
+	return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(scom_addr_fops, scom_addr_get, scom_addr_set,
+			"0x%llx\n");
+
+static int scom_val_set(void *data, u64 val)
+{
+	struct scom_debug_entry *ent = data;
+
+	if (!scom_map_ok(ent->map))
+		return -EFAULT;
+
+	scom_write(ent->map, 0, val);
+
+	return 0;
+}
+
+static int scom_val_get(void *data, u64 *val)
+{
+	struct scom_debug_entry *ent = data;
+
+	if (!scom_map_ok(ent->map))
+		return -EFAULT;
+
+	*val = scom_read(ent->map, 0);
+	return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(scom_val_fops, scom_val_get, scom_val_set,
+			"0x%llx\n");
+
+static int scom_debug_init_one(struct dentry *root, struct device_node *dn,
+			       int i)
+{
+	struct scom_debug_entry *ent;
+	struct dentry *dir;
+
+	ent = kzalloc(sizeof(*ent), GFP_KERNEL);
+	if (!ent)
+		return -ENOMEM;
+
+	ent->dn = of_node_get(dn);
+	ent->map = SCOM_MAP_INVALID;
+	spin_lock_init(&ent->lock);
+	snprintf(ent->name, 8, "scom%d", i);
+	ent->blob.data = dn->full_name;
+	ent->blob.size = strlen(dn->full_name);
+
+	dir = debugfs_create_dir(ent->name, root);
+	if (!dir) {
+		of_node_put(dn);
+		kfree(ent);
+		return -1;
+	}
+
+	debugfs_create_file("addr", 0600, dir, ent, &scom_addr_fops);
+	debugfs_create_file("value", 0600, dir, ent, &scom_val_fops);
+	debugfs_create_blob("path", 0400, dir, &ent->blob);
+
+	return 0;
+}
+
+static int scom_debug_init(void)
+{
+	struct device_node *dn;
+	struct dentry *root;
+	int i, rc;
+
+	root = debugfs_create_dir("scom", powerpc_debugfs_root);
+	if (!root)
+		return -1;
+
+	i = rc = 0;
+	for_each_node_with_property(dn, "scom-controller")
+		rc |= scom_debug_init_one(root, dn, i++);
+
+	return rc;
+}
+device_initcall(scom_debug_init);
+#endif /* CONFIG_SCOM_DEBUGFS */
-- 
1.7.1

[PATCH 04/15] powerpc/xics: xics.h relies on linux/interrupt.h

[Michael Ellerman]

From: Michael Ellerman <michael@ellerman.id.au>

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/xics.h |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/include/asm/xics.h b/arch/powerpc/include/asm/xics.h
index 146aad8..c4ed4c5 100644
--- a/arch/powerpc/include/asm/xics.h
+++ b/arch/powerpc/include/asm/xics.h
@@ -6,6 +6,8 @@
 #ifndef _XICS_H
 #define _XICS_H
 
+#include <linux/interrupt.h>
+
 #define XICS_IPI		2
 #define XICS_IRQ_SPURIOUS	0
 
-- 
1.7.1

[PATCH 05/15] powerpc/xics: Move irq_host matching into the ics backend

[Michael Ellerman]

From: Michael Ellerman <michael@ellerman.id.au>

An upcoming new ics backend will need to implement different matching
semantics to the current ones, which are essentially the RTAS ics
backends. So move the current match into the RTAS backend, and allow
other ics backends to override.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/xics.h        |    1 +
 arch/powerpc/sysdev/xics/ics-rtas.c    |   11 +++++++++++
 arch/powerpc/sysdev/xics/xics-common.c |   12 +++++++-----
 3 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/arch/powerpc/include/asm/xics.h b/arch/powerpc/include/asm/xics.h
index c4ed4c5..6c06306 100644
--- a/arch/powerpc/include/asm/xics.h
+++ b/arch/powerpc/include/asm/xics.h
@@ -59,6 +59,7 @@ struct ics {
 	int (*map)(struct ics *ics, unsigned int virq);
 	void (*mask_unknown)(struct ics *ics, unsigned long vec);
 	long (*get_server)(struct ics *ics, unsigned long vec);
+	int (*host_match)(struct ics *ics, struct device_node *node);
 	char data[];
 };
 
diff --git a/arch/powerpc/sysdev/xics/ics-rtas.c b/arch/powerpc/sysdev/xics/ics-rtas.c
index ff76268..dd713b0 100644
--- a/arch/powerpc/sysdev/xics/ics-rtas.c
+++ b/arch/powerpc/sysdev/xics/ics-rtas.c
@@ -26,12 +26,14 @@ static int ibm_int_off;
 static int ics_rtas_map(struct ics *ics, unsigned int virq);
 static void ics_rtas_mask_unknown(struct ics *ics, unsigned long vec);
 static long ics_rtas_get_server(struct ics *ics, unsigned long vec);
+static int ics_rtas_host_match(struct ics *ics, struct device_node *node);
 
 /* Only one global & state struct ics */
 static struct ics ics_rtas = {
 	.map		= ics_rtas_map,
 	.mask_unknown	= ics_rtas_mask_unknown,
 	.get_server	= ics_rtas_get_server,
+	.host_match	= ics_rtas_host_match,
 };
 
 static void ics_rtas_unmask_irq(struct irq_data *d)
@@ -202,6 +204,15 @@ static long ics_rtas_get_server(struct ics *ics, unsigned long vec)
 	return status[0];
 }
 
+static int ics_rtas_host_match(struct ics *ics, struct device_node *node)
+{
+	/* IBM machines have interrupt parents of various funky types for things
+	 * like vdevices, events, etc... The trick we use here is to match
+	 * everything here except the legacy 8259 which is compatible "chrp,iic"
+	 */
+	return !of_device_is_compatible(node, "chrp,iic");
+}
+
 int ics_rtas_init(void)
 {
 	ibm_get_xive = rtas_token("ibm,get-xive");
diff --git a/arch/powerpc/sysdev/xics/xics-common.c b/arch/powerpc/sysdev/xics/xics-common.c
index 7c9cba3..dd7579a 100644
--- a/arch/powerpc/sysdev/xics/xics-common.c
+++ b/arch/powerpc/sysdev/xics/xics-common.c
@@ -331,11 +331,13 @@ int xics_get_irq_server(unsigned int virq, const struct cpumask *cpumask,
 
 static int xics_host_match(struct irq_host *h, struct device_node *node)
 {
-	/* IBM machines have interrupt parents of various funky types for things
-	 * like vdevices, events, etc... The trick we use here is to match
-	 * everything here except the legacy 8259 which is compatible "chrp,iic"
-	 */
-	return !of_device_is_compatible(node, "chrp,iic");
+	struct ics *ics;
+
+	list_for_each_entry(ics, &ics_list, link)
+		if (ics->host_match(ics, node))
+			return 1;
+
+	return 0;
 }
 
 /* Dummies */
-- 
1.7.1

[PATCH 06/15] powerpc/nvram: Search for nvram using compatible

[Michael Ellerman]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

As well as searching for nodes with type = "nvram", search for nodes
that have compatible = "nvram". This can't be converted into a single
call to of_find_compatible_node() with a non-NULL type, because that
searches for a node that has _both_ type & compatible = "nvram".

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/sysdev/mmio_nvram.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/sysdev/mmio_nvram.c b/arch/powerpc/sysdev/mmio_nvram.c
index 2073242..ddc877a 100644
--- a/arch/powerpc/sysdev/mmio_nvram.c
+++ b/arch/powerpc/sysdev/mmio_nvram.c
@@ -115,6 +115,8 @@ int __init mmio_nvram_init(void)
 	int ret;
 
 	nvram_node = of_find_node_by_type(NULL, "nvram");
+	if (!nvram_node)
+		nvram_node = of_find_compatible_node(NULL, NULL, "nvram");
 	if (!nvram_node) {
 		printk(KERN_WARNING "nvram: no node found in device-tree\n");
 		return -ENODEV;
-- 
1.7.1

[PATCH 07/15] powerpc: Move CPU_FTRS_BASE_BOOK3E into cputable.h & update FTR masks

[Michael Ellerman]

From: Michael Ellerman <michael@ellerman.id.au>

Where they belong with all the others. Remove SMT which may not be
true for all BOOK3E parts.

Currently the FTRS_POSSIBLE & FTRS_ALWAYS are defined for 64 or 32 bit.
Now that we have BOOK3E we need to split it three ways, BOOK3S, BOOK3E,
and 32-bit.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/cputable.h |   17 +++++++++++++----
 arch/powerpc/kernel/cputable.c      |    3 ---
 2 files changed, 13 insertions(+), 7 deletions(-)

diff --git a/arch/powerpc/include/asm/cputable.h b/arch/powerpc/include/asm/cputable.h
index 7d2eb44..32a970d 100644
--- a/arch/powerpc/include/asm/cputable.h
+++ b/arch/powerpc/include/asm/cputable.h
@@ -435,12 +435,18 @@ extern const char *powerpc_base_platform;
 	    CPU_FTR_PURR | CPU_FTR_REAL_LE | CPU_FTR_NO_SLBIE_B)
 #define CPU_FTRS_COMPATIBLE	(CPU_FTR_USE_TB | CPU_FTR_PPCAS_ARCH_V2)
 
-#ifdef __powerpc64__
+#define CPU_FTRS_BASE_BOOK3E	(CPU_FTR_USE_TB | CPU_FTR_PPCAS_ARCH_V2 | \
+				 CPU_FTR_NODSISRALIGN | CPU_FTR_NOEXECUTE)
+
+#if defined(CONFIG_PPC_BOOK3S_64)
 #define CPU_FTRS_POSSIBLE	\
 	    (CPU_FTRS_POWER3 | CPU_FTRS_RS64 | CPU_FTRS_POWER4 |	\
 	    CPU_FTRS_PPC970 | CPU_FTRS_POWER5 | CPU_FTRS_POWER6 |	\
 	    CPU_FTRS_POWER7 | CPU_FTRS_CELL | CPU_FTRS_PA6T |		\
 	    CPU_FTR_1T_SEGMENT | CPU_FTR_VSX)
+#elif defined(CONFIG_PPC_BOOK3E_64)
+#define CPU_FTRS_POSSIBLE	\
+	    (CPU_FTRS_BASE_BOOK3E)
 #else
 enum {
 	CPU_FTRS_POSSIBLE =
@@ -477,13 +483,16 @@ enum {
 #endif
 	    0,
 };
-#endif /* __powerpc64__ */
+#endif /* defined(CONFIG_PPC_BOOK3S_64) */
 
-#ifdef __powerpc64__
+#if defined(CONFIG_PPC_BOOK3S_64)
 #define CPU_FTRS_ALWAYS		\
 	    (CPU_FTRS_POWER3 & CPU_FTRS_RS64 & CPU_FTRS_POWER4 &	\
 	    CPU_FTRS_PPC970 & CPU_FTRS_POWER5 & CPU_FTRS_POWER6 &	\
 	    CPU_FTRS_POWER7 & CPU_FTRS_CELL & CPU_FTRS_PA6T & CPU_FTRS_POSSIBLE)
+#elif defined(CONFIG_PPC_BOOK3E_64)
+#define CPU_FTRS_ALWAYS	\
+	    (CPU_FTRS_BASE_BOOK3E)
 #else
 enum {
 	CPU_FTRS_ALWAYS =
@@ -517,7 +526,7 @@ enum {
 #endif
 	    CPU_FTRS_POSSIBLE,
 };
-#endif /* __powerpc64__ */
+#endif /* defined(CONFIG_PPC_BOOK3S_64) */
 
 static inline int cpu_has_feature(unsigned long feature)
 {
diff --git a/arch/powerpc/kernel/cputable.c b/arch/powerpc/kernel/cputable.c
index 867ae30..70cbae1 100644
--- a/arch/powerpc/kernel/cputable.c
+++ b/arch/powerpc/kernel/cputable.c
@@ -2015,9 +2015,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
 	{	/* This is a default entry to get going, to be replaced by
 		 * a real one at some stage
 		 */
-#define CPU_FTRS_BASE_BOOK3E	(CPU_FTR_USE_TB | \
-	    CPU_FTR_PPCAS_ARCH_V2 | CPU_FTR_SMT | \
-	    CPU_FTR_NODSISRALIGN | CPU_FTR_NOEXECUTE)
 		.pvr_mask		= 0x00000000,
 		.pvr_value		= 0x00000000,
 		.cpu_name		= "Book3E",
-- 
1.7.1

Re: [PATCH 07/15] powerpc: Move CPU_FTRS_BASE_BOOK3E into cputable.h & update FTR masks

[Kumar Gala]

On Apr 15, 2011, at 3:32 AM, Michael Ellerman wrote:

> From: Michael Ellerman <michael@ellerman.id.au>
> 
> Where they belong with all the others. Remove SMT which may not be
> true for all BOOK3E parts.
> 
> Currently the FTRS_POSSIBLE & FTRS_ALWAYS are defined for 64 or 32 bit.
> Now that we have BOOK3E we need to split it three ways, BOOK3S, BOOK3E,
> and 32-bit.
> 
> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> ---
> arch/powerpc/include/asm/cputable.h |   17 +++++++++++++----
> arch/powerpc/kernel/cputable.c      |    3 ---
> 2 files changed, 13 insertions(+), 7 deletions(-)

will probably need some updating based on:

commit 11ed0db9f6c7811233632d2ab79c50c011b89902
Author: Kumar Gala <galak@kernel.crashing.org>
Date:   Wed Apr 6 00:11:06 2011 -0500

    powerpc/book3e: Fix CPU feature handling on 64-bit e5500
    
    The CPU_FTRS_POSSIBLE and CPU_FTRS_ALWAYS defines did not encompass
    e5500 CPU features when built for 64-bit.  This causes issues with
    cpu_has_feature() as it utilizes the POSSIBLE & ALWAYS defines as part
    of its check.
    
    Create a unique CPU_FTRS_E5500 (as its different from CPU_FTRS_E500MC),
    created a new group for 64-bit Book3e based CPUs and add CPU_FTRS_E5500
    to that group.
    
    Signed-off-by: Kumar Gala <galak@kernel.crashing.org>

- k

[PATCH 08/15] powerpc: Add A2 cpu support

[Michael Ellerman]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Add the cputable entry, regs and setup & restore entries for
the PPC A2.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/cputable.h    |    6 +-
 arch/powerpc/include/asm/reg_a2.h      |  156 ++++++++++++++++++++++++++++++++
 arch/powerpc/kernel/Makefile           |    1 +
 arch/powerpc/kernel/cpu_setup_a2.S     |  114 +++++++++++++++++++++++
 arch/powerpc/kernel/cputable.c         |   25 +++++-
 arch/powerpc/platforms/Kconfig.cputype |    4 +
 6 files changed, 302 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/cputable.h b/arch/powerpc/include/asm/cputable.h
index 32a970d..0dd4266 100644
--- a/arch/powerpc/include/asm/cputable.h
+++ b/arch/powerpc/include/asm/cputable.h
@@ -438,6 +438,8 @@ extern const char *powerpc_base_platform;
 #define CPU_FTRS_BASE_BOOK3E	(CPU_FTR_USE_TB | CPU_FTR_PPCAS_ARCH_V2 | \
 				 CPU_FTR_NODSISRALIGN | CPU_FTR_NOEXECUTE)
 
+#define CPU_FTRS_A2 (CPU_FTRS_BASE_BOOK3E | CPU_FTR_SMT | CPU_FTR_DBELL)
+
 #if defined(CONFIG_PPC_BOOK3S_64)
 #define CPU_FTRS_POSSIBLE	\
 	    (CPU_FTRS_POWER3 | CPU_FTRS_RS64 | CPU_FTRS_POWER4 |	\
@@ -446,7 +448,7 @@ extern const char *powerpc_base_platform;
 	    CPU_FTR_1T_SEGMENT | CPU_FTR_VSX)
 #elif defined(CONFIG_PPC_BOOK3E_64)
 #define CPU_FTRS_POSSIBLE	\
-	    (CPU_FTRS_BASE_BOOK3E)
+	    (CPU_FTRS_BASE_BOOK3E | CPU_FTRS_A2)
 #else
 enum {
 	CPU_FTRS_POSSIBLE =
@@ -492,7 +494,7 @@ enum {
 	    CPU_FTRS_POWER7 & CPU_FTRS_CELL & CPU_FTRS_PA6T & CPU_FTRS_POSSIBLE)
 #elif defined(CONFIG_PPC_BOOK3E_64)
 #define CPU_FTRS_ALWAYS	\
-	    (CPU_FTRS_BASE_BOOK3E)
+	    (CPU_FTRS_BASE_BOOK3E & CPU_FTRS_A2)
 #else
 enum {
 	CPU_FTRS_ALWAYS =
diff --git a/arch/powerpc/include/asm/reg_a2.h b/arch/powerpc/include/asm/reg_a2.h
new file mode 100644
index 0000000..3ba9c6f
--- /dev/null
+++ b/arch/powerpc/include/asm/reg_a2.h
@@ -0,0 +1,156 @@
+/*
+ *  Register definitions specific to the A2 core
+ *
+ *  Copyright (C) 2008 Ben. Herrenschmidt (benh@kernel.crashing.org), IBM Corp.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+
+#ifndef __ASM_POWERPC_REG_A2_H__
+#define __ASM_POWERPC_REG_A2_H__
+
+#define SPRN_TENSR	0x1b5
+#define SPRN_TENS	0x1b6	/* Thread ENable Set */
+#define SPRN_TENC	0x1b7	/* Thread ENable Clear */
+
+#define SPRN_A2_CCR0	0x3f0	/* Core Configuration Register 0 */
+#define SPRN_A2_CCR1	0x3f1	/* Core Configuration Register 1 */
+#define SPRN_A2_CCR2	0x3f2	/* Core Configuration Register 2 */
+#define SPRN_MMUCR0	0x3fc	/* MMU Control Register 0 */
+#define SPRN_MMUCR1	0x3fd	/* MMU Control Register 1 */
+#define SPRN_MMUCR2	0x3fe	/* MMU Control Register 2 */
+#define SPRN_MMUCR3	0x3ff	/* MMU Control Register 3 */
+
+#define SPRN_IAR	0x372
+
+#define SPRN_IUCR0	0x3f3
+#define IUCR0_ICBI_ACK	0x1000
+
+#define SPRN_XUCR0	0x3f6	/* Execution Unit Config Register 0 */
+
+#define A2_IERAT_SIZE	16
+#define A2_DERAT_SIZE	32
+
+/* A2 MMUCR0 bits */
+#define MMUCR0_ECL	0x80000000	/* Extended Class for TLB fills */
+#define MMUCR0_TID_NZ	0x40000000	/* TID is non-zero */
+#define MMUCR0_TS	0x10000000	/* Translation space for TLB fills */
+#define MMUCR0_TGS	0x20000000	/* Guest space for TLB fills */
+#define MMUCR0_TLBSEL	0x0c000000	/* TLB or ERAT target for TLB fills */
+#define MMUCR0_TLBSEL_U	0x00000000	/*  TLBSEL = UTLB */
+#define MMUCR0_TLBSEL_I	0x08000000	/*  TLBSEL = I-ERAT */
+#define MMUCR0_TLBSEL_D	0x0c000000	/*  TLBSEL = D-ERAT */
+#define MMUCR0_LOCKSRSH	0x02000000	/* Use TLB lock on tlbsx. */
+#define MMUCR0_TID_MASK	0x000000ff	/* TID field */
+
+/* A2 MMUCR1 bits */
+#define MMUCR1_IRRE		0x80000000	/* I-ERAT round robin enable */
+#define MMUCR1_DRRE		0x40000000	/* D-ERAT round robin enable */
+#define MMUCR1_REE		0x20000000	/* Reference Exception Enable*/
+#define MMUCR1_CEE		0x10000000	/* Change exception enable */
+#define MMUCR1_CSINV_ALL	0x00000000	/* Inval ERAT on all CS evts */
+#define MMUCR1_CSINV_NISYNC	0x04000000	/* Inval ERAT on all ex isync*/
+#define MMUCR1_CSINV_NEVER	0x0c000000	/* Don't inval ERAT on CS */
+#define MMUCR1_ICTID		0x00080000	/* IERAT class field as TID */
+#define MMUCR1_ITTID		0x00040000	/* IERAT thdid field as TID */
+#define MMUCR1_DCTID		0x00020000	/* DERAT class field as TID */
+#define MMUCR1_DTTID		0x00010000	/* DERAT thdid field as TID */
+#define MMUCR1_DCCD		0x00008000	/* DERAT class ignore */
+#define MMUCR1_TLBWE_BINV	0x00004000	/* back invalidate on tlbwe */
+
+/* A2 MMUCR2 bits */
+#define MMUCR2_PSSEL_SHIFT	4
+
+/* A2 MMUCR3 bits */
+#define MMUCR3_THID		0x0000000f	/* Thread ID */
+
+/* *** ERAT TLB bits definitions */
+#define TLB0_EPN_MASK		ASM_CONST(0xfffffffffffff000)
+#define TLB0_CLASS_MASK		ASM_CONST(0x0000000000000c00)
+#define TLB0_CLASS_00		ASM_CONST(0x0000000000000000)
+#define TLB0_CLASS_01		ASM_CONST(0x0000000000000400)
+#define TLB0_CLASS_10		ASM_CONST(0x0000000000000800)
+#define TLB0_CLASS_11		ASM_CONST(0x0000000000000c00)
+#define TLB0_V			ASM_CONST(0x0000000000000200)
+#define TLB0_X			ASM_CONST(0x0000000000000100)
+#define TLB0_SIZE_MASK		ASM_CONST(0x00000000000000f0)
+#define TLB0_SIZE_4K		ASM_CONST(0x0000000000000010)
+#define TLB0_SIZE_64K		ASM_CONST(0x0000000000000030)
+#define TLB0_SIZE_1M		ASM_CONST(0x0000000000000050)
+#define TLB0_SIZE_16M		ASM_CONST(0x0000000000000070)
+#define TLB0_SIZE_1G		ASM_CONST(0x00000000000000a0)
+#define TLB0_THDID_MASK		ASM_CONST(0x000000000000000f)
+#define TLB0_THDID_0		ASM_CONST(0x0000000000000001)
+#define TLB0_THDID_1		ASM_CONST(0x0000000000000002)
+#define TLB0_THDID_2		ASM_CONST(0x0000000000000004)
+#define TLB0_THDID_3		ASM_CONST(0x0000000000000008)
+#define TLB0_THDID_ALL		ASM_CONST(0x000000000000000f)
+
+#define TLB1_RESVATTR		ASM_CONST(0x00f0000000000000)
+#define TLB1_U0			ASM_CONST(0x0008000000000000)
+#define TLB1_U1			ASM_CONST(0x0004000000000000)
+#define TLB1_U2			ASM_CONST(0x0002000000000000)
+#define TLB1_U3			ASM_CONST(0x0001000000000000)
+#define TLB1_R			ASM_CONST(0x0000800000000000)
+#define TLB1_C			ASM_CONST(0x0000400000000000)
+#define TLB1_RPN_MASK		ASM_CONST(0x000003fffffff000)
+#define TLB1_W			ASM_CONST(0x0000000000000800)
+#define TLB1_I			ASM_CONST(0x0000000000000400)
+#define TLB1_M			ASM_CONST(0x0000000000000200)
+#define TLB1_G			ASM_CONST(0x0000000000000100)
+#define TLB1_E			ASM_CONST(0x0000000000000080)
+#define TLB1_VF			ASM_CONST(0x0000000000000040)
+#define TLB1_UX			ASM_CONST(0x0000000000000020)
+#define TLB1_SX			ASM_CONST(0x0000000000000010)
+#define TLB1_UW			ASM_CONST(0x0000000000000008)
+#define TLB1_SW			ASM_CONST(0x0000000000000004)
+#define TLB1_UR			ASM_CONST(0x0000000000000002)
+#define TLB1_SR			ASM_CONST(0x0000000000000001)
+
+/* A2 erativax attributes definitions */
+#define ERATIVAX_RS_IS_ALL		0x000
+#define ERATIVAX_RS_IS_TID		0x040
+#define ERATIVAX_RS_IS_CLASS		0x080
+#define ERATIVAX_RS_IS_FULLMATCH	0x0c0
+#define ERATIVAX_CLASS_00		0x000
+#define ERATIVAX_CLASS_01		0x010
+#define ERATIVAX_CLASS_10		0x020
+#define ERATIVAX_CLASS_11		0x030
+#define ERATIVAX_PSIZE_4K		(TLB_PSIZE_4K >> 1)
+#define ERATIVAX_PSIZE_64K		(TLB_PSIZE_64K >> 1)
+#define ERATIVAX_PSIZE_1M		(TLB_PSIZE_1M >> 1)
+#define ERATIVAX_PSIZE_16M		(TLB_PSIZE_16M >> 1)
+#define ERATIVAX_PSIZE_1G		(TLB_PSIZE_1G >> 1)
+
+/* A2 eratilx attributes definitions */
+#define ERATILX_T_ALL			0
+#define ERATILX_T_TID			1
+#define ERATILX_T_TGS			2
+#define ERATILX_T_FULLMATCH		3
+#define ERATILX_T_CLASS0		4
+#define ERATILX_T_CLASS1		5
+#define ERATILX_T_CLASS2		6
+#define ERATILX_T_CLASS3		7
+
+/* XUCR0 bits */
+#define XUCR0_TRACE_UM_T0		0x40000000	/* Thread 0 */
+#define XUCR0_TRACE_UM_T1		0x20000000	/* Thread 1 */
+#define XUCR0_TRACE_UM_T2		0x10000000	/* Thread 2 */
+#define XUCR0_TRACE_UM_T3		0x08000000	/* Thread 3 */
+
+/* A2 CCR0 register */
+#define A2_CCR0_PME_DISABLED		0x00000000
+#define A2_CCR0_PME_SLEEP		0x40000000
+#define A2_CCR0_PME_RVW			0x80000000
+#define A2_CCR0_PME_DISABLED2		0xc0000000
+
+/* A2 CCR2 register */
+#define A2_CCR2_ERAT_ONLY_MODE		0x00000001
+#define A2_CCR2_ENABLE_ICSWX		0x00000002
+#define A2_CCR2_ENABLE_PC		0x20000000
+#define A2_CCR2_ENABLE_TRACE		0x40000000
+
+#endif /* __ASM_POWERPC_REG_A2_H__ */
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index b0ba78f..82e0bed 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -41,6 +41,7 @@ obj-$(CONFIG_PPC_BOOK3S_64)	+= cpu_setup_ppc970.o cpu_setup_pa6t.o
 obj-$(CONFIG_PPC_BOOK3S_64)	+= cpu_setup_power7.o
 obj64-$(CONFIG_RELOCATABLE)	+= reloc_64.o
 obj-$(CONFIG_PPC_BOOK3E_64)	+= exceptions-64e.o idle_book3e.o
+obj-$(CONFIG_PPC_A2)		+= cpu_setup_a2.o
 obj-$(CONFIG_PPC64)		+= vdso64/
 obj-$(CONFIG_ALTIVEC)		+= vecemu.o
 obj-$(CONFIG_PPC_970_NAP)	+= idle_power4.o
diff --git a/arch/powerpc/kernel/cpu_setup_a2.S b/arch/powerpc/kernel/cpu_setup_a2.S
new file mode 100644
index 0000000..7f818fe
--- /dev/null
+++ b/arch/powerpc/kernel/cpu_setup_a2.S
@@ -0,0 +1,114 @@
+/*
+ *  A2 specific assembly support code
+ *
+ *  Copyright 2009 Ben Herrenschmidt, IBM Corp.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+
+#include <asm/asm-offsets.h>
+#include <asm/ppc_asm.h>
+#include <asm/ppc-opcode.h>
+#include <asm/processor.h>
+#include <asm/reg_a2.h>
+#include <asm/reg.h>
+#include <asm/thread_info.h>
+
+/*
+ * Disable thdid and class fields in ERATs to bump PID to full 14 bits capacity.
+ * This also prevents external LPID accesses but that isn't a problem when not a
+ * guest. Under PV, this setting will be ignored and MMUCR will return the right
+ * number of PID bits we can use.
+ */
+#define MMUCR1_EXTEND_PID \
+	(MMUCR1_ICTID | MMUCR1_ITTID | MMUCR1_DCTID | \
+	 MMUCR1_DTTID | MMUCR1_DCCD)
+
+/*
+ * Use extended PIDs if enabled.
+ * Don't clear the ERATs on context sync events and enable I & D LRU.
+ * Enable ERAT back invalidate when tlbwe overwrites an entry.
+ */
+#define INITIAL_MMUCR1 \
+	(MMUCR1_EXTEND_PID | MMUCR1_CSINV_NEVER | MMUCR1_IRRE | \
+	 MMUCR1_DRRE | MMUCR1_TLBWE_BINV)
+
+_GLOBAL(__setup_cpu_a2)
+	/* Some of these are actually thread local and some are
+	 * core local but doing it always won't hurt
+	 */
+
+#ifdef CONFIG_PPC_WSP_COPRO
+	/* Make sure ACOP starts out as zero */
+	li	r3,0
+	mtspr   SPRN_ACOP,r3
+
+	/* Enable icswx instruction */
+	mfspr   r3,SPRN_A2_CCR2
+	ori     r3,r3,A2_CCR2_ENABLE_ICSWX
+	mtspr   SPRN_A2_CCR2,r3
+
+	/* Unmask all CTs in HACOP */
+	li      r3,-1
+	mtspr   SPRN_HACOP,r3
+#endif /* CONFIG_PPC_WSP_COPRO */
+
+	/* Enable doorbell */
+	mfspr   r3,SPRN_A2_CCR2
+	oris     r3,r3,A2_CCR2_ENABLE_PC@h
+	mtspr   SPRN_A2_CCR2,r3
+	isync
+
+	/* Setup CCR0 to disable power saving for now as it's busted
+	 * in the current implementations. Setup CCR1 to wake on
+	 * interrupts normally (we write the default value but who
+	 * knows what FW may have clobbered...)
+	 */
+	li	r3,0
+	mtspr	SPRN_A2_CCR0, r3
+	LOAD_REG_IMMEDIATE(r3,0x0f0f0f0f)
+	mtspr	SPRN_A2_CCR1, r3
+
+	/* Initialise MMUCR1 */
+	lis	r3,INITIAL_MMUCR1@h
+	ori	r3,r3,INITIAL_MMUCR1@l
+	mtspr	SPRN_MMUCR1,r3
+
+	/* Set MMUCR2 to enable 4K, 64K, 1M, 16M and 1G pages */
+	LOAD_REG_IMMEDIATE(r3, 0x000a7531)
+	mtspr	SPRN_MMUCR2,r3
+
+	/* Set MMUCR3 to write all thids bit to the TLB */
+	LOAD_REG_IMMEDIATE(r3, 0x0000000f)
+	mtspr	SPRN_MMUCR3,r3
+
+	/* Don't do ERAT stuff if running guest mode */
+	mfmsr	r3
+	andis.	r0,r3,MSR_GS@h
+	bne	1f
+
+	/* Now set the I-ERAT watermark to 15 */
+	lis	r4,(MMUCR0_TLBSEL_I|MMUCR0_ECL)@h
+	mtspr	SPRN_MMUCR0, r4
+	li	r4,A2_IERAT_SIZE-1
+	PPC_ERATWE(r4,r4,3)
+
+	/* Now set the D-ERAT watermark to 31 */
+	lis	r4,(MMUCR0_TLBSEL_D|MMUCR0_ECL)@h
+	mtspr	SPRN_MMUCR0, r4
+	li	r4,A2_DERAT_SIZE-1
+	PPC_ERATWE(r4,r4,3)
+
+	/* And invalidate the beast just in case. That won't get rid of
+	 * a bolted entry though it will be in LRU and so will go away eventually
+	 * but let's not bother for now
+	 */
+	PPC_ERATILX(0,0,0)
+1:
+	blr
+
+_GLOBAL(__restore_cpu_a2)
+	b	__setup_cpu_a2
diff --git a/arch/powerpc/kernel/cputable.c b/arch/powerpc/kernel/cputable.c
index 70cbae1..69d1e76 100644
--- a/arch/powerpc/kernel/cputable.c
+++ b/arch/powerpc/kernel/cputable.c
@@ -62,10 +62,12 @@ extern void __setup_cpu_745x(unsigned long offset, struct cpu_spec* spec);
 extern void __setup_cpu_ppc970(unsigned long offset, struct cpu_spec* spec);
 extern void __setup_cpu_ppc970MP(unsigned long offset, struct cpu_spec* spec);
 extern void __setup_cpu_pa6t(unsigned long offset, struct cpu_spec* spec);
+extern void __setup_cpu_a2(unsigned long offset, struct cpu_spec* spec);
 extern void __restore_cpu_pa6t(void);
 extern void __restore_cpu_ppc970(void);
 extern void __setup_cpu_power7(unsigned long offset, struct cpu_spec* spec);
 extern void __restore_cpu_power7(void);
+extern void __restore_cpu_a2(void);
 #endif /* CONFIG_PPC64 */
 #if defined(CONFIG_E500)
 extern void __setup_cpu_e5500(unsigned long offset, struct cpu_spec* spec);
@@ -2011,7 +2013,26 @@ static struct cpu_spec __initdata cpu_specs[] = {
 #endif /* CONFIG_PPC32 */
 #endif /* CONFIG_E500 */
 
-#ifdef CONFIG_PPC_BOOK3E_64
+#ifdef CONFIG_PPC_A2
+	{	/* Standard A2 (>= DD2) + FPU core */
+		.pvr_mask		= 0xffff0000,
+		.pvr_value		= 0x00480000,
+		.cpu_name		= "A2 (>= DD2)",
+		.cpu_features		= CPU_FTRS_A2,
+		.cpu_user_features	= COMMON_USER_PPC64,
+		.mmu_features		= MMU_FTR_TYPE_3E | MMU_FTR_USE_TLBILX |
+					  MMU_FTR_USE_TLBIVAX_BCAST |
+					  MMU_FTR_LOCK_BCAST_INVAL |
+					  MMU_FTR_USE_TLBRSRV |
+					  MMU_FTR_USE_PAIRED_MAS,
+		.icache_bsize		= 64,
+		.dcache_bsize		= 64,
+		.num_pmcs		= 0,
+		.cpu_setup		= __setup_cpu_a2,
+		.cpu_restore		= __restore_cpu_a2,
+		.machine_check		= machine_check_generic,
+		.platform		= "ppca2",
+	},
 	{	/* This is a default entry to get going, to be replaced by
 		 * a real one at some stage
 		 */
@@ -2029,7 +2050,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
 		.machine_check		= machine_check_generic,
 		.platform		= "power6",
 	},
-#endif
+#endif /* CONFIG_PPC_A2 */
 };
 
 static struct cpu_spec the_cpu_spec;
diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
index 111138c..7c1e1c6 100644
--- a/arch/powerpc/platforms/Kconfig.cputype
+++ b/arch/powerpc/platforms/Kconfig.cputype
@@ -107,6 +107,10 @@ config POWER4
 	depends on PPC64 && PPC_BOOK3S
 	def_bool y
 
+config PPC_A2
+	bool
+	depends on PPC_BOOK3E_64
+
 config TUNE_CELL
 	bool "Optimize for Cell Broadband Engine"
 	depends on PPC64 && PPC_BOOK3S
-- 
1.7.1

[PATCH 09/15] powerpc: Add TLB size detection for TYPE_3E MMUs

[Michael Ellerman]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/mmu-book3e.h |   15 +++++++++++++++
 arch/powerpc/mm/mmu_context_nohash.c  |   12 +++++++++++-
 2 files changed, 26 insertions(+), 1 deletions(-)

diff --git a/arch/powerpc/include/asm/mmu-book3e.h b/arch/powerpc/include/asm/mmu-book3e.h
index f6e7d40..3903d0c 100644
--- a/arch/powerpc/include/asm/mmu-book3e.h
+++ b/arch/powerpc/include/asm/mmu-book3e.h
@@ -137,6 +137,21 @@
 #define MMUCSR0_TLB2PS	0x00078000	/* TLB2 Page Size */
 #define MMUCSR0_TLB3PS	0x00780000	/* TLB3 Page Size */
 
+/* MMUCFG bits */
+#define MMUCFG_MAVN_NASK	0x00000003
+#define MMUCFG_MAVN_V1_0	0x00000000
+#define MMUCFG_MAVN_V2_0	0x00000001
+#define MMUCFG_NTLB_MASK	0x0000000c
+#define MMUCFG_NTLB_SHIFT	2
+#define MMUCFG_PIDSIZE_MASK	0x000007c0
+#define MMUCFG_PIDSIZE_SHIFT	6
+#define MMUCFG_TWC		0x00008000
+#define MMUCFG_LRAT		0x00010000
+#define MMUCFG_RASIZE_MASK	0x00fe0000
+#define MMUCFG_RASIZE_SHIFT	17
+#define MMUCFG_LPIDSIZE_MASK	0x0f000000
+#define MMUCFG_LPIDSIZE_SHIFT	24
+
 /* TLBnCFG encoding */
 #define TLBnCFG_N_ENTRY		0x00000fff	/* number of entries */
 #define TLBnCFG_HES		0x00002000	/* HW select supported */
diff --git a/arch/powerpc/mm/mmu_context_nohash.c b/arch/powerpc/mm/mmu_context_nohash.c
index 4d8fa91..336807d 100644
--- a/arch/powerpc/mm/mmu_context_nohash.c
+++ b/arch/powerpc/mm/mmu_context_nohash.c
@@ -409,7 +409,17 @@ void __init mmu_context_init(void)
 	} else if (mmu_has_feature(MMU_FTR_TYPE_47x)) {
 		first_context = 1;
 		last_context = 65535;
-	} else {
+	} else
+#ifdef CONFIG_PPC_BOOK3E_MMU
+	if (mmu_has_feature(MMU_FTR_TYPE_3E)) {
+		u32 mmucfg = mfspr(SPRN_MMUCFG);
+		u32 pid_bits = (mmucfg & MMUCFG_PIDSIZE_MASK)
+				>> MMUCFG_PIDSIZE_SHIFT;
+		first_context = 1;
+		last_context = (1UL << (pid_bits + 1)) - 1;
+	} else
+#endif
+	{
 		first_context = 1;
 		last_context = 255;
 	}
-- 
1.7.1

[PATCH 10/15] powerpc: Define slb0_limit() for BOOK3E

[Michael Ellerman]

From: Michael Ellerman <michael@ellerman.id.au>

On BOOK3E we don't have an SLB 0, but the equivalent concept is the
bolted entry mapping the kernel. Currently this is a 1G entry, so
for now hardcode that. This will probably need to be reworked in
future.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/kernel/setup_64.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index 91a5cc5..3d0daf4 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c
@@ -436,10 +436,14 @@ void __init setup_system(void)
 
 static u64 slb0_limit(void)
 {
+#ifdef CONFIG_PPC_BOOK3E
+	return 1 << 30;
+#else
 	if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) {
 		return 1UL << SID_SHIFT_1T;
 	}
 	return 1UL << SID_SHIFT;
+#endif
 }
 
 static void __init irqstack_early_init(void)
-- 
1.7.1

Re: [PATCH 10/15] powerpc: Define slb0_limit() for BOOK3E

[Olof Johansson]

Hi,

On Fri, Apr 15, 2011 at 1:32 AM, Michael Ellerman <michael@ozlabs.org> wrot=
e:
> From: Michael Ellerman <michael@ellerman.id.au>
>
> On BOOK3E we don't have an SLB 0, but the equivalent concept is the
> bolted entry mapping the kernel. Currently this is a 1G entry, so
> for now hardcode that. This will probably need to be reworked in
> future.

A year from now when someone sees the ifdef, the above comment would
be nice to have in the code instead of running a git annotate/log to
find out why it's hardcoded the way it is. :)


-Olof


>
> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> ---
> =A0arch/powerpc/kernel/setup_64.c | =A0 =A04 ++++
> =A01 files changed, 4 insertions(+), 0 deletions(-)
>
> diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_6=
4.c
> index 91a5cc5..3d0daf4 100644
> --- a/arch/powerpc/kernel/setup_64.c
> +++ b/arch/powerpc/kernel/setup_64.c
> @@ -436,10 +436,14 @@ void __init setup_system(void)
>
> =A0static u64 slb0_limit(void)
> =A0{
> +#ifdef CONFIG_PPC_BOOK3E
> + =A0 =A0 =A0 return 1 << 30;
> +#else
> =A0 =A0 =A0 =A0if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) {
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0return 1UL << SID_SHIFT_1T;
> =A0 =A0 =A0 =A0}
> =A0 =A0 =A0 =A0return 1UL << SID_SHIFT;
> +#endif
> =A0}
>
> =A0static void __init irqstack_early_init(void)
> --
> 1.7.1
>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev
>

Re: [PATCH 10/15] powerpc: Define slb0_limit() for BOOK3E

[Kumar Gala]

On Apr 15, 2011, at 3:32 AM, Michael Ellerman wrote:

> From: Michael Ellerman <michael@ellerman.id.au>
>=20
> On BOOK3E we don't have an SLB 0, but the equivalent concept is the
> bolted entry mapping the kernel. Currently this is a 1G entry, so
> for now hardcode that. This will probably need to be reworked in
> future.
>=20
> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> ---
> arch/powerpc/kernel/setup_64.c |    4 ++++
> 1 files changed, 4 insertions(+), 0 deletions(-)
>=20
> diff --git a/arch/powerpc/kernel/setup_64.c =
b/arch/powerpc/kernel/setup_64.c
> index 91a5cc5..3d0daf4 100644
> --- a/arch/powerpc/kernel/setup_64.c
> +++ b/arch/powerpc/kernel/setup_64.c
> @@ -436,10 +436,14 @@ void __init setup_system(void)
>=20
> static u64 slb0_limit(void)
> {
> +#ifdef CONFIG_PPC_BOOK3E
> +	return 1 << 30;
> +#else
> 	if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) {
> 		return 1UL << SID_SHIFT_1T;
> 	}
> 	return 1UL << SID_SHIFT;
> +#endif
> }
>=20

Let's rename this function to something 'linear_map'.  As on FSL Book-E =
64 we do things a bit differently and have more covered in linear map =
than 1G

> static void __init irqstack_early_init(void)
> --=20
> 1.7.1
>=20
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev

Re: [PATCH 10/15] powerpc: Define slb0_limit() for BOOK3E

[Benjamin Herrenschmidt]

On Mon, 2011-04-18 at 07:42 -0500, Kumar Gala wrote:
> Let's rename this function to something 'linear_map'.  As on FSL
> Book-E 64 we do things a bit differently and have more covered in
> linear map than 1G

It's not quite linear_map. It's whatever can be accessed without taking
exceptions. IE. What is bolted.

It should probably go into a variable initialized by the mm code tho.

Cheers,
Ben.

Re: [PATCH 10/15] powerpc: Define slb0_limit() for BOOK3E

[Kumar Gala]

On Apr 18, 2011, at 4:41 PM, Benjamin Herrenschmidt wrote:

> On Mon, 2011-04-18 at 07:42 -0500, Kumar Gala wrote:
>> Let's rename this function to something 'linear_map'.  As on FSL
>> Book-E 64 we do things a bit differently and have more covered in
>> linear map than 1G
> 
> It's not quite linear_map. It's whatever can be accessed without taking
> exceptions. IE. What is bolted.
> 
> It should probably go into a variable initialized by the mm code tho.
> 
> Cheers,
> Ben.

Is this 'ppc64_rma_size' ?

- k

Re: [PATCH 10/15] powerpc: Define slb0_limit() for BOOK3E

[Benjamin Herrenschmidt]

On Mon, 2011-04-18 at 17:30 -0500, Kumar Gala wrote:
> > On Mon, 2011-04-18 at 07:42 -0500, Kumar Gala wrote:
> >> Let's rename this function to something 'linear_map'.  As on FSL
> >> Book-E 64 we do things a bit differently and have more covered in
> >> linear map than 1G
> > 
> > It's not quite linear_map. It's whatever can be accessed without
> taking
> > exceptions. IE. What is bolted.
> > 
> > It should probably go into a variable initialized by the mm code
> tho.
> > 
> > Cheers,
> > Ben.
> 
> Is this 'ppc64_rma_size' ? 

Not quite either :-)

The RMA (on server) is the amount of memory that is accessible in real
mode (with MMU off).

SLB0 is segment 0, which is 256M or 1T depending on what segment size we
use, and is a bolted segment in Linux. Thus all accesses to SLB0 are
guaranteed to not generate an SLB miss fault. Since on server our linear
mapping is bolted in the hash, what that means is that any access to
memory in segment 0 is guaranteed to not take an exception (ie not
clobber SRR0/1 for example).

We re-use the SLB0 concept on BookE to represent the bolted portion of
the linear mapping since it has the same characteristics, ie, accesses
will not cause an exception that can clobber SRR0/1.

The patch Michael posted is indeed a bit gross (it's my code btw :-) and
we'll need to clean it up, my idea is to have a global set by the
platform, but maybe we can have some generic code that performs a TLB
search for IPROT entries :-)

Cheers,
Ben.

[PATCH 12/15] powerpc/book3e: Use way 3 for linear mapping bolted entry

[Michael Ellerman]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

An erratum on A2 can lead to the bolted entry we insert for the linear
mapping being evicted, to avoid that write the bolted entry to way 3.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/kernel/exceptions-64e.S |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/kernel/exceptions-64e.S b/arch/powerpc/kernel/exceptions-64e.S
index 5c43063..e6c0926 100644
--- a/arch/powerpc/kernel/exceptions-64e.S
+++ b/arch/powerpc/kernel/exceptions-64e.S
@@ -864,8 +864,9 @@ have_hes:
 	 * that will have to be made dependent on whether we are running under
 	 * a hypervisor I suppose.
 	 */
-	ori	r3,r3,MAS0_HES | MAS0_WQ_ALLWAYS
-	mtspr	SPRN_MAS0,r3
+	ori	r11,r3,MAS0_WQ_ALLWAYS
+	oris	r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 erratum 376 */
+	mtspr	SPRN_MAS0,r11
 	lis	r3,(MAS1_VALID | MAS1_IPROT)@h
 	ori	r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT
 	mtspr	SPRN_MAS1,r3
-- 
1.7.1

Re: [PATCH 12/15] powerpc/book3e: Use way 3 for linear mapping bolted entry

[Kumar Gala]

On Apr 15, 2011, at 3:32 AM, Michael Ellerman wrote:

> From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>=20
> An erratum on A2 can lead to the bolted entry we insert for the linear
> mapping being evicted, to avoid that write the bolted entry to way 3.
>=20
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> ---
> arch/powerpc/kernel/exceptions-64e.S |    5 +++--
> 1 files changed, 3 insertions(+), 2 deletions(-)
>=20
> diff --git a/arch/powerpc/kernel/exceptions-64e.S =
b/arch/powerpc/kernel/exceptions-64e.S
> index 5c43063..e6c0926 100644
> --- a/arch/powerpc/kernel/exceptions-64e.S
> +++ b/arch/powerpc/kernel/exceptions-64e.S
> @@ -864,8 +864,9 @@ have_hes:
> 	 * that will have to be made dependent on whether we are running =
under
> 	 * a hypervisor I suppose.
> 	 */
> -	ori	r3,r3,MAS0_HES | MAS0_WQ_ALLWAYS
> -	mtspr	SPRN_MAS0,r3
> +	ori	r11,r3,MAS0_WQ_ALLWAYS
> +	oris	r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 =
erratum 376 */
> +	mtspr	SPRN_MAS0,r11
> 	lis	r3,(MAS1_VALID | MAS1_IPROT)@h
> 	ori	r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT
> 	mtspr	SPRN_MAS1,r3

Seems like this should have a MMU Feature bit or something for A2.

- k=

Re: [PATCH 12/15] powerpc/book3e: Use way 3 for linear mapping bolted entry

[Benjamin Herrenschmidt]

On Mon, 2011-04-18 at 07:43 -0500, Kumar Gala wrote:
> On Apr 15, 2011, at 3:32 AM, Michael Ellerman wrote:
> 
> > From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> > 
> > An erratum on A2 can lead to the bolted entry we insert for the linear
> > mapping being evicted, to avoid that write the bolted entry to way 3.
> > 
> > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> > Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> > ---
> > arch/powerpc/kernel/exceptions-64e.S |    5 +++--
> > 1 files changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/arch/powerpc/kernel/exceptions-64e.S b/arch/powerpc/kernel/exceptions-64e.S
> > index 5c43063..e6c0926 100644
> > --- a/arch/powerpc/kernel/exceptions-64e.S
> > +++ b/arch/powerpc/kernel/exceptions-64e.S
> > @@ -864,8 +864,9 @@ have_hes:
> > 	 * that will have to be made dependent on whether we are running under
> > 	 * a hypervisor I suppose.
> > 	 */
> > -	ori	r3,r3,MAS0_HES | MAS0_WQ_ALLWAYS
> > -	mtspr	SPRN_MAS0,r3
> > +	ori	r11,r3,MAS0_WQ_ALLWAYS
> > +	oris	r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 erratum 376 */
> > +	mtspr	SPRN_MAS0,r11
> > 	lis	r3,(MAS1_VALID | MAS1_IPROT)@h
> > 	ori	r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT
> > 	mtspr	SPRN_MAS1,r3
> 
> Seems like this should have a MMU Feature bit or something for A2.

Too early. We haven't detected the CPU and are establishing the initial
TLB entry here.

Any reason why that wouldn't work on something else anyways ?

Cheers,
Ben.

Re: [PATCH 12/15] powerpc/book3e: Use way 3 for linear mapping bolted entry

[Kumar Gala]

On Apr 18, 2011, at 4:40 PM, Benjamin Herrenschmidt wrote:

> On Mon, 2011-04-18 at 07:43 -0500, Kumar Gala wrote:
>> On Apr 15, 2011, at 3:32 AM, Michael Ellerman wrote:
>>=20
>>> From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>>>=20
>>> An erratum on A2 can lead to the bolted entry we insert for the =
linear
>>> mapping being evicted, to avoid that write the bolted entry to way =
3.
>>>=20
>>> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>>> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
>>> ---
>>> arch/powerpc/kernel/exceptions-64e.S |    5 +++--
>>> 1 files changed, 3 insertions(+), 2 deletions(-)
>>>=20
>>> diff --git a/arch/powerpc/kernel/exceptions-64e.S =
b/arch/powerpc/kernel/exceptions-64e.S
>>> index 5c43063..e6c0926 100644
>>> --- a/arch/powerpc/kernel/exceptions-64e.S
>>> +++ b/arch/powerpc/kernel/exceptions-64e.S
>>> @@ -864,8 +864,9 @@ have_hes:
>>> 	 * that will have to be made dependent on whether we are running =
under
>>> 	 * a hypervisor I suppose.
>>> 	 */
>>> -	ori	r3,r3,MAS0_HES | MAS0_WQ_ALLWAYS
>>> -	mtspr	SPRN_MAS0,r3
>>> +	ori	r11,r3,MAS0_WQ_ALLWAYS
>>> +	oris	r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 =
erratum 376 */
>>> +	mtspr	SPRN_MAS0,r11
>>> 	lis	r3,(MAS1_VALID | MAS1_IPROT)@h
>>> 	ori	r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT
>>> 	mtspr	SPRN_MAS1,r3
>>=20
>> Seems like this should have a MMU Feature bit or something for A2.
>=20
> Too early. We haven't detected the CPU and are establishing the =
initial
> TLB entry here.

How about wrapping with CONFIG_PPC_A2

> Any reason why that wouldn't work on something else anyways ?

No, I wasn't paying attention to what this code exactly was (not enough =
context in the diff), I see its our initial setup so not a big deal.  =
Was thinking this was run-time exception handler code.

- k

Re: [PATCH 12/15] powerpc/book3e: Use way 3 for linear mapping bolted entry

[Benjamin Herrenschmidt]

> >> Seems like this should have a MMU Feature bit or something for A2.
> > 
> > Too early. We haven't detected the CPU and are establishing the initial
> > TLB entry here.
> 
> How about wrapping with CONFIG_PPC_A2

YUCK :-)

> > Any reason why that wouldn't work on something else anyways ?
> 
> No, I wasn't paying attention to what this code exactly was (not enough
> context in the diff), I see its our initial setup so not a big deal.
>   Was thinking this was run-time exception handler code.

No, it's the boot time bolted entry.

Cheers,
Ben.

[PATCH 11/15] powerpc: Index crit/dbg/mcheck stacks using cpu number on 64bit

[Michael Ellerman]

From: Michael Ellerman <michael@ellerman.id.au>

In exc_lvl_ctx_init() we index into the crit/dbg/mcheck stacks using
the hard cpu id, but that assumes the hard cpu id is zero based and
contiguous. That is not the case on A2.

The root of the problem is that the 32bit code has no equivalent of the
paca to allow it to do the hard->soft mapping in assembler. Until the
32bit code is updated to handle that, index the stacks using the soft
cpu ids on 64bit and hard on 32 bit.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/kernel/irq.c |   26 +++++++++++++++-----------
 1 files changed, 15 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index ac32ba1..4f5d6e7 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -397,24 +397,28 @@ struct thread_info *mcheckirq_ctx[NR_CPUS] __read_mostly;
 void exc_lvl_ctx_init(void)
 {
 	struct thread_info *tp;
-	int i, hw_cpu;
+	int i, cpu_nr;
 
 	for_each_possible_cpu(i) {
-		hw_cpu = get_hard_smp_processor_id(i);
-		memset((void *)critirq_ctx[hw_cpu], 0, THREAD_SIZE);
-		tp = critirq_ctx[hw_cpu];
-		tp->cpu = i;
+#ifdef CONFIG_PPC64
+		cpu_nr = i;
+#else
+		cpu_nr = get_hard_smp_processor_id(i);
+#endif
+		memset((void *)critirq_ctx[cpu_nr], 0, THREAD_SIZE);
+		tp = critirq_ctx[cpu_nr];
+		tp->cpu = cpu_nr;
 		tp->preempt_count = 0;
 
 #ifdef CONFIG_BOOKE
-		memset((void *)dbgirq_ctx[hw_cpu], 0, THREAD_SIZE);
-		tp = dbgirq_ctx[hw_cpu];
-		tp->cpu = i;
+		memset((void *)dbgirq_ctx[cpu_nr], 0, THREAD_SIZE);
+		tp = dbgirq_ctx[cpu_nr];
+		tp->cpu = cpu_nr;
 		tp->preempt_count = 0;
 
-		memset((void *)mcheckirq_ctx[hw_cpu], 0, THREAD_SIZE);
-		tp = mcheckirq_ctx[hw_cpu];
-		tp->cpu = i;
+		memset((void *)mcheckirq_ctx[cpu_nr], 0, THREAD_SIZE);
+		tp = mcheckirq_ctx[cpu_nr];
+		tp->cpu = cpu_nr;
 		tp->preempt_count = HARDIRQ_OFFSET;
 #endif
 	}
-- 
1.7.1

[PATCH 13/15] powerpc/book3e: Flush IPROT protected TLB entries leftover by firmware

[Michael Ellerman]

From: Jack Miller <jack@codezen.org>

When we set up the TLB for ourselves on Book3E, we need to flush out any
old mappings established by the firmware or bootloader.  At present we
attempt this with a tlbilx to flush everything, but this will leave behind
any entries with the IPROT bit set.

There are several good reason firmware might establish mappings with IPROT,
and in fact ePAPR compliant firmwares are required to establish their
initial mapped area with IPROT.

This patch, therefore adds more complex code to scan through the TLB upon
entry and flush away any entries that are not our own.

Signed-off-by: Jack Miller <jack@codezen.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/mmu-book3e.h |    1 +
 arch/powerpc/kernel/exceptions-64e.S  |   45 ++++++++++++++++++++++++++++++++-
 2 files changed, 45 insertions(+), 1 deletions(-)

diff --git a/arch/powerpc/include/asm/mmu-book3e.h b/arch/powerpc/include/asm/mmu-book3e.h
index 3903d0c..84948e3 100644
--- a/arch/powerpc/include/asm/mmu-book3e.h
+++ b/arch/powerpc/include/asm/mmu-book3e.h
@@ -43,6 +43,7 @@
 #define MAS0_TLBSEL(x)		(((x) << 28) & 0x30000000)
 #define MAS0_ESEL(x)		(((x) << 16) & 0x0FFF0000)
 #define MAS0_NV(x)		((x) & 0x00000FFF)
+#define MAS0_ESEL_MASK		0x0FFF0000
 #define MAS0_HES		0x00004000
 #define MAS0_WQ_ALLWAYS		0x00000000
 #define MAS0_WQ_COND		0x00001000
diff --git a/arch/powerpc/kernel/exceptions-64e.S b/arch/powerpc/kernel/exceptions-64e.S
index e6c0926..ab5c4dd 100644
--- a/arch/powerpc/kernel/exceptions-64e.S
+++ b/arch/powerpc/kernel/exceptions-64e.S
@@ -886,8 +886,51 @@ have_hes:
 	bctr
 
 1:	/* We are now running at PAGE_OFFSET, clean the TLB of everything
-	 * else (XXX we should scan for bolted crap from the firmware too)
+	 * else (including IPROTed things left by firmware)
+	 * r4 = TLBnCFG
+	 * r3 = current address (more or less)
 	 */
+
+	li	r5,0
+	mtspr	SPRN_MAS6,r5
+	tlbsx	0,r3
+
+	rlwinm	r9,r4,0,TLBnCFG_N_ENTRY
+	rlwinm	r10,r4,8,0xff
+	addi	r10,r10,-1	/* Get inner loop mask */
+
+	li	r3,1
+
+	mfspr	r5,SPRN_MAS1
+	rlwinm	r5,r5,0,(~(MAS1_VALID|MAS1_IPROT))
+
+	mfspr	r6,SPRN_MAS2
+	rldicr	r6,r6,0,51		/* Extract EPN */
+
+	mfspr	r7,SPRN_MAS0
+	rlwinm	r7,r7,0,0xffff0fff	/* Clear HES and WQ */
+
+	rlwinm	r8,r7,16,0xfff		/* Extract ESEL */
+
+2:	add	r4,r3,r8
+	and	r4,r4,r10
+
+	rlwimi	r7,r4,16,MAS0_ESEL_MASK
+
+	mtspr	SPRN_MAS0,r7
+	mtspr	SPRN_MAS1,r5
+	mtspr	SPRN_MAS2,r6
+	tlbwe
+
+	addi	r3,r3,1
+	and.	r4,r3,r10
+
+	bne	3f
+	addis	r6,r6,(1<<30)@h
+3:
+	cmpw	r3,r9
+	blt	2b
+
 	PPC_TLBILX(0,0,0)
 	sync
 	isync
-- 
1.7.1

[PATCH 14/15] powerpc: Add WSP platform

[Michael Ellerman]

From: David Gibson <david@gibson.dropbear.id.au>

Add a platform for the Wire Speed Processor, based on the PPC A2.

This includes code for the ICS & OPB interrupt controllers, as well
as a SCOM backend, and SCOM based cpu bringup.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Jack Miller <jack@codezen.org>
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/include/asm/wsp.h        |   14 +
 arch/powerpc/kernel/exceptions-64e.S  |   23 +
 arch/powerpc/platforms/Kconfig        |    1 +
 arch/powerpc/platforms/Makefile       |    1 +
 arch/powerpc/platforms/wsp/Kconfig    |   28 ++
 arch/powerpc/platforms/wsp/Makefile   |    6 +
 arch/powerpc/platforms/wsp/ics.c      |  716 +++++++++++++++++++++++++++++++++
 arch/powerpc/platforms/wsp/ics.h      |   20 +
 arch/powerpc/platforms/wsp/opb_pic.c  |  332 +++++++++++++++
 arch/powerpc/platforms/wsp/psr2.c     |   95 +++++
 arch/powerpc/platforms/wsp/scom_smp.c |  427 ++++++++++++++++++++
 arch/powerpc/platforms/wsp/scom_wsp.c |   77 ++++
 arch/powerpc/platforms/wsp/setup.c    |   36 ++
 arch/powerpc/platforms/wsp/smp.c      |   87 ++++
 arch/powerpc/platforms/wsp/wsp.h      |   17 +
 arch/powerpc/sysdev/xics/icp-native.c |    1 +
 16 files changed, 1881 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/include/asm/wsp.h b/arch/powerpc/include/asm/wsp.h
new file mode 100644
index 0000000..c7dc830
--- /dev/null
+++ b/arch/powerpc/include/asm/wsp.h
@@ -0,0 +1,14 @@
+/*
+ *  Copyright 2011 Michael Ellerman, IBM Corp.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+#ifndef __ASM_POWERPC_WSP_H
+#define __ASM_POWERPC_WSP_H
+
+extern int wsp_get_chip_id(struct device_node *dn);
+
+#endif /* __ASM_POWERPC_WSP_H */
diff --git a/arch/powerpc/kernel/exceptions-64e.S b/arch/powerpc/kernel/exceptions-64e.S
index ab5c4dd..8e5d2c1 100644
--- a/arch/powerpc/kernel/exceptions-64e.S
+++ b/arch/powerpc/kernel/exceptions-64e.S
@@ -864,6 +864,20 @@ have_hes:
 	 * that will have to be made dependent on whether we are running under
 	 * a hypervisor I suppose.
 	 */
+
+	/* BEWARE, MAGIC
+	 * This code is called as an ordinary function on the boot CPU. But to
+	 * avoid duplication, this code is also used in SCOM bringup of
+	 * secondary CPUs. We read the code between the initial_tlb_code_start
+	 * and initial_tlb_code_end labels one instruction at a time and RAM it
+	 * into the new core via SCOM. That doesn't process branches, so there
+	 * must be none between those two labels. It also means if this code
+	 * ever takes any parameters, the SCOM code must also be updated to
+	 * provide them.
+	 */
+	.globl a2_tlbinit_code_start
+a2_tlbinit_code_start:
+
 	ori	r11,r3,MAS0_WQ_ALLWAYS
 	oris	r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 erratum 376 */
 	mtspr	SPRN_MAS0,r11
@@ -880,6 +894,9 @@ have_hes:
 	/* Write the TLB entry */
 	tlbwe
 
+	.globl a2_tlbinit_after_linear_map
+a2_tlbinit_after_linear_map:
+
 	/* Now we branch the new virtual address mapped by this entry */
 	LOAD_REG_IMMEDIATE(r3,1f)
 	mtctr	r3
@@ -931,10 +948,16 @@ have_hes:
 	cmpw	r3,r9
 	blt	2b
 
+	.globl  a2_tlbinit_after_iprot_flush
+a2_tlbinit_after_iprot_flush:
+
 	PPC_TLBILX(0,0,0)
 	sync
 	isync
 
+	.globl a2_tlbinit_code_end
+a2_tlbinit_code_end:
+
 	/* We translate LR and return */
 	mflr	r3
 	tovirt(r3,r3)
diff --git a/arch/powerpc/platforms/Kconfig b/arch/powerpc/platforms/Kconfig
index 114d748..2359454 100644
--- a/arch/powerpc/platforms/Kconfig
+++ b/arch/powerpc/platforms/Kconfig
@@ -20,6 +20,7 @@ source "arch/powerpc/platforms/embedded6xx/Kconfig"
 source "arch/powerpc/platforms/44x/Kconfig"
 source "arch/powerpc/platforms/40x/Kconfig"
 source "arch/powerpc/platforms/amigaone/Kconfig"
+source "arch/powerpc/platforms/wsp/Kconfig"
 
 config KVM_GUEST
 	bool "KVM Guest support"
diff --git a/arch/powerpc/platforms/Makefile b/arch/powerpc/platforms/Makefile
index fdb9f0b..73e2116 100644
--- a/arch/powerpc/platforms/Makefile
+++ b/arch/powerpc/platforms/Makefile
@@ -22,3 +22,4 @@ obj-$(CONFIG_PPC_CELL)		+= cell/
 obj-$(CONFIG_PPC_PS3)		+= ps3/
 obj-$(CONFIG_EMBEDDED6xx)	+= embedded6xx/
 obj-$(CONFIG_AMIGAONE)		+= amigaone/
+obj-$(CONFIG_PPC_WSP)		+= wsp/
diff --git a/arch/powerpc/platforms/wsp/Kconfig b/arch/powerpc/platforms/wsp/Kconfig
new file mode 100644
index 0000000..c3c48eb
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/Kconfig
@@ -0,0 +1,28 @@
+config PPC_WSP
+	bool
+	default n
+
+menu "WSP platform selection"
+	depends on PPC_BOOK3E_64
+
+config PPC_PSR2
+	bool "PSR-2 platform"
+	select PPC_A2
+	select GENERIC_TBSYNC
+	select PPC_SCOM
+	select EPAPR_BOOT
+	select PPC_WSP
+	select PPC_XICS
+	select PPC_ICP_NATIVE
+	default y
+
+endmenu
+
+config PPC_A2_DD2
+	bool "Support for DD2 based A2/WSP systems"
+	depends on PPC_A2
+
+config WORKAROUND_ERRATUM_463
+	depends on PPC_A2_DD2
+	bool "Workaround erratum 463"
+	default y
diff --git a/arch/powerpc/platforms/wsp/Makefile b/arch/powerpc/platforms/wsp/Makefile
new file mode 100644
index 0000000..095be73
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/Makefile
@@ -0,0 +1,6 @@
+ccflags-y			+= -mno-minimal-toc
+
+obj-y				+= setup.o ics.o
+obj-$(CONFIG_PPC_PSR2)		+= psr2.o opb_pic.o
+obj-$(CONFIG_PPC_WSP)		+= scom_wsp.o
+obj-$(CONFIG_SMP)		+= smp.o scom_smp.o
diff --git a/arch/powerpc/platforms/wsp/ics.c b/arch/powerpc/platforms/wsp/ics.c
new file mode 100644
index 0000000..9d40798
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/ics.c
@@ -0,0 +1,716 @@
+/*
+ * Copyright 2008-2011 IBM Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/cpu.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/msi.h>
+#include <linux/of.h>
+#include <linux/slab.h>
+#include <linux/smp.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+
+#include <asm/io.h>
+#include <asm/irq.h>
+#include <asm/xics.h>
+
+#include "wsp.h"
+#include "ics.h"
+
+
+/* WSP ICS */
+
+struct wsp_ics {
+	struct ics ics;
+	struct device_node *dn;
+	void __iomem *regs;
+	spinlock_t lock;
+	unsigned long *bitmap;
+	u32 chip_id;
+	u32 lsi_base;
+	u32 lsi_count;
+	u64 hwirq_start;
+	u64 count;
+#ifdef CONFIG_SMP
+	int *hwirq_cpu_map;
+#endif
+};
+
+#define to_wsp_ics(ics)	container_of(ics, struct wsp_ics, ics)
+
+#define INT_SRC_LAYER_BUID_REG(base)	((base) + 0x00)
+#define IODA_TBL_ADDR_REG(base)		((base) + 0x18)
+#define IODA_TBL_DATA_REG(base)		((base) + 0x20)
+#define XIVE_UPDATE_REG(base)		((base) + 0x28)
+#define ICS_INT_CAPS_REG(base)		((base) + 0x30)
+
+#define TBL_AUTO_INCREMENT	((1UL << 63) | (1UL << 15))
+#define TBL_SELECT_XIST		(1UL << 48)
+#define TBL_SELECT_XIVT		(1UL << 49)
+
+#define IODA_IRQ(irq)		((irq) & (0x7FFULL))	/* HRM 5.1.3.4 */
+
+#define XIST_REQUIRED		0x8
+#define XIST_REJECTED		0x4
+#define XIST_PRESENTED		0x2
+#define XIST_PENDING		0x1
+
+#define XIVE_SERVER_SHIFT	42
+#define XIVE_SERVER_MASK	0xFFFFULL
+#define XIVE_PRIORITY_MASK	0xFFULL
+#define XIVE_PRIORITY_SHIFT	32
+#define XIVE_WRITE_ENABLE	(1ULL << 63)
+
+/*
+ * The docs refer to a 6 bit field called ChipID, which consists of a
+ * 3 bit NodeID and a 3 bit ChipID. On WSP the ChipID is always zero
+ * so we ignore it, and every where we use "chip id" in this code we
+ * mean the NodeID.
+ */
+#define WSP_ICS_CHIP_SHIFT		17
+
+
+static struct wsp_ics *ics_list;
+static int num_ics;
+
+/* ICS Source controller accessors */
+
+static u64 wsp_ics_get_xive(struct wsp_ics *ics, unsigned int irq)
+{
+	unsigned long flags;
+	u64 xive;
+
+	spin_lock_irqsave(&ics->lock, flags);
+	out_be64(IODA_TBL_ADDR_REG(ics->regs), TBL_SELECT_XIVT | IODA_IRQ(irq));
+	xive = in_be64(IODA_TBL_DATA_REG(ics->regs));
+	spin_unlock_irqrestore(&ics->lock, flags);
+
+	return xive;
+}
+
+static void wsp_ics_set_xive(struct wsp_ics *ics, unsigned int irq, u64 xive)
+{
+	xive &= ~XIVE_ADDR_MASK;
+	xive |= (irq & XIVE_ADDR_MASK);
+	xive |= XIVE_WRITE_ENABLE;
+
+	out_be64(XIVE_UPDATE_REG(ics->regs), xive);
+}
+
+static u64 xive_set_server(u64 xive, unsigned int server)
+{
+	u64 mask = ~(XIVE_SERVER_MASK << XIVE_SERVER_SHIFT);
+
+	xive &= mask;
+	xive |= (server & XIVE_SERVER_MASK) << XIVE_SERVER_SHIFT;
+
+	return xive;
+}
+
+static u64 xive_set_priority(u64 xive, unsigned int priority)
+{
+	u64 mask = ~(XIVE_PRIORITY_MASK << XIVE_PRIORITY_SHIFT);
+
+	xive &= mask;
+	xive |= (priority & XIVE_PRIORITY_MASK) << XIVE_PRIORITY_SHIFT;
+
+	return xive;
+}
+
+
+#ifdef CONFIG_SMP
+/* Find logical CPUs within mask on a given chip and store result in ret */
+void cpus_on_chip(int chip_id, cpumask_t *mask, cpumask_t *ret)
+{
+	int cpu, chip;
+	struct device_node *cpu_dn, *dn;
+	const u32 *prop;
+
+	cpumask_clear(ret);
+	for_each_cpu(cpu, mask) {
+		cpu_dn = of_get_cpu_node(cpu, NULL);
+		if (!cpu_dn)
+			continue;
+
+		prop = of_get_property(cpu_dn, "at-node", NULL);
+		if (!prop) {
+			of_node_put(cpu_dn);
+			continue;
+		}
+
+		dn = of_find_node_by_phandle(*prop);
+		of_node_put(cpu_dn);
+
+		chip = wsp_get_chip_id(dn);
+		if (chip == chip_id)
+			cpumask_set_cpu(cpu, ret);
+
+		of_node_put(dn);
+	}
+}
+
+/* Store a suitable CPU to handle a hwirq in the ics->hwirq_cpu_map cache */
+static int cache_hwirq_map(struct wsp_ics *ics, unsigned int hwirq,
+			   const cpumask_t *affinity)
+{
+	cpumask_var_t avail, newmask;
+	int ret = -ENOMEM, cpu, cpu_rover = 0, target;
+	int index = hwirq - ics->hwirq_start;
+	unsigned int nodeid;
+
+	BUG_ON(index < 0 || index >= ics->count);
+
+	if (!ics->hwirq_cpu_map)
+		return -ENOMEM;
+
+	if (!distribute_irqs) {
+		ics->hwirq_cpu_map[hwirq - ics->hwirq_start] = xics_default_server;
+		return 0;
+	}
+
+	/* Allocate needed CPU masks */
+	if (!alloc_cpumask_var(&avail, GFP_KERNEL))
+		goto ret;
+	if (!alloc_cpumask_var(&newmask, GFP_KERNEL))
+		goto freeavail;
+
+	/* Find PBus attached to the source of this IRQ */
+	nodeid = (hwirq >> WSP_ICS_CHIP_SHIFT) & 0x3; /* 12:14 */
+
+	/* Find CPUs that could handle this IRQ */
+	if (affinity)
+		cpumask_and(avail, cpu_online_mask, affinity);
+	else
+		cpumask_copy(avail, cpu_online_mask);
+
+	/* Narrow selection down to logical CPUs on the same chip */
+	cpus_on_chip(nodeid, avail, newmask);
+
+	/* Ensure we haven't narrowed it down to 0 */
+	if (unlikely(cpumask_empty(newmask))) {
+		if (unlikely(cpumask_empty(avail))) {
+			ret = -1;
+			goto out;
+		}
+		cpumask_copy(newmask, avail);
+	}
+
+	/* Choose a CPU out of those we narrowed it down to in round robin */
+	target = hwirq % cpumask_weight(newmask);
+	for_each_cpu(cpu, newmask) {
+		if (cpu_rover++ >= target) {
+			ics->hwirq_cpu_map[index] = get_hard_smp_processor_id(cpu);
+			ret = 0;
+			goto out;
+		}
+	}
+
+	/* Shouldn't happen */
+	WARN_ON(1);
+
+out:
+	free_cpumask_var(newmask);
+freeavail:
+	free_cpumask_var(avail);
+ret:
+	if (ret < 0) {
+		ics->hwirq_cpu_map[index] = cpumask_first(cpu_online_mask);
+		pr_warning("Error, falling hwirq 0x%x routing back to CPU %i\n",
+			   hwirq, ics->hwirq_cpu_map[index]);
+	}
+	return ret;
+}
+
+static void alloc_irq_map(struct wsp_ics *ics)
+{
+	int i;
+
+	ics->hwirq_cpu_map = kmalloc(sizeof(int) * ics->count, GFP_KERNEL);
+	if (!ics->hwirq_cpu_map) {
+		pr_warning("Allocate hwirq_cpu_map failed, "
+			   "IRQ balancing disabled\n");
+		return;
+	}
+
+	for (i=0; i < ics->count; i++)
+		ics->hwirq_cpu_map[i] = xics_default_server;
+}
+
+static int get_irq_server(struct wsp_ics *ics, unsigned int hwirq)
+{
+	int index = hwirq - ics->hwirq_start;
+
+	BUG_ON(index < 0 || index >= ics->count);
+
+	if (!ics->hwirq_cpu_map)
+		return xics_default_server;
+
+	return ics->hwirq_cpu_map[index];
+}
+#else /* !CONFIG_SMP */
+static int cache_hwirq_map(struct wsp_ics *ics, unsigned int hwirq,
+			   const cpumask_t *affinity)
+{
+	return 0;
+}
+
+static int get_irq_server(struct wsp_ics *ics, unsigned int hwirq)
+{
+	return xics_default_server;
+}
+
+static void alloc_irq_map(struct wsp_ics *ics) { }
+#endif
+
+static void wsp_chip_unmask_irq(struct irq_data *d)
+{
+	struct wsp_ics *ics;
+	unsigned int irq;
+	int server;
+	u64 xive;
+
+	irq = (unsigned int)irq_map[d->irq].hwirq;
+	if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+		return;
+
+	ics = d->chip_data;
+	if (WARN_ON(!ics))
+		return;
+
+	server = get_irq_server(ics, irq);
+
+	xive = wsp_ics_get_xive(ics, irq);
+	xive = xive_set_server(xive, server);
+	xive = xive_set_priority(xive, DEFAULT_PRIORITY);
+	wsp_ics_set_xive(ics, irq, xive);
+}
+
+static unsigned int wsp_chip_startup(struct irq_data *d)
+{
+	/* unmask it */
+	wsp_chip_unmask_irq(d);
+	return 0;
+}
+
+static void wsp_mask_real_irq(unsigned int irq, struct wsp_ics *ics)
+{
+	u64 xive;
+
+	if (irq == XICS_IPI)
+		return;
+
+	if (WARN_ON(!ics))
+		return;
+	xive = wsp_ics_get_xive(ics, irq);
+	xive = xive_set_server(xive, xics_default_server);
+	xive = xive_set_priority(xive, LOWEST_PRIORITY);
+	wsp_ics_set_xive(ics, irq, xive);
+}
+
+static void wsp_chip_mask_irq(struct irq_data *d)
+{
+	struct wsp_ics *ics = d->chip_data;
+	unsigned int irq;
+
+	irq = (unsigned int)irq_map[d->irq].hwirq;
+	if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+		return;
+
+	wsp_mask_real_irq(irq, ics);
+}
+
+static int wsp_chip_set_affinity(struct irq_data *d,
+				 const struct cpumask *cpumask, bool force)
+{
+	struct wsp_ics *ics;
+	unsigned int irq;
+	int ret;
+	u64 xive;
+
+	irq = (unsigned int)irq_map[d->irq].hwirq;
+	if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+		return -1;
+
+	ics = d->chip_data;
+	if (WARN_ON(!ics))
+		return -1;
+	xive = wsp_ics_get_xive(ics, irq);
+
+	/*
+	 * For the moment only implement delivery to all cpus or one cpu.
+	 * Get current irq_server for the given irq
+	 */
+	ret = cache_hwirq_map(ics, d->irq, cpumask);
+	if (ret == -1) {
+		char cpulist[128];
+		cpumask_scnprintf(cpulist, sizeof(cpulist), cpumask);
+		pr_warning("%s: No online cpus in the mask %s for irq %d\n",
+			   __func__, cpulist, d->irq);
+		return -1;
+	} else if (ret == -ENOMEM) {
+		pr_warning("%s: Out of memory\n", __func__);
+		return -1;
+	}
+
+	xive = xive_set_server(xive, get_irq_server(ics, irq));
+	wsp_ics_set_xive(ics, irq, xive);
+
+	return 0;
+}
+
+static struct irq_chip wsp_irq_chip = {
+	.name = "WSP ICS",
+	.irq_startup		= wsp_chip_startup,
+	.irq_mask		= wsp_chip_mask_irq,
+	.irq_unmask		= wsp_chip_unmask_irq,
+	.irq_set_affinity	= wsp_chip_set_affinity
+};
+
+static int wsp_ics_host_match(struct ics *ics, struct device_node *dn)
+{
+	/* All ICSs in the system implement a global irq number space,
+	 * so match against them all. */
+	return of_device_is_compatible(dn, "ibm,ppc-xics");
+}
+
+static int wsp_ics_match_hwirq(struct wsp_ics *wsp_ics, unsigned int hwirq)
+{
+	if (hwirq >= wsp_ics->hwirq_start &&
+	    hwirq <  wsp_ics->hwirq_start + wsp_ics->count)
+		return 1;
+
+	return 0;
+}
+
+static int wsp_ics_map(struct ics *ics, unsigned int virq)
+{
+	struct wsp_ics *wsp_ics = to_wsp_ics(ics);
+	unsigned long flags;
+	unsigned int hw_irq;
+
+	hw_irq = (unsigned int)irq_map[virq].hwirq;
+	if (!wsp_ics_match_hwirq(wsp_ics, hw_irq))
+		return -ENOENT;
+
+	irq_set_chip_and_handler(virq, &wsp_irq_chip, handle_fasteoi_irq);
+
+	irq_set_chip_data(virq, wsp_ics);
+
+	spin_lock_irqsave(&wsp_ics->lock, flags);
+	bitmap_allocate_region(wsp_ics->bitmap, hw_irq - wsp_ics->hwirq_start, 0);
+	spin_unlock_irqrestore(&wsp_ics->lock, flags);
+
+	return 0;
+}
+
+static void wsp_ics_mask_unknown(struct ics *ics, unsigned long hw_irq)
+{
+	struct wsp_ics *wsp_ics = to_wsp_ics(ics);
+
+	if (!wsp_ics_match_hwirq(wsp_ics, hw_irq))
+		return;
+
+	pr_err("%s: IRQ %lu (real) is invalid, disabling it.\n", __func__, hw_irq);
+	wsp_mask_real_irq(hw_irq, wsp_ics);
+}
+
+static long wsp_ics_get_server(struct ics *ics, unsigned long hw_irq)
+{
+	struct wsp_ics *wsp_ics = to_wsp_ics(ics);
+
+	if (!wsp_ics_match_hwirq(wsp_ics, hw_irq))
+		return -ENOENT;
+
+	return get_irq_server(wsp_ics, hw_irq);
+}
+
+/* HW Number allocation API */
+
+static struct wsp_ics *wsp_ics_find_dn_ics(struct device_node *dn)
+{
+	struct device_node *iparent;
+	int i;
+
+	iparent = of_irq_find_parent(dn);
+	if (!iparent) {
+		pr_err("wsp_ics: Failed to find interrupt parent!\n");
+		return NULL;
+	}
+
+	for(i = 0; i < num_ics; i++) {
+		if(ics_list[i].dn == iparent)
+			break;
+	}
+
+	if (i >= num_ics) {
+		pr_err("wsp_ics: Unable to find parent bitmap!\n");
+		return NULL;
+	}
+
+	return &ics_list[i];
+}
+
+int wsp_ics_alloc_irq(struct device_node *dn, int num)
+{
+	struct wsp_ics *ics;
+	int order, offset;
+
+	ics = wsp_ics_find_dn_ics(dn);
+	if (!ics)
+		return -ENODEV;
+
+	/* Fast, but overly strict if num isn't a power of two */
+	order = get_count_order(num);
+
+	spin_lock_irq(&ics->lock);
+	offset = bitmap_find_free_region(ics->bitmap, ics->count, order);
+	spin_unlock_irq(&ics->lock);
+
+	if (offset < 0)
+		return offset;
+
+	return offset + ics->hwirq_start;
+}
+
+void wsp_ics_free_irq(struct device_node *dn, unsigned int irq)
+{
+	struct wsp_ics *ics;
+
+	ics = wsp_ics_find_dn_ics(dn);
+	if (WARN_ON(!ics))
+		return;
+
+	spin_lock_irq(&ics->lock);
+	bitmap_release_region(ics->bitmap, irq, 0);
+	spin_unlock_irq(&ics->lock);
+}
+
+/* Initialisation */
+
+static int __init wsp_ics_bitmap_setup(struct wsp_ics *ics,
+				      struct device_node *dn)
+{
+	int len, i, j, size;
+	u32 start, count;
+	const u32 *p;
+
+	size = BITS_TO_LONGS(ics->count) * sizeof(long);
+	ics->bitmap = kzalloc(size, GFP_KERNEL);
+	if (!ics->bitmap) {
+		pr_err("wsp_ics: ENOMEM allocating IRQ bitmap!\n");
+		return -ENOMEM;
+	}
+
+	spin_lock_init(&ics->lock);
+
+	p = of_get_property(dn, "available-ranges", &len);
+	if (!p || !len) {
+		/* FIXME this should be a WARN() once mambo is updated */
+		pr_err("wsp_ics: No available-ranges defined for %s\n",
+			dn->full_name);
+		return 0;
+	}
+
+	if (len % (2 * sizeof(u32)) != 0) {
+		/* FIXME this should be a WARN() once mambo is updated */
+		pr_err("wsp_ics: Invalid available-ranges for %s\n",
+			dn->full_name);
+		return 0;
+	}
+
+	bitmap_fill(ics->bitmap, ics->count);
+
+	for (i = 0; i < len / sizeof(u32); i += 2) {
+		start = of_read_number(p + i, 1);
+		count = of_read_number(p + i + 1, 1);
+
+		pr_devel("%s: start: %d count: %d\n", __func__, start, count);
+
+		if ((start + count) > (ics->hwirq_start + ics->count) ||
+		     start < ics->hwirq_start) {
+			pr_err("wsp_ics: Invalid range! -> %d to %d\n",
+					start, start + count);
+			break;
+		}
+
+		for (j = 0; j < count; j++)
+			bitmap_release_region(ics->bitmap,
+				(start + j) - ics->hwirq_start, 0);
+	}
+
+	/* Ensure LSIs are not available for allocation */
+	bitmap_allocate_region(ics->bitmap, ics->lsi_base,
+			       get_count_order(ics->lsi_count));
+
+	return 0;
+}
+
+static int __init wsp_ics_setup(struct wsp_ics *ics, struct device_node *dn)
+{
+	u32 lsi_buid, msi_buid, msi_base, msi_count;
+	void __iomem *regs;
+	const u32 *p;
+	int rc, len, i;
+	u64 caps, buid;
+
+	p = of_get_property(dn, "interrupt-ranges", &len);
+	if (!p || len < (2 * sizeof(u32))) {
+		pr_err("wsp_ics: No/bad interrupt-ranges found on %s\n",
+			dn->full_name);
+		return -ENOENT;
+	}
+
+	if (len > (2 * sizeof(u32))) {
+		pr_err("wsp_ics: Multiple ics ranges not supported.\n");
+		return -EINVAL;
+	}
+
+	regs = of_iomap(dn, 0);
+	if (!regs) {
+		pr_err("wsp_ics: of_iomap(%s) failed\n", dn->full_name);
+		return -ENXIO;
+	}
+
+	ics->hwirq_start = of_read_number(p, 1);
+	ics->count = of_read_number(p + 1, 1);
+	ics->regs = regs;
+
+	ics->chip_id = wsp_get_chip_id(dn);
+	if (WARN_ON(ics->chip_id < 0))
+		ics->chip_id = 0;
+
+	/* Get some informations about the critter */
+	caps = in_be64(ICS_INT_CAPS_REG(ics->regs));
+	buid = in_be64(INT_SRC_LAYER_BUID_REG(ics->regs));
+	ics->lsi_count = caps >> 56;
+	msi_count = (caps >> 44) & 0x7ff;
+
+	/* Note: LSI BUID is 9 bits, but really only 3 are BUID and the
+	 * rest is mixed in the interrupt number. We store the whole
+	 * thing though
+	 */
+	lsi_buid = (buid >> 48) & 0x1ff;
+	ics->lsi_base = (ics->chip_id << WSP_ICS_CHIP_SHIFT) | lsi_buid << 5;
+	msi_buid = (buid >> 37) & 0x7;
+	msi_base = (ics->chip_id << WSP_ICS_CHIP_SHIFT) | msi_buid << 11;
+
+	pr_info("wsp_ics: Found %s\n", dn->full_name);
+	pr_info("wsp_ics:    irq range : 0x%06llx..0x%06llx\n",
+		ics->hwirq_start, ics->hwirq_start + ics->count - 1);
+	pr_info("wsp_ics:    %4d LSIs : 0x%06x..0x%06x\n",
+		ics->lsi_count, ics->lsi_base,
+		ics->lsi_base + ics->lsi_count - 1);
+	pr_info("wsp_ics:    %4d MSIs : 0x%06x..0x%06x\n",
+		msi_count, msi_base,
+		msi_base + msi_count - 1);
+
+	/* Let's check the HW config is sane */
+	if (ics->lsi_base < ics->hwirq_start ||
+	    (ics->lsi_base + ics->lsi_count) > (ics->hwirq_start + ics->count))
+		pr_warning("wsp_ics: WARNING ! LSIs out of interrupt-ranges !\n");
+	if (msi_base < ics->hwirq_start ||
+	    (msi_base + msi_count) > (ics->hwirq_start + ics->count))
+		pr_warning("wsp_ics: WARNING ! MSIs out of interrupt-ranges !\n");
+
+	/* We don't check for overlap between LSI and MSI, which will happen
+	 * if we use the same BUID, I'm not sure yet how legit that is.
+	 */
+
+	rc = wsp_ics_bitmap_setup(ics, dn);
+	if (rc) {
+		iounmap(regs);
+		return rc;
+	}
+
+	ics->dn = of_node_get(dn);
+	alloc_irq_map(ics);
+
+	for(i = 0; i < ics->count; i++)
+		wsp_mask_real_irq(ics->hwirq_start + i, ics);
+
+	ics->ics.map = wsp_ics_map;
+	ics->ics.mask_unknown = wsp_ics_mask_unknown;
+	ics->ics.get_server = wsp_ics_get_server;
+	ics->ics.host_match = wsp_ics_host_match;
+
+	xics_register_ics(&ics->ics);
+
+	return 0;
+}
+
+static void __init wsp_ics_set_default_server(void)
+{
+	struct device_node *np;
+	u32 hwid;
+
+	/* Find the server number for the boot cpu. */
+	np = of_get_cpu_node(boot_cpuid, NULL);
+	BUG_ON(!np);
+
+	hwid = get_hard_smp_processor_id(boot_cpuid);
+
+	pr_info("wsp_ics: default server is %#x, CPU %s\n", hwid, np->full_name);
+	xics_default_server = hwid;
+
+	of_node_put(np);
+}
+
+static int __init wsp_ics_init(void)
+{
+	struct device_node *dn;
+	struct wsp_ics *ics;
+	int rc, found;
+
+	wsp_ics_set_default_server();
+
+	found = 0;
+	for_each_compatible_node(dn, NULL, "ibm,ppc-xics")
+		found++;
+
+	if (found == 0) {
+		pr_err("wsp_ics: No ICS's found!\n");
+		return -ENODEV;
+	}
+
+	ics_list = kmalloc(sizeof(*ics) * found, GFP_KERNEL);
+	if (!ics_list) {
+		pr_err("wsp_ics: No memory for structs.\n");
+		return -ENOMEM;
+	}
+
+	num_ics = 0;
+	ics = ics_list;
+	for_each_compatible_node(dn, NULL, "ibm,wsp-xics") {
+		rc = wsp_ics_setup(ics, dn);
+		if (rc == 0) {
+			ics++;
+			num_ics++;
+		}
+	}
+
+	if (found != num_ics) {
+		pr_err("wsp_ics: Failed setting up %d ICS's\n",
+			found - num_ics);
+		return -1;
+	}
+
+	return 0;
+}
+
+void __init wsp_init_irq(void)
+{
+	wsp_ics_init();
+	xics_init();
+
+	/* We need to patch our irq chip's EOI to point to the right ICP */
+	wsp_irq_chip.irq_eoi = icp_ops->eoi;
+}
diff --git a/arch/powerpc/platforms/wsp/ics.h b/arch/powerpc/platforms/wsp/ics.h
new file mode 100644
index 0000000..e34d531
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/ics.h
@@ -0,0 +1,20 @@
+/*
+ * Copyright 2009 IBM Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+
+#ifndef __ICS_H
+#define __ICS_H
+
+#define XIVE_ADDR_MASK		0x7FFULL
+
+extern void wsp_init_irq(void);
+
+extern int wsp_ics_alloc_irq(struct device_node *dn, int num);
+extern void wsp_ics_free_irq(struct device_node *dn, unsigned int irq);
+
+#endif /* __ICS_H */
diff --git a/arch/powerpc/platforms/wsp/opb_pic.c b/arch/powerpc/platforms/wsp/opb_pic.c
new file mode 100644
index 0000000..4ed9a7a
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/opb_pic.c
@@ -0,0 +1,332 @@
+/*
+ * IBM Onboard Peripheral Bus Interrupt Controller
+ *
+ * Copyright 2010 Jack Miller, IBM Corporation.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/irq.h>
+#include <linux/of.h>
+#include <linux/slab.h>
+#include <linux/time.h>
+
+#include <asm/reg_a2.h>
+#include <asm/irq.h>
+
+#define OPB_NR_IRQS 32
+
+#define OPB_MLSASIER	0x04    /* MLS Accumulated Status IER */
+#define OPB_MLSIR	0x50	/* MLS Interrupt Register */
+#define OPB_MLSIER	0x54	/* MLS Interrupt Enable Register */
+#define OPB_MLSIPR	0x58	/* MLS Interrupt Polarity Register */
+#define OPB_MLSIIR	0x5c	/* MLS Interrupt Inputs Register */
+
+static int opb_index = 0;
+
+struct opb_pic {
+	struct irq_host *host;
+	void *regs;
+	int index;
+	spinlock_t lock;
+};
+
+static u32 opb_in(struct opb_pic *opb, int offset)
+{
+	return in_be32(opb->regs + offset);
+}
+
+static void opb_out(struct opb_pic *opb, int offset, u32 val)
+{
+	out_be32(opb->regs + offset, val);
+}
+
+static void opb_unmask_irq(struct irq_data *d)
+{
+	struct opb_pic *opb;
+	unsigned long flags;
+	u32 ier, bitset;
+
+	opb = d->chip_data;
+	bitset = (1 << (31 - irq_map[d->irq].hwirq));
+
+	spin_lock_irqsave(&opb->lock, flags);
+
+	ier = opb_in(opb, OPB_MLSIER);
+	opb_out(opb, OPB_MLSIER, ier | bitset);
+	ier = opb_in(opb, OPB_MLSIER);
+
+	spin_unlock_irqrestore(&opb->lock, flags);
+}
+
+static void opb_mask_irq(struct irq_data *d)
+{
+	struct opb_pic *opb;
+	unsigned long flags;
+	u32 ier, mask;
+
+	opb = d->chip_data;
+	mask = ~(1 << (31 - irq_map[d->irq].hwirq));
+
+	spin_lock_irqsave(&opb->lock, flags);
+
+	ier = opb_in(opb, OPB_MLSIER);
+	opb_out(opb, OPB_MLSIER, ier & mask);
+	ier = opb_in(opb, OPB_MLSIER); // Flush posted writes
+
+	spin_unlock_irqrestore(&opb->lock, flags);
+}
+
+static void opb_ack_irq(struct irq_data *d)
+{
+	struct opb_pic *opb;
+	unsigned long flags;
+	u32 bitset;
+
+	opb = d->chip_data;
+	bitset = (1 << (31 - irq_map[d->irq].hwirq));
+
+	spin_lock_irqsave(&opb->lock, flags);
+
+	opb_out(opb, OPB_MLSIR, bitset);
+	opb_in(opb, OPB_MLSIR); // Flush posted writes
+
+	spin_unlock_irqrestore(&opb->lock, flags);
+}
+
+static void opb_mask_ack_irq(struct irq_data *d)
+{
+	struct opb_pic *opb;
+	unsigned long flags;
+	u32 bitset;
+	u32 ier, ir;
+
+	opb = d->chip_data;
+	bitset = (1 << (31 - irq_map[d->irq].hwirq));
+
+	spin_lock_irqsave(&opb->lock, flags);
+
+	ier = opb_in(opb, OPB_MLSIER);
+	opb_out(opb, OPB_MLSIER, ier & ~bitset);
+	ier = opb_in(opb, OPB_MLSIER); // Flush posted writes
+
+	opb_out(opb, OPB_MLSIR, bitset);
+	ir = opb_in(opb, OPB_MLSIR); // Flush posted writes
+
+	spin_unlock_irqrestore(&opb->lock, flags);
+}
+
+static int opb_set_irq_type(struct irq_data *d, unsigned int flow)
+{
+	struct opb_pic *opb;
+	unsigned long flags;
+	int invert, ipr, mask, bit;
+
+	opb = d->chip_data;
+
+	/* The only information we're interested in in the type is whether it's
+	 * a high or low trigger. For high triggered interrupts, the polarity
+	 * set for it in the MLS Interrupt Polarity Register is 0, for low
+	 * interrupts it's 1 so that the proper input in the MLS Interrupt Input
+	 * Register is interrupted as asserting the interrupt. */
+
+	switch (flow) {
+		case IRQ_TYPE_NONE:
+			opb_mask_irq(d);
+			return 0;
+
+		case IRQ_TYPE_LEVEL_HIGH:
+			invert = 0;
+			break;
+
+		case IRQ_TYPE_LEVEL_LOW:
+			invert = 1;
+			break;
+
+		default:
+			return -EINVAL;
+	}
+
+	bit = (1 << (31 - irq_map[d->irq].hwirq));
+	mask = ~bit;
+
+	spin_lock_irqsave(&opb->lock, flags);
+
+	ipr = opb_in(opb, OPB_MLSIPR);
+	ipr = (ipr & mask) | (invert ? bit : 0);
+	opb_out(opb, OPB_MLSIPR, ipr);
+	ipr = opb_in(opb, OPB_MLSIPR);  // Flush posted writes
+
+	spin_unlock_irqrestore(&opb->lock, flags);
+
+	/* Record the type in the interrupt descriptor */
+	irqd_set_trigger_type(d, flow);
+
+	return 0;
+}
+
+static struct irq_chip opb_irq_chip = {
+	.name		= "OPB",
+	.irq_mask	= opb_mask_irq,
+	.irq_unmask	= opb_unmask_irq,
+	.irq_mask_ack	= opb_mask_ack_irq,
+	.irq_ack	= opb_ack_irq,
+	.irq_set_type	= opb_set_irq_type
+};
+
+static int opb_host_map(struct irq_host *host, unsigned int virq,
+		irq_hw_number_t hwirq)
+{
+	struct opb_pic *opb;
+
+	opb = host->host_data;
+
+	/* Most of the important stuff is handled by the generic host code, like
+	 * the lookup, so just attach some info to the virtual irq */
+
+	irq_set_chip_data(virq, opb);
+	irq_set_chip_and_handler(virq, &opb_irq_chip, handle_level_irq);
+	irq_set_irq_type(virq, IRQ_TYPE_NONE);
+
+	return 0;
+}
+
+static int opb_host_xlate(struct irq_host *host, struct device_node *dn,
+		const u32 *intspec, unsigned int intsize,
+		irq_hw_number_t *out_hwirq, unsigned int *out_type)
+{
+	/* Interrupt size must == 2 */
+	BUG_ON(intsize != 2);
+	*out_hwirq = intspec[0];
+	*out_type = intspec[1];
+	return 0;
+}
+
+static struct irq_host_ops opb_host_ops = {
+	.map = opb_host_map,
+	.xlate = opb_host_xlate,
+};
+
+irqreturn_t opb_irq_handler(int irq, void *private)
+{
+	struct opb_pic *opb;
+	u32 ir, src, subvirq;
+
+	opb = (struct opb_pic *) private;
+
+	/* Read the OPB MLS Interrupt Register for
+	 * asserted interrupts */
+	ir = opb_in(opb, OPB_MLSIR);
+	if (!ir)
+		return IRQ_NONE;
+
+	do {
+		/* Get 1 - 32 source, *NOT* bit */
+		src = 32 - ffs(ir);
+
+		/* Translate from the OPB's conception of interrupt number to
+		 * Linux's virtual IRQ */
+
+		subvirq = irq_linear_revmap(opb->host, src);
+
+		generic_handle_irq(subvirq);
+	} while ((ir = opb_in(opb, OPB_MLSIR)));
+
+	return IRQ_HANDLED;
+}
+
+struct opb_pic *opb_pic_init_one(struct device_node *dn)
+{
+	struct opb_pic *opb;
+	struct resource res;
+
+	if (of_address_to_resource(dn, 0, &res)) {
+		printk(KERN_ERR "opb: Couldn't translate resource\n");
+		return  NULL;
+	}
+
+	opb = kzalloc(sizeof(struct opb_pic), GFP_KERNEL);
+	if (!opb) {
+		printk(KERN_ERR "opb: Failed to allocate opb struct!\n");
+		return NULL;
+	}
+
+	/* Get access to the OPB MMIO registers */
+	opb->regs = ioremap(res.start + 0x10000, 0x1000);
+	if (!opb->regs) {
+		printk(KERN_ERR "opb: Failed to allocate register space!\n");
+		goto free_opb;
+	}
+
+	/* Allocate an irq host so that Linux knows that despite only
+	 * having one interrupt to issue, we're the controller for multiple
+	 * hardware IRQs, so later we can lookup their virtual IRQs. */
+
+	opb->host = irq_alloc_host(dn, IRQ_HOST_MAP_LINEAR,
+			OPB_NR_IRQS, &opb_host_ops, -1);
+
+	if (!opb->host) {
+		printk(KERN_ERR "opb: Failed to allocate IRQ host!\n");
+		goto free_regs;
+	}
+
+	opb->index = opb_index++;
+	spin_lock_init(&opb->lock);
+	opb->host->host_data = opb;
+
+	/* Disable all interrupts by default */
+	opb_out(opb, OPB_MLSASIER, 0);
+	opb_out(opb, OPB_MLSIER, 0);
+
+	/* ACK any interrupts left by FW */
+	opb_out(opb, OPB_MLSIR, 0xFFFFFFFF);
+
+	return opb;
+
+free_regs:
+	iounmap(opb->regs);
+free_opb:
+	kfree(opb);
+	return NULL;
+}
+
+void __init opb_pic_init(void)
+{
+	struct device_node *dn;
+	struct opb_pic *opb;
+	int virq;
+	int rc;
+
+	/* Call init_one for each OPB device */
+	for_each_compatible_node(dn, NULL, "ibm,opb") {
+
+		/* Fill in an OPB struct */
+		opb = opb_pic_init_one(dn);
+		if (!opb) {
+			printk(KERN_WARNING "opb: Failed to init node, skipped!\n");
+			continue;
+		}
+
+		/* Map / get opb's hardware virtual irq */
+		virq = irq_of_parse_and_map(dn, 0);
+		if (virq <= 0) {
+			printk("opb: irq_op_parse_and_map failed!\n");
+			continue;
+		}
+
+		/* Attach opb interrupt handler to new virtual IRQ */
+		rc = request_irq(virq, opb_irq_handler, 0, "OPB LS Cascade", opb);
+		if (rc) {
+			printk("opb: request_irq failed: %d\n", rc);
+			continue;
+		}
+
+		printk("OPB%d init with %d IRQs at %p\n", opb->index,
+				OPB_NR_IRQS, opb->regs);
+	}
+}
diff --git a/arch/powerpc/platforms/wsp/psr2.c b/arch/powerpc/platforms/wsp/psr2.c
new file mode 100644
index 0000000..40f2891
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/psr2.c
@@ -0,0 +1,95 @@
+/*
+ * Copyright 2008-2011, IBM Corporation
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/of.h>
+#include <linux/smp.h>
+
+#include <asm/machdep.h>
+#include <asm/system.h>
+#include <asm/time.h>
+#include <asm/udbg.h>
+
+#include "ics.h"
+#include "wsp.h"
+
+
+static void psr2_spin(void)
+{
+	hard_irq_disable();
+	for (;;) ;
+}
+
+static void psr2_restart(char *cmd)
+{
+	psr2_spin();
+}
+
+static int psr2_probe_devices(void)
+{
+	struct device_node *np;
+
+	/* Our RTC is a ds1500. It seems to be programatically compatible
+	 * with the ds1511 for which we have a driver so let's use that
+	 */
+	np = of_find_compatible_node(NULL, NULL, "dallas,ds1500");
+	if (np != NULL) {
+		struct resource res;
+		if (of_address_to_resource(np, 0, &res) == 0)
+			platform_device_register_simple("ds1511", 0, &res, 1);
+	}
+	return 0;
+}
+machine_arch_initcall(psr2_md, psr2_probe_devices);
+
+static void __init psr2_setup_arch(void)
+{
+	/* init to some ~sane value until calibrate_delay() runs */
+	loops_per_jiffy = 50000000;
+
+	scom_init_wsp();
+
+	/* Setup SMP callback */
+#ifdef CONFIG_SMP
+	a2_setup_smp();
+#endif
+}
+
+static int __init psr2_probe(void)
+{
+	unsigned long root = of_get_flat_dt_root();
+
+	if (!of_flat_dt_is_compatible(root, "ibm,psr2"))
+		return 0;
+
+	return 1;
+}
+
+static void __init psr2_init_irq(void)
+{
+	wsp_init_irq();
+	opb_pic_init();
+}
+
+define_machine(psr2_md) {
+	.name			= "PSR2 A2",
+	.probe			= psr2_probe,
+	.setup_arch		= psr2_setup_arch,
+	.restart		= psr2_restart,
+	.power_off		= psr2_spin,
+	.halt			= psr2_spin,
+	.calibrate_decr		= generic_calibrate_decr,
+	.init_IRQ		= psr2_init_irq,
+	.progress		= udbg_progress,
+	.power_save		= book3e_idle,
+};
diff --git a/arch/powerpc/platforms/wsp/scom_smp.c b/arch/powerpc/platforms/wsp/scom_smp.c
new file mode 100644
index 0000000..141e780
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/scom_smp.c
@@ -0,0 +1,427 @@
+/*
+ * SCOM support for A2 platforms
+ *
+ * Copyright 2007-2011 Benjamin Herrenschmidt, David Gibson,
+ *		       Michael Ellerman, IBM Corp.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/cpumask.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+
+#include <asm/cputhreads.h>
+#include <asm/reg_a2.h>
+#include <asm/scom.h>
+#include <asm/udbg.h>
+
+#include "wsp.h"
+
+#define SCOM_RAMC		0x2a		/* Ram Command */
+#define SCOM_RAMC_TGT1_EXT	0x80000000
+#define SCOM_RAMC_SRC1_EXT	0x40000000
+#define SCOM_RAMC_SRC2_EXT	0x20000000
+#define SCOM_RAMC_SRC3_EXT	0x10000000
+#define SCOM_RAMC_ENABLE	0x00080000
+#define SCOM_RAMC_THREADSEL	0x00060000
+#define SCOM_RAMC_EXECUTE	0x00010000
+#define SCOM_RAMC_MSR_OVERRIDE	0x00008000
+#define SCOM_RAMC_MSR_PR	0x00004000
+#define SCOM_RAMC_MSR_GS	0x00002000
+#define SCOM_RAMC_FORCE		0x00001000
+#define SCOM_RAMC_FLUSH		0x00000800
+#define SCOM_RAMC_INTERRUPT	0x00000004
+#define SCOM_RAMC_ERROR		0x00000002
+#define SCOM_RAMC_DONE		0x00000001
+#define SCOM_RAMI		0x29		/* Ram Instruction */
+#define SCOM_RAMIC		0x28		/* Ram Instruction and Command */
+#define SCOM_RAMIC_INSN		0xffffffff00000000
+#define SCOM_RAMD		0x2d		/* Ram Data */
+#define SCOM_RAMDH		0x2e		/* Ram Data High */
+#define SCOM_RAMDL		0x2f		/* Ram Data Low */
+#define SCOM_PCCR0		0x33		/* PC Configuration Register 0 */
+#define SCOM_PCCR0_ENABLE_DEBUG	0x80000000
+#define SCOM_PCCR0_ENABLE_RAM	0x40000000
+#define SCOM_THRCTL		0x30		/* Thread Control and Status */
+#define SCOM_THRCTL_T0_STOP	0x80000000
+#define SCOM_THRCTL_T1_STOP	0x40000000
+#define SCOM_THRCTL_T2_STOP	0x20000000
+#define SCOM_THRCTL_T3_STOP	0x10000000
+#define SCOM_THRCTL_T0_STEP	0x08000000
+#define SCOM_THRCTL_T1_STEP	0x04000000
+#define SCOM_THRCTL_T2_STEP	0x02000000
+#define SCOM_THRCTL_T3_STEP	0x01000000
+#define SCOM_THRCTL_T0_RUN	0x00800000
+#define SCOM_THRCTL_T1_RUN	0x00400000
+#define SCOM_THRCTL_T2_RUN	0x00200000
+#define SCOM_THRCTL_T3_RUN	0x00100000
+#define SCOM_THRCTL_T0_PM	0x00080000
+#define SCOM_THRCTL_T1_PM	0x00040000
+#define SCOM_THRCTL_T2_PM	0x00020000
+#define SCOM_THRCTL_T3_PM	0x00010000
+#define SCOM_THRCTL_T0_UDE	0x00008000
+#define SCOM_THRCTL_T1_UDE	0x00004000
+#define SCOM_THRCTL_T2_UDE	0x00002000
+#define SCOM_THRCTL_T3_UDE	0x00001000
+#define SCOM_THRCTL_ASYNC_DIS	0x00000800
+#define SCOM_THRCTL_TB_DIS	0x00000400
+#define SCOM_THRCTL_DEC_DIS	0x00000200
+#define SCOM_THRCTL_AND		0x31		/* Thread Control and Status */
+#define SCOM_THRCTL_OR		0x32		/* Thread Control and Status */
+
+
+static DEFINE_PER_CPU(scom_map_t, scom_ptrs);
+
+static scom_map_t get_scom(int cpu, struct device_node *np, int *first_thread)
+{
+	scom_map_t scom = per_cpu(scom_ptrs, cpu);
+	int tcpu;
+
+	if (scom_map_ok(scom)) {
+		*first_thread = 0;
+		return scom;
+	}
+
+	*first_thread = 1;
+
+	scom = scom_map_device(np, 0);
+
+	for (tcpu = cpu_first_thread_sibling(cpu);
+	     tcpu <= cpu_last_thread_sibling(cpu); tcpu++)
+		per_cpu(scom_ptrs, tcpu) = scom;
+
+	/* Hack: for the boot core, this will actually get called on
+	 * the second thread up, not the first so our test above will
+	 * set first_thread incorrectly. */
+	if (cpu_first_thread_sibling(cpu) == 0)
+		*first_thread = 0;
+
+	return scom;
+}
+
+static int a2_scom_ram(scom_map_t scom, int thread, u32 insn, int extmask)
+{
+	u64 cmd, mask, val;
+	int n = 0;
+
+	cmd = ((u64)insn << 32) | (((u64)extmask & 0xf) << 28)
+		| ((u64)thread << 17) | SCOM_RAMC_ENABLE | SCOM_RAMC_EXECUTE;
+	mask = SCOM_RAMC_DONE | SCOM_RAMC_INTERRUPT | SCOM_RAMC_ERROR;
+
+	scom_write(scom, SCOM_RAMIC, cmd);
+
+	while (!((val = scom_read(scom, SCOM_RAMC)) & mask)) {
+		pr_devel("Waiting on RAMC = 0x%llx\n", val);
+		if (++n == 3) {
+			pr_err("RAMC timeout on instruction 0x%08x, thread %d\n",
+			       insn, thread);
+			return -1;
+		}
+	}
+
+	if (val & SCOM_RAMC_INTERRUPT) {
+		pr_err("RAMC interrupt on instruction 0x%08x, thread %d\n",
+		       insn, thread);
+		return -SCOM_RAMC_INTERRUPT;
+	}
+
+	if (val & SCOM_RAMC_ERROR) {
+		pr_err("RAMC error on instruction 0x%08x, thread %d\n",
+		       insn, thread);
+		return -SCOM_RAMC_ERROR;
+	}
+
+	return 0;
+}
+
+static int a2_scom_getgpr(scom_map_t scom, int thread, int gpr, int alt,
+			  u64 *out_gpr)
+{
+	int rc;
+
+	/* or rN, rN, rN */
+	u32 insn = 0x7c000378 | (gpr << 21) | (gpr << 16) | (gpr << 11);
+	rc = a2_scom_ram(scom, thread, insn, alt ? 0xf : 0x0);
+	if (rc)
+		return rc;
+
+	*out_gpr = scom_read(scom, SCOM_RAMD);
+
+	return 0;
+}
+
+static int a2_scom_getspr(scom_map_t scom, int thread, int spr, u64 *out_spr)
+{
+	int rc, sprhi, sprlo;
+	u32 insn;
+
+	sprhi = spr >> 5;
+	sprlo = spr & 0x1f;
+	insn = 0x7c2002a6 | (sprlo << 16) | (sprhi << 11); /* mfspr r1,spr */
+
+	if (spr == 0x0ff0)
+		insn = 0x7c2000a6; /* mfmsr r1 */
+
+	rc = a2_scom_ram(scom, thread, insn, 0xf);
+	if (rc)
+		return rc;
+	return a2_scom_getgpr(scom, thread, 1, 1, out_spr);
+}
+
+static int a2_scom_setgpr(scom_map_t scom, int thread, int gpr,
+			  int alt, u64 val)
+{
+	u32 lis = 0x3c000000 | (gpr << 21);
+	u32 li = 0x38000000 | (gpr << 21);
+	u32 oris = 0x64000000 | (gpr << 21) | (gpr << 16);
+	u32 ori = 0x60000000 | (gpr << 21) | (gpr << 16);
+	u32 rldicr32 = 0x780007c6 | (gpr << 21) | (gpr << 16);
+	u32 highest = val >> 48;
+	u32 higher = (val >> 32) & 0xffff;
+	u32 high = (val >> 16) & 0xffff;
+	u32 low = val & 0xffff;
+	int lext = alt ? 0x8 : 0x0;
+	int oext = alt ? 0xf : 0x0;
+	int rc = 0;
+
+	if (highest)
+		rc |= a2_scom_ram(scom, thread, lis | highest, lext);
+
+	if (higher) {
+		if (highest)
+			rc |= a2_scom_ram(scom, thread, oris | higher, oext);
+		else
+			rc |= a2_scom_ram(scom, thread, li | higher, lext);
+	}
+
+	if (highest || higher)
+		rc |= a2_scom_ram(scom, thread, rldicr32, oext);
+
+	if (high) {
+		if (highest || higher)
+			rc |= a2_scom_ram(scom, thread, oris | high, oext);
+		else
+			rc |= a2_scom_ram(scom, thread, lis | high, lext);
+	}
+
+	if (highest || higher || high)
+		rc |= a2_scom_ram(scom, thread, ori | low, oext);
+	else
+		rc |= a2_scom_ram(scom, thread, li | low, lext);
+
+	return rc;
+}
+
+static int a2_scom_setspr(scom_map_t scom, int thread, int spr, u64 val)
+{
+	int sprhi = spr >> 5;
+	int sprlo = spr & 0x1f;
+	/* mtspr spr, r1 */
+	u32 insn = 0x7c2003a6 | (sprlo << 16) | (sprhi << 11);
+
+	if (spr == 0x0ff0)
+		insn = 0x7c200124; /* mtmsr r1 */
+
+	if (a2_scom_setgpr(scom, thread, 1, 1, val))
+		return -1;
+
+	return a2_scom_ram(scom, thread, insn, 0xf);
+}
+
+static int a2_scom_initial_tlb(scom_map_t scom, int thread)
+{
+	extern u32 a2_tlbinit_code_start[], a2_tlbinit_code_end[];
+	extern u32 a2_tlbinit_after_iprot_flush[];
+	extern u32 a2_tlbinit_after_linear_map[];
+	u32 assoc, entries, i;
+	u64 epn, tlbcfg;
+	u32 *p;
+	int rc;
+
+	/* Invalidate all entries (including iprot) */
+
+	rc = a2_scom_getspr(scom, thread, SPRN_TLB0CFG, &tlbcfg);
+	if (rc)
+		goto scom_fail;
+	entries = tlbcfg & TLBnCFG_N_ENTRY;
+	assoc = (tlbcfg & TLBnCFG_ASSOC) >> 24;
+	epn = 0;
+
+	/* Set MMUCR2 to enable 4K, 64K, 1M, 16M and 1G pages */
+	a2_scom_setspr(scom, thread, SPRN_MMUCR2, 0x000a7531);
+	/* Set MMUCR3 to write all thids bit to the TLB */
+	a2_scom_setspr(scom, thread, SPRN_MMUCR3, 0x0000000f);
+
+	/* Set MAS1 for 1G page size, and MAS2 to our initial EPN */
+	a2_scom_setspr(scom, thread, SPRN_MAS1, MAS1_TSIZE(BOOK3E_PAGESZ_1GB));
+	a2_scom_setspr(scom, thread, SPRN_MAS2, epn);
+	for (i = 0; i < entries; i++) {
+
+		a2_scom_setspr(scom, thread, SPRN_MAS0, MAS0_ESEL(i % assoc));
+
+		/* tlbwe */
+		rc = a2_scom_ram(scom, thread, 0x7c0007a4, 0);
+		if (rc)
+			goto scom_fail;
+
+		/* Next entry is new address? */
+		if((i + 1) % assoc == 0) {
+			epn += (1 << 30);
+			a2_scom_setspr(scom, thread, SPRN_MAS2, epn);
+		}
+	}
+
+	/* Setup args for linear mapping */
+	rc = a2_scom_setgpr(scom, thread, 3, 0, MAS0_TLBSEL(0));
+	if (rc)
+		goto scom_fail;
+
+	/* Linear mapping */
+	for (p = a2_tlbinit_code_start; p < a2_tlbinit_after_linear_map; p++) {
+		rc = a2_scom_ram(scom, thread, *p, 0);
+		if (rc)
+			goto scom_fail;
+	}
+
+	/*
+	 * For the boot thread, between the linear mapping and the debug
+	 * mappings there is a loop to flush iprot mappings. Ramming doesn't do
+	 * branches, but the secondary threads don't need to be nearly as smart
+	 * (i.e. we don't need to worry about invalidating the mapping we're
+	 * standing on).
+	 */
+
+	/* Debug mappings. Expects r11 = MAS0 from linear map (set above) */
+	for (p = a2_tlbinit_after_iprot_flush; p < a2_tlbinit_code_end; p++) {
+		rc = a2_scom_ram(scom, thread, *p, 0);
+		if (rc)
+			goto scom_fail;
+	}
+
+scom_fail:
+	if (rc)
+		pr_err("Setting up initial TLB failed, err %d\n", rc);
+
+	if (rc == -SCOM_RAMC_INTERRUPT) {
+		/* Interrupt, dump some status */
+		int rc[10];
+		u64 iar, srr0, srr1, esr, mas0, mas1, mas2, mas7_3, mas8, ccr2;
+		rc[0] = a2_scom_getspr(scom, thread, SPRN_IAR, &iar);
+		rc[1] = a2_scom_getspr(scom, thread, SPRN_SRR0, &srr0);
+		rc[2] = a2_scom_getspr(scom, thread, SPRN_SRR1, &srr1);
+		rc[3] = a2_scom_getspr(scom, thread, SPRN_ESR, &esr);
+		rc[4] = a2_scom_getspr(scom, thread, SPRN_MAS0, &mas0);
+		rc[5] = a2_scom_getspr(scom, thread, SPRN_MAS1, &mas1);
+		rc[6] = a2_scom_getspr(scom, thread, SPRN_MAS2, &mas2);
+		rc[7] = a2_scom_getspr(scom, thread, SPRN_MAS7_MAS3, &mas7_3);
+		rc[8] = a2_scom_getspr(scom, thread, SPRN_MAS8, &mas8);
+		rc[9] = a2_scom_getspr(scom, thread, SPRN_A2_CCR2, &ccr2);
+		pr_err(" -> retreived IAR =0x%llx (err %d)\n", iar, rc[0]);
+		pr_err("    retreived SRR0=0x%llx (err %d)\n", srr0, rc[1]);
+		pr_err("    retreived SRR1=0x%llx (err %d)\n", srr1, rc[2]);
+		pr_err("    retreived ESR =0x%llx (err %d)\n", esr, rc[3]);
+		pr_err("    retreived MAS0=0x%llx (err %d)\n", mas0, rc[4]);
+		pr_err("    retreived MAS1=0x%llx (err %d)\n", mas1, rc[5]);
+		pr_err("    retreived MAS2=0x%llx (err %d)\n", mas2, rc[6]);
+		pr_err("    retreived MS73=0x%llx (err %d)\n", mas7_3, rc[7]);
+		pr_err("    retreived MAS8=0x%llx (err %d)\n", mas8, rc[8]);
+		pr_err("    retreived CCR2=0x%llx (err %d)\n", ccr2, rc[9]);
+	}
+
+	return rc;
+}
+
+int __devinit a2_scom_startup_cpu(unsigned int lcpu, int thr_idx,
+				  struct device_node *np)
+{
+	u64 init_iar, init_msr, init_ccr2;
+	unsigned long start_here;
+	int rc, core_setup;
+	scom_map_t scom;
+	u64 pccr0;
+
+	scom = get_scom(lcpu, np, &core_setup);
+	if (!scom) {
+		printk(KERN_ERR "Couldn't map SCOM for CPU%d\n", lcpu);
+		return -1;
+	}
+
+	pr_devel("Bringing up CPU%d using SCOM...\n", lcpu);
+
+	pccr0 = scom_read(scom, SCOM_PCCR0);
+	scom_write(scom, SCOM_PCCR0, pccr0 | SCOM_PCCR0_ENABLE_DEBUG |
+				     SCOM_PCCR0_ENABLE_RAM);
+
+	/* Stop the thead with THRCTL. If we are setting up the TLB we stop all
+	 * threads. We also disable asynchronous interrupts while RAMing.
+	 */
+	if (core_setup)
+		scom_write(scom, SCOM_THRCTL_OR,
+			      SCOM_THRCTL_T0_STOP |
+			      SCOM_THRCTL_T1_STOP |
+			      SCOM_THRCTL_T2_STOP |
+			      SCOM_THRCTL_T3_STOP |
+			      SCOM_THRCTL_ASYNC_DIS);
+	else
+		scom_write(scom, SCOM_THRCTL_OR, SCOM_THRCTL_T0_STOP >> thr_idx);
+
+	/* Flush its pipeline just in case */
+	scom_write(scom, SCOM_RAMC, ((u64)thr_idx << 17) |
+		      SCOM_RAMC_FLUSH | SCOM_RAMC_ENABLE);
+
+	a2_scom_getspr(scom, thr_idx, SPRN_IAR, &init_iar);
+	a2_scom_getspr(scom, thr_idx, 0x0ff0, &init_msr);
+	a2_scom_getspr(scom, thr_idx, SPRN_A2_CCR2, &init_ccr2);
+
+	/* Set MSR to MSR_CM (0x0ff0 is magic value for MSR_CM) */
+	rc = a2_scom_setspr(scom, thr_idx, 0x0ff0, MSR_CM);
+	if (rc) {
+		pr_err("Failed to set MSR ! err %d\n", rc);
+		return rc;
+	}
+
+	/* RAM in an sync/isync for the sake of it */
+	a2_scom_ram(scom, thr_idx, 0x7c0004ac, 0);
+	a2_scom_ram(scom, thr_idx, 0x4c00012c, 0);
+
+	if (core_setup) {
+		pr_devel("CPU%d is first thread in core, initializing TLB...\n",
+			 lcpu);
+		rc = a2_scom_initial_tlb(scom, thr_idx);
+		if (rc)
+			goto fail;
+	}
+
+	start_here = *(unsigned long *)(core_setup ? generic_secondary_smp_init
+					: generic_secondary_thread_init);
+	pr_devel("CPU%d entry point at 0x%lx...\n", lcpu, start_here);
+
+	rc |= a2_scom_setspr(scom, thr_idx, SPRN_IAR, start_here);
+	rc |= a2_scom_setgpr(scom, thr_idx, 3, 0,
+			     get_hard_smp_processor_id(lcpu));
+	/*
+	 * Tell book3e_secondary_core_init not to set up the TLB, we've
+	 * already done that.
+	 */
+	rc |= a2_scom_setgpr(scom, thr_idx, 4, 0, 1);
+
+	rc |= a2_scom_setspr(scom, thr_idx, SPRN_TENS, 0x1 << thr_idx);
+
+	scom_write(scom, SCOM_RAMC, 0);
+	scom_write(scom, SCOM_THRCTL_AND, ~(SCOM_THRCTL_T0_STOP >> thr_idx));
+	scom_write(scom, SCOM_PCCR0, pccr0);
+fail:
+	pr_devel("  SCOM initialization %s\n", rc ? "failed" : "succeeded");
+	if (rc) {
+		pr_err("Old IAR=0x%08llx MSR=0x%08llx CCR2=0x%08llx\n",
+		       init_iar, init_msr, init_ccr2);
+	}
+
+	return rc;
+}
diff --git a/arch/powerpc/platforms/wsp/scom_wsp.c b/arch/powerpc/platforms/wsp/scom_wsp.c
new file mode 100644
index 0000000..4052e22
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/scom_wsp.c
@@ -0,0 +1,77 @@
+/*
+ *  SCOM backend for WSP
+ *
+ *  Copyright 2010 Benjamin Herrenschmidt, IBM Corp.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/cpumask.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+
+#include <asm/cputhreads.h>
+#include <asm/reg_a2.h>
+#include <asm/scom.h>
+#include <asm/udbg.h>
+
+#include "wsp.h"
+
+
+static scom_map_t wsp_scom_map(struct device_node *dev, u64 reg, u64 count)
+{
+	struct resource r;
+	u64 xscom_addr;
+
+	if (!of_get_property(dev, "scom-controller", NULL)) {
+		pr_err("%s: device %s is not a SCOM controller\n",
+			__func__, dev->full_name);
+		return SCOM_MAP_INVALID;
+	}
+
+	if (of_address_to_resource(dev, 0, &r)) {
+		pr_debug("Failed to find SCOM controller address\n");
+		return 0;
+	}
+
+	/* Transform the SCOM address into an XSCOM offset */
+	xscom_addr = ((reg & 0x7f000000) >> 1) | ((reg & 0xfffff) << 3);
+
+	return (scom_map_t)ioremap(r.start + xscom_addr, count << 3);
+}
+
+static void wsp_scom_unmap(scom_map_t map)
+{
+	iounmap((void *)map);
+}
+
+static u64 wsp_scom_read(scom_map_t map, u32 reg)
+{
+	u64 __iomem *addr = (u64 __iomem *)map;
+
+	return in_be64(addr + reg);
+}
+
+static void wsp_scom_write(scom_map_t map, u32 reg, u64 value)
+{
+	u64 __iomem *addr = (u64 __iomem *)map;
+
+	return out_be64(addr + reg, value);
+}
+
+static const struct scom_controller wsp_scom_controller = {
+	.map	= wsp_scom_map,
+	.unmap	= wsp_scom_unmap,
+	.read	= wsp_scom_read,
+	.write	= wsp_scom_write
+};
+
+void scom_init_wsp(void)
+{
+	scom_init(&wsp_scom_controller);
+}
diff --git a/arch/powerpc/platforms/wsp/setup.c b/arch/powerpc/platforms/wsp/setup.c
new file mode 100644
index 0000000..11ac2f0
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/setup.c
@@ -0,0 +1,36 @@
+/*
+ * Copyright 2010 Michael Ellerman, IBM Corporation
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/kernel.h>
+#include <linux/of_platform.h>
+
+#include "wsp.h"
+
+/*
+ * Find chip-id by walking up device tree looking for ibm,wsp-chip-id property.
+ * Won't work for nodes that are not a descendant of a wsp node.
+ */
+int wsp_get_chip_id(struct device_node *dn)
+{
+	const u32 *p;
+	int rc;
+
+	/* Start looking at the specified node, not its parent */
+	dn = of_node_get(dn);
+	while (dn && !(p = of_get_property(dn, "ibm,wsp-chip-id", NULL)))
+		dn = of_get_next_parent(dn);
+
+	if (!dn)
+		return -1;
+
+	rc = *p;
+	of_node_put(dn);
+
+	return rc;
+}
diff --git a/arch/powerpc/platforms/wsp/smp.c b/arch/powerpc/platforms/wsp/smp.c
new file mode 100644
index 0000000..c7b8db9
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/smp.c
@@ -0,0 +1,87 @@
+/*
+ *  SMP Support for A2 platforms
+ *
+ *  Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ *
+ */
+
+#include <linux/cpumask.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/of.h>
+#include <linux/smp.h>
+
+#include <asm/dbell.h>
+#include <asm/machdep.h>
+#include <asm/xics.h>
+
+#include "ics.h"
+#include "wsp.h"
+
+static void __devinit smp_a2_setup_cpu(int cpu)
+{
+	doorbell_setup_this_cpu();
+
+	if (cpu != boot_cpuid)
+		xics_setup_cpu();
+}
+
+int __devinit smp_a2_kick_cpu(int nr)
+{
+	const char *enable_method;
+	struct device_node *np;
+	int thr_idx;
+
+	if (nr < 0 || nr >= NR_CPUS)
+		return -ENOENT;
+
+	np = of_get_cpu_node(nr, &thr_idx);
+	if (!np)
+		return -ENODEV;
+
+	enable_method = of_get_property(np, "enable-method", NULL);
+	pr_devel("CPU%d has enable-method: \"%s\"\n", nr, enable_method);
+
+	if (!enable_method) {
+                printk(KERN_ERR "CPU%d has no enable-method\n", nr);
+		return -ENOENT;
+	} else if (strcmp(enable_method, "ibm,a2-scom") == 0) {
+		if (a2_scom_startup_cpu(nr, thr_idx, np))
+			return -1;
+	} else {
+		printk(KERN_ERR "CPU%d: Don't understand enable-method \"%s\"\n",
+                       nr, enable_method);
+		return -EINVAL;
+	}
+
+	/*
+	 * The processor is currently spinning, waiting for the
+	 * cpu_start field to become non-zero After we set cpu_start,
+	 * the processor will continue on to secondary_start
+	 */
+	paca[nr].cpu_start = 1;
+
+	return 0;
+}
+
+static int __init smp_a2_probe(void)
+{
+	return cpus_weight(cpu_possible_map);
+}
+
+static struct smp_ops_t a2_smp_ops = {
+	.message_pass	= doorbell_message_pass,
+	.probe		= smp_a2_probe,
+	.kick_cpu	= smp_a2_kick_cpu,
+	.setup_cpu	= smp_a2_setup_cpu,
+};
+
+void __init a2_setup_smp(void)
+{
+	smp_ops = &a2_smp_ops;
+}
diff --git a/arch/powerpc/platforms/wsp/wsp.h b/arch/powerpc/platforms/wsp/wsp.h
new file mode 100644
index 0000000..7c3e087
--- /dev/null
+++ b/arch/powerpc/platforms/wsp/wsp.h
@@ -0,0 +1,17 @@
+#ifndef __WSP_H
+#define __WSP_H
+
+#include <asm/wsp.h>
+
+extern void wsp_setup_pci(void);
+extern void scom_init_wsp(void);
+
+extern void a2_setup_smp(void);
+extern int a2_scom_startup_cpu(unsigned int lcpu, int thr_idx,
+			       struct device_node *np);
+int smp_a2_cpu_bootable(unsigned int nr);
+int __devinit smp_a2_kick_cpu(int nr);
+
+void opb_pic_init(void);
+
+#endif /*  __WSP_H */
diff --git a/arch/powerpc/sysdev/xics/icp-native.c b/arch/powerpc/sysdev/xics/icp-native.c
index be5e3d7..efc79e1 100644
--- a/arch/powerpc/sysdev/xics/icp-native.c
+++ b/arch/powerpc/sysdev/xics/icp-native.c
@@ -7,6 +7,7 @@
  *  2 of the License, or (at your option) any later version.
  *
  */
+
 #include <linux/types.h>
 #include <linux/kernel.h>
 #include <linux/irq.h>
-- 
1.7.1

Re: [PATCH 14/15] powerpc: Add WSP platform

[Kumar Gala]

On Apr 15, 2011, at 3:32 AM, Michael Ellerman wrote:

> diff --git a/arch/powerpc/kernel/exceptions-64e.S =
b/arch/powerpc/kernel/exceptions-64e.S
> index ab5c4dd..8e5d2c1 100644
> --- a/arch/powerpc/kernel/exceptions-64e.S
> +++ b/arch/powerpc/kernel/exceptions-64e.S
> @@ -864,6 +864,20 @@ have_hes:
> 	 * that will have to be made dependent on whether we are running =
under
> 	 * a hypervisor I suppose.
> 	 */
> +
> +	/* BEWARE, MAGIC
> +	 * This code is called as an ordinary function on the boot CPU. =
But to
> +	 * avoid duplication, this code is also used in SCOM bringup of
> +	 * secondary CPUs. We read the code between the =
initial_tlb_code_start
> +	 * and initial_tlb_code_end labels one instruction at a time and =
RAM it
> +	 * into the new core via SCOM. That doesn't process branches, so =
there
> +	 * must be none between those two labels. It also means if this =
code
> +	 * ever takes any parameters, the SCOM code must also be updated =
to
> +	 * provide them.
> +	 */
> +	.globl a2_tlbinit_code_start
> +a2_tlbinit_code_start:
> +
> 	ori	r11,r3,MAS0_WQ_ALLWAYS
> 	oris	r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 =
erratum 376 */
> 	mtspr	SPRN_MAS0,r11
> @@ -880,6 +894,9 @@ have_hes:
> 	/* Write the TLB entry */
> 	tlbwe
>=20
> +	.globl a2_tlbinit_after_linear_map
> +a2_tlbinit_after_linear_map:
> +
> 	/* Now we branch the new virtual address mapped by this entry */
> 	LOAD_REG_IMMEDIATE(r3,1f)
> 	mtctr	r3
> @@ -931,10 +948,16 @@ have_hes:
> 	cmpw	r3,r9
> 	blt	2b
>=20
> +	.globl  a2_tlbinit_after_iprot_flush
> +a2_tlbinit_after_iprot_flush:
> +
> 	PPC_TLBILX(0,0,0)
> 	sync
> 	isync
>=20
> +	.globl a2_tlbinit_code_end
> +a2_tlbinit_code_end:
> +
> 	/* We translate LR and return */
> 	mflr	r3
> 	tovirt(r3,r3)

Can this be moved into kernel/cpu_setup_a2.S.  This and the debug patch =
that follows are bit ugly.

- k=

Re: [PATCH 14/15] powerpc: Add WSP platform

[Benjamin Herrenschmidt]

On Mon, 2011-04-18 at 07:46 -0500, Kumar Gala wrote:
> 
> Can this be moved into kernel/cpu_setup_a2.S.  This and the debug
> patch that follows are bit ugly.

Well, the code per-se is not really A2 specific... the labels and
comments are :-)

Cheers,
Ben.

[PATCH 15/15] powerpc: Add early debug for WSP platforms

[Michael Ellerman]

From: Jack Miller <jack@codezen.org>

Signed-off-by: Jack Miller <jack@codezen.org>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 arch/powerpc/Kconfig.debug           |    5 +++
 arch/powerpc/include/asm/reg_a2.h    |    9 ++++++
 arch/powerpc/include/asm/udbg.h      |    1 +
 arch/powerpc/kernel/exceptions-64e.S |   17 +++++++++++
 arch/powerpc/kernel/udbg.c           |    2 +
 arch/powerpc/kernel/udbg_16550.c     |   51 ++++++++++++++++++++++++++++++++++
 6 files changed, 85 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/Kconfig.debug b/arch/powerpc/Kconfig.debug
index 2d38a50..a597dd7 100644
--- a/arch/powerpc/Kconfig.debug
+++ b/arch/powerpc/Kconfig.debug
@@ -267,6 +267,11 @@ config PPC_EARLY_DEBUG_USBGECKO
 	  Select this to enable early debugging for Nintendo GameCube/Wii
 	  consoles via an external USB Gecko adapter.
 
+config PPC_EARLY_DEBUG_WSP
+	bool "Early debugging via WSP's internal UART"
+	depends on PPC_WSP
+	select PPC_UDBG_16550
+
 endchoice
 
 config PPC_EARLY_DEBUG_44x_PHYSLOW
diff --git a/arch/powerpc/include/asm/reg_a2.h b/arch/powerpc/include/asm/reg_a2.h
index 3ba9c6f..3d52a11 100644
--- a/arch/powerpc/include/asm/reg_a2.h
+++ b/arch/powerpc/include/asm/reg_a2.h
@@ -110,6 +110,15 @@
 #define TLB1_UR			ASM_CONST(0x0000000000000002)
 #define TLB1_SR			ASM_CONST(0x0000000000000001)
 
+#ifdef CONFIG_PPC_EARLY_DEBUG_WSP
+#define WSP_UART_PHYS	0xffc000c000
+/* This needs to be careful chosen to hit a !0 congruence class
+ * in the TLB since we bolt it in way 3, which is already occupied
+ * by our linear mapping primary bolted entry in CC 0.
+ */
+#define WSP_UART_VIRT	0xf000000000001000
+#endif
+
 /* A2 erativax attributes definitions */
 #define ERATIVAX_RS_IS_ALL		0x000
 #define ERATIVAX_RS_IS_TID		0x040
diff --git a/arch/powerpc/include/asm/udbg.h b/arch/powerpc/include/asm/udbg.h
index 11ae699..58580e9 100644
--- a/arch/powerpc/include/asm/udbg.h
+++ b/arch/powerpc/include/asm/udbg.h
@@ -52,6 +52,7 @@ extern void __init udbg_init_44x_as1(void);
 extern void __init udbg_init_40x_realmode(void);
 extern void __init udbg_init_cpm(void);
 extern void __init udbg_init_usbgecko(void);
+extern void __init udbg_init_wsp(void);
 
 #endif /* __KERNEL__ */
 #endif /* _ASM_POWERPC_UDBG_H */
diff --git a/arch/powerpc/kernel/exceptions-64e.S b/arch/powerpc/kernel/exceptions-64e.S
index 8e5d2c1..f4ed2cb 100644
--- a/arch/powerpc/kernel/exceptions-64e.S
+++ b/arch/powerpc/kernel/exceptions-64e.S
@@ -17,6 +17,7 @@
 #include <asm/cputable.h>
 #include <asm/setup.h>
 #include <asm/thread_info.h>
+#include <asm/reg_a2.h>
 #include <asm/exception-64e.h>
 #include <asm/bug.h>
 #include <asm/irqflags.h>
@@ -951,6 +952,22 @@ a2_tlbinit_after_linear_map:
 	.globl  a2_tlbinit_after_iprot_flush
 a2_tlbinit_after_iprot_flush:
 
+#ifdef CONFIG_PPC_EARLY_DEBUG_WSP
+	/* Now establish early debug mappings if applicable */
+	/* Restore the MAS0 we used for linear mapping load */
+	mtspr	SPRN_MAS0,r11
+
+	lis	r3,(MAS1_VALID | MAS1_IPROT)@h
+	ori	r3,r3,(BOOK3E_PAGESZ_4K << MAS1_TSIZE_SHIFT)
+	mtspr	SPRN_MAS1,r3
+	LOAD_REG_IMMEDIATE(r3, WSP_UART_VIRT | MAS2_I | MAS2_G)
+	mtspr	SPRN_MAS2,r3
+	LOAD_REG_IMMEDIATE(r3, WSP_UART_PHYS | MAS3_SR | MAS3_SW)
+	mtspr	SPRN_MAS7_MAS3,r3
+	/* re-use the MAS8 value from the linear mapping */
+	tlbwe
+#endif /* CONFIG_PPC_EARLY_DEBUG_WSP */
+
 	PPC_TLBILX(0,0,0)
 	sync
 	isync
diff --git a/arch/powerpc/kernel/udbg.c b/arch/powerpc/kernel/udbg.c
index e39cad8..23d65ab 100644
--- a/arch/powerpc/kernel/udbg.c
+++ b/arch/powerpc/kernel/udbg.c
@@ -62,6 +62,8 @@ void __init udbg_early_init(void)
 	udbg_init_cpm();
 #elif defined(CONFIG_PPC_EARLY_DEBUG_USBGECKO)
 	udbg_init_usbgecko();
+#elif defined(CONFIG_PPC_EARLY_DEBUG_WSP)
+	udbg_init_wsp();
 #endif
 
 #ifdef CONFIG_PPC_EARLY_DEBUG
diff --git a/arch/powerpc/kernel/udbg_16550.c b/arch/powerpc/kernel/udbg_16550.c
index b4b167b..ed7ac57 100644
--- a/arch/powerpc/kernel/udbg_16550.c
+++ b/arch/powerpc/kernel/udbg_16550.c
@@ -11,6 +11,7 @@
 #include <linux/types.h>
 #include <asm/udbg.h>
 #include <asm/io.h>
+#include <asm/reg_a2.h>
 
 extern u8 real_readb(volatile u8 __iomem  *addr);
 extern void real_writeb(u8 data, volatile u8 __iomem *addr);
@@ -298,3 +299,53 @@ void __init udbg_init_40x_realmode(void)
 	udbg_getc_poll = NULL;
 }
 #endif /* CONFIG_PPC_EARLY_DEBUG_40x */
+
+#ifdef CONFIG_PPC_EARLY_DEBUG_WSP
+static void udbg_wsp_flush(void)
+{
+	if (udbg_comport) {
+		while ((readb(&udbg_comport->lsr) & LSR_THRE) == 0)
+			/* wait for idle */;
+	}
+}
+
+static void udbg_wsp_putc(char c)
+{
+	if (udbg_comport) {
+		if (c == '\n')
+			udbg_wsp_putc('\r');
+		udbg_wsp_flush();
+		writeb(c, &udbg_comport->thr); eieio();
+	}
+}
+
+static int udbg_wsp_getc(void)
+{
+	if (udbg_comport) {
+		while ((readb(&udbg_comport->lsr) & LSR_DR) == 0)
+			; /* wait for char */
+		return readb(&udbg_comport->rbr);
+	}
+	return -1;
+}
+
+static int udbg_wsp_getc_poll(void)
+{
+	if (udbg_comport)
+		if (readb(&udbg_comport->lsr) & LSR_DR)
+			return readb(&udbg_comport->rbr);
+	return -1;
+}
+
+void __init udbg_init_wsp(void)
+{
+	udbg_comport = (struct NS16550 __iomem *)WSP_UART_VIRT;
+
+	udbg_init_uart(udbg_comport, 57600, 50000000);
+
+	udbg_putc = udbg_wsp_putc;
+	udbg_flush = udbg_wsp_flush;
+	udbg_getc = udbg_wsp_getc;
+	udbg_getc_poll = udbg_wsp_getc_poll;
+}
+#endif /* CONFIG_PPC_EARLY_DEBUG_WSP */
-- 
1.7.1