Merge branch 'for-2.6.22' of master.kernel.org:/pub/scm/linux/kernel/git/arnd/cell-2.6 into for-2.6.22

5 files changed, 530 insertions, 58 deletions

diff --git a/arch/powerpc/platforms/cell/cbe_cpufreq.c b/arch/powerpc/platforms/cell/cbe_cpufreq.c
index 9c5d63b7e76c..4495973bff59 100644
--- a/arch/powerpc/platforms/cell/cbe_cpufreq.c
+++ b/arch/powerpc/platforms/cell/cbe_cpufreq.c
@@ -28,6 +28,8 @@
 #include <asm/processor.h>
 #include <asm/prom.h>
 #include <asm/time.h>
+#include <asm/pmi.h>
+#include <asm/of_platform.h>
 
 #include "cbe_regs.h"
 
@@ -68,6 +70,38 @@ static u64 MIC_Slow_Next_Timer_table[] = {
  * hardware specific functions
  */
 
+static struct of_device *pmi_dev;
+
+static int set_pmode_pmi(int cpu, unsigned int pmode)
+{
+	int ret;
+	pmi_message_t pmi_msg;
+#ifdef DEBUG
+	u64 time;
+#endif
+
+	pmi_msg.type = PMI_TYPE_FREQ_CHANGE;
+	pmi_msg.data1 =	cbe_cpu_to_node(cpu);
+	pmi_msg.data2 = pmode;
+
+#ifdef DEBUG
+	time = (u64) get_cycles();
+#endif
+
+	pmi_send_message(pmi_dev, pmi_msg);
+	ret = pmi_msg.data2;
+
+	pr_debug("PMI returned slow mode %d\n", ret);
+
+#ifdef DEBUG
+	time = (u64) get_cycles() - time; /* actual cycles (not cpu cycles!) */
+	time = 1000000000 * time / CLOCK_TICK_RATE; /* time in ns (10^-9) */
+	pr_debug("had to wait %lu ns for a transition\n", time);
+#endif
+	return ret;
+}
+
+
 static int get_pmode(int cpu)
 {
 	int ret;
@@ -79,7 +113,7 @@ static int get_pmode(int cpu)
 	return ret;
 }
 
-static int set_pmode(int cpu, unsigned int pmode)
+static int set_pmode_reg(int cpu, unsigned int pmode)
 {
 	struct cbe_pmd_regs __iomem *pmd_regs;
 	struct cbe_mic_tm_regs __iomem *mic_tm_regs;
@@ -120,6 +154,39 @@ static int set_pmode(int cpu, unsigned int pmode)
 	return 0;
 }
 
+static int set_pmode(int cpu, unsigned int slow_mode) {
+	if(pmi_dev)
+		return set_pmode_pmi(cpu, slow_mode);
+	else
+		return set_pmode_reg(cpu, slow_mode);
+}
+
+static void cbe_cpufreq_handle_pmi(struct of_device *dev, pmi_message_t pmi_msg)
+{
+	struct cpufreq_policy policy;
+	u8 cpu;
+	u8 cbe_pmode_new;
+
+	BUG_ON (pmi_msg.type != PMI_TYPE_FREQ_CHANGE);
+
+	cpu = cbe_node_to_cpu(pmi_msg.data1);
+	cbe_pmode_new = pmi_msg.data2;
+
+	cpufreq_get_policy(&policy, cpu);
+
+	policy.max = min(policy.max, cbe_freqs[cbe_pmode_new].frequency);
+	policy.min = min(policy.min, policy.max);
+
+	pr_debug("cbe_handle_pmi: new policy.min=%d policy.max=%d\n", policy.min, policy.max);
+	cpufreq_set_policy(&policy);
+}
+
+static struct pmi_handler cbe_pmi_handler = {
+	.type			= PMI_TYPE_FREQ_CHANGE,
+	.handle_pmi_message	= cbe_cpufreq_handle_pmi,
+};
+
+
 /*
  * cpufreq functions
  */
@@ -234,11 +301,23 @@ static struct cpufreq_driver cbe_cpufreq_driver = {
 
 static int __init cbe_cpufreq_init(void)
 {
+	struct device_node *np;
+
+	np = of_find_node_by_type(NULL, "ibm,pmi");
+
+	pmi_dev = of_find_device_by_node(np);
+
+	if (pmi_dev)
+		pmi_register_handler(pmi_dev, &cbe_pmi_handler);
+
 	return cpufreq_register_driver(&cbe_cpufreq_driver);
 }
 
 static void __exit cbe_cpufreq_exit(void)
 {
+	if(pmi_dev)
+		pmi_unregister_handler(pmi_dev, &cbe_pmi_handler);
+
 	cpufreq_unregister_driver(&cbe_cpufreq_driver);
 }
 
diff --git a/arch/powerpc/platforms/cell/cbe_regs.c b/arch/powerpc/platforms/cell/cbe_regs.c
index 7c94af4ac439..12c9674b4b1f 100644
--- a/arch/powerpc/platforms/cell/cbe_regs.c
+++ b/arch/powerpc/platforms/cell/cbe_regs.c
@@ -14,6 +14,8 @@
 #include <asm/pgtable.h>
 #include <asm/prom.h>
 #include <asm/ptrace.h>
+#include <asm/of_device.h>
+#include <asm/of_platform.h>
 
 #include "cbe_regs.h"
 
@@ -27,6 +29,7 @@
 static struct cbe_regs_map
 {
 	struct device_node *cpu_node;
+	struct device_node *be_node;
 	struct cbe_pmd_regs __iomem *pmd_regs;
 	struct cbe_iic_regs __iomem *iic_regs;
 	struct cbe_mic_tm_regs __iomem *mic_tm_regs;
@@ -37,30 +40,43 @@ static int cbe_regs_map_count;
 static struct cbe_thread_map
 {
 	struct device_node *cpu_node;
+	struct device_node *be_node;
 	struct cbe_regs_map *regs;
+	unsigned int thread_id;
+	unsigned int cbe_id;
 } cbe_thread_map[NR_CPUS];
 
+static cpumask_t cbe_local_mask[MAX_CBE] = { [0 ... MAX_CBE-1] = CPU_MASK_NONE };
+static cpumask_t cbe_first_online_cpu = CPU_MASK_NONE;
+
 static struct cbe_regs_map *cbe_find_map(struct device_node *np)
 {
 	int i;
 	struct device_node *tmp_np;
 
-	if (strcasecmp(np->type, "spe") == 0) {
-		if (np->data == NULL) {
-			/* walk up path until cpu node was found */
-			tmp_np = np->parent;
-			while (tmp_np != NULL && strcasecmp(tmp_np->type, "cpu") != 0)
-				tmp_np = tmp_np->parent;
+	if (strcasecmp(np->type, "spe")) {
+		for (i = 0; i < cbe_regs_map_count; i++)
+			if (cbe_regs_maps[i].cpu_node == np ||
+			    cbe_regs_maps[i].be_node == np)
+				return &cbe_regs_maps[i];
+		return NULL;
+	}
 
-			np->data = cbe_find_map(tmp_np);
-		}
+	if (np->data)
 		return np->data;
-	}
 
-	for (i = 0; i < cbe_regs_map_count; i++)
-		if (cbe_regs_maps[i].cpu_node == np)
-			return &cbe_regs_maps[i];
-	return NULL;
+	/* walk up path until cpu or be node was found */
+	tmp_np = np;
+	do {
+		tmp_np = tmp_np->parent;
+		/* on a correct devicetree we wont get up to root */
+		BUG_ON(!tmp_np);
+	} while (strcasecmp(tmp_np->type, "cpu") &&
+		 strcasecmp(tmp_np->type, "be"));
+
+	np->data = cbe_find_map(tmp_np);
+
+	return np->data;
 }
 
 struct cbe_pmd_regs __iomem *cbe_get_pmd_regs(struct device_node *np)
@@ -130,49 +146,69 @@ struct cbe_mic_tm_regs __iomem *cbe_get_cpu_mic_tm_regs(int cpu)
 }
 EXPORT_SYMBOL_GPL(cbe_get_cpu_mic_tm_regs);
 
-/* FIXME
- * This is little more than a stub at the moment.  It should be
- * fleshed out so that it works for both SMT and non-SMT, no
- * matter if the passed cpu is odd or even.
- * For SMT enabled, returns 0 for even-numbered cpu; otherwise 1.
- * For SMT disabled, returns 0 for all cpus.
- */
 u32 cbe_get_hw_thread_id(int cpu)
 {
-	return (cpu & 1);
+	return cbe_thread_map[cpu].thread_id;
 }
 EXPORT_SYMBOL_GPL(cbe_get_hw_thread_id);
 
-void __init cbe_regs_init(void)
+u32 cbe_cpu_to_node(int cpu)
 {
-	int i;
-	struct device_node *cpu;
+	return cbe_thread_map[cpu].cbe_id;
+}
+EXPORT_SYMBOL_GPL(cbe_cpu_to_node);
 
-	/* Build local fast map of CPUs */
-	for_each_possible_cpu(i)
-		cbe_thread_map[i].cpu_node = of_get_cpu_node(i, NULL);
+u32 cbe_node_to_cpu(int node)
+{
+	return find_first_bit( (unsigned long *) &cbe_local_mask[node], sizeof(cpumask_t));
+}
+EXPORT_SYMBOL_GPL(cbe_node_to_cpu);
 
-	/* Find maps for each device tree CPU */
-	for_each_node_by_type(cpu, "cpu") {
-		struct cbe_regs_map *map = &cbe_regs_maps[cbe_regs_map_count++];
+static struct device_node *cbe_get_be_node(int cpu_id)
+{
+	struct device_node *np;
+
+	for_each_node_by_type (np, "be") {
+		int len,i;
+		const phandle *cpu_handle;
+
+		cpu_handle = of_get_property(np, "cpus", &len);
+
+		for (i=0; i<len; i++)
+			if (of_find_node_by_phandle(cpu_handle[i]) == of_get_cpu_node(cpu_id, NULL))
+				return np;
+	}
+
+	return NULL;
+}
+
+void __init cbe_fill_regs_map(struct cbe_regs_map *map)
+{
+	if(map->be_node) {
+		struct device_node *be, *np;
+
+		be = map->be_node;
 
+		for_each_node_by_type(np, "pervasive")
+			if (of_get_parent(np) == be)
+				map->pmd_regs = of_iomap(np, 0);
+
+		for_each_node_by_type(np, "CBEA-Internal-Interrupt-Controller")
+			if (of_get_parent(np) == be)
+				map->iic_regs = of_iomap(np, 2);
+
+		for_each_node_by_type(np, "mic-tm")
+			if (of_get_parent(np) == be)
+				map->mic_tm_regs = of_iomap(np, 0);
+	} else {
+		struct device_node *cpu;
 		/* That hack must die die die ! */
 		const struct address_prop {
 			unsigned long address;
 			unsigned int len;
 		} __attribute__((packed)) *prop;
 
-
-		if (cbe_regs_map_count > MAX_CBE) {
-			printk(KERN_ERR "cbe_regs: More BE chips than supported"
-			       "!\n");
-			cbe_regs_map_count--;
-			return;
-		}
-		map->cpu_node = cpu;
-		for_each_possible_cpu(i)
-			if (cbe_thread_map[i].cpu_node == cpu)
-				cbe_thread_map[i].regs = map;
+		cpu = map->cpu_node;
 
 		prop = of_get_property(cpu, "pervasive", NULL);
 		if (prop != NULL)
@@ -188,3 +224,50 @@ void __init cbe_regs_init(void)
 	}
 }
 
+
+void __init cbe_regs_init(void)
+{
+	int i;
+	unsigned int thread_id;
+	struct device_node *cpu;
+
+	/* Build local fast map of CPUs */
+	for_each_possible_cpu(i) {
+		cbe_thread_map[i].cpu_node = of_get_cpu_node(i, &thread_id);
+		cbe_thread_map[i].be_node = cbe_get_be_node(i);
+		cbe_thread_map[i].thread_id = thread_id;
+	}
+
+	/* Find maps for each device tree CPU */
+	for_each_node_by_type(cpu, "cpu") {
+		struct cbe_regs_map *map;
+		unsigned int cbe_id;
+
+		cbe_id = cbe_regs_map_count++;
+		map = &cbe_regs_maps[cbe_id];
+
+		if (cbe_regs_map_count > MAX_CBE) {
+			printk(KERN_ERR "cbe_regs: More BE chips than supported"
+			       "!\n");
+			cbe_regs_map_count--;
+			return;
+		}
+		map->cpu_node = cpu;
+
+		for_each_possible_cpu(i) {
+			struct cbe_thread_map *thread = &cbe_thread_map[i];
+
+			if (thread->cpu_node == cpu) {
+				thread->regs = map;
+				thread->cbe_id = cbe_id;
+				map->be_node = thread->be_node;
+				cpu_set(i, cbe_local_mask[cbe_id]);
+				if(thread->thread_id == 0)
+					cpu_set(i, cbe_first_online_cpu);
+			}
+		}
+
+		cbe_fill_regs_map(map);
+	}
+}
+
diff --git a/arch/powerpc/platforms/cell/cbe_regs.h b/arch/powerpc/platforms/cell/cbe_regs.h
index 440a7ecc66ea..17d597144877 100644
--- a/arch/powerpc/platforms/cell/cbe_regs.h
+++ b/arch/powerpc/platforms/cell/cbe_regs.h
@@ -255,6 +255,11 @@ struct cbe_mic_tm_regs {
 extern struct cbe_mic_tm_regs __iomem *cbe_get_mic_tm_regs(struct device_node *np);
 extern struct cbe_mic_tm_regs __iomem *cbe_get_cpu_mic_tm_regs(int cpu);
 
+/* some utility functions to deal with SMT */
+extern u32 cbe_get_hw_thread_id(int cpu);
+extern u32 cbe_cpu_to_node(int cpu);
+extern u32 cbe_node_to_cpu(int node);
+
 /* Init this module early */
 extern void cbe_regs_init(void);
 
diff --git a/arch/powerpc/platforms/cell/cbe_thermal.c b/arch/powerpc/platforms/cell/cbe_thermal.c
index e8bcd2a767ce..f370f0fa6f4c 100644
--- a/arch/powerpc/platforms/cell/cbe_thermal.c
+++ b/arch/powerpc/platforms/cell/cbe_thermal.c
@@ -1,6 +1,31 @@
 /*
  * thermal support for the cell processor
  *
+ * This module adds some sysfs attributes to cpu and spu nodes.
+ * Base for measurements are the digital thermal sensors (DTS)
+ * located on the chip.
+ * The accuracy is 2 degrees, starting from 65 up to 125 degrees celsius
+ * The attributes can be found under
+ * /sys/devices/system/cpu/cpuX/thermal
+ * /sys/devices/system/spu/spuX/thermal
+ *
+ * The following attributes are added for each node:
+ * temperature:
+ *	contains the current temperature measured by the DTS
+ * throttle_begin:
+ *	throttling begins when temperature is greater or equal to
+ *	throttle_begin. Setting this value to 125 prevents throttling.
+ * throttle_end:
+ *	throttling is being ceased, if the temperature is lower than
+ *	throttle_end. Due to a delay between applying throttling and
+ *	a reduced temperature this value should be less than throttle_begin.
+ *	A value equal to throttle_begin provides only a very little hysteresis.
+ * throttle_full_stop:
+ *	If the temperatrue is greater or equal to throttle_full_stop,
+ *	full throttling is applied to the cpu or spu. This value should be
+ *	greater than throttle_begin and throttle_end. Setting this value to
+ *	65 prevents the unit from running code at all.
+ *
  * (C) Copyright IBM Deutschland Entwicklung GmbH 2005
  *
  * Author: Christian Krafft <krafft@de.ibm.com>
@@ -31,6 +56,26 @@
 #include "cbe_regs.h"
 #include "spu_priv1_mmio.h"
 
+#define TEMP_MIN 65
+#define TEMP_MAX 125
+
+#define SYSDEV_PREFIX_ATTR(_prefix,_name,_mode)			\
+struct sysdev_attribute attr_ ## _prefix ## _ ## _name = {	\
+	.attr = { .name = __stringify(_name), .mode = _mode },	\
+	.show	= _prefix ## _show_ ## _name,			\
+	.store	= _prefix ## _store_ ## _name,			\
+};
+
+static inline u8 reg_to_temp(u8 reg_value)
+{
+	return ((reg_value & 0x3f) << 1) + TEMP_MIN;
+}
+
+static inline u8 temp_to_reg(u8 temp)
+{
+	return ((temp - TEMP_MIN) >> 1) & 0x3f;
+}
+
 static struct cbe_pmd_regs __iomem *get_pmd_regs(struct sys_device *sysdev)
 {
 	struct spu *spu;
@@ -58,20 +103,81 @@ static u8 spu_read_register_value(struct sys_device *sysdev, union spe_reg __iom
 
 static ssize_t spu_show_temp(struct sys_device *sysdev, char *buf)
 {
-	int value;
+	u8 value;
 	struct cbe_pmd_regs __iomem *pmd_regs;
 
 	pmd_regs = get_pmd_regs(sysdev);
 
 	value = spu_read_register_value(sysdev, &pmd_regs->ts_ctsr1);
-	/* clear all other bits */
+
+	return sprintf(buf, "%d\n", reg_to_temp(value));
+}
+
+static ssize_t show_throttle(struct cbe_pmd_regs __iomem *pmd_regs, char *buf, int pos)
+{
+	u64 value;
+
+	value = in_be64(&pmd_regs->tm_tpr.val);
+	/* access the corresponding byte */
+	value >>= pos;
 	value &= 0x3F;
-	/* temp is stored in steps of 2 degrees */
-	value *= 2;
-	/* base temp is 65 degrees */
-	value += 65;
 
-	return sprintf(buf, "%d\n", (int) value);
+	return sprintf(buf, "%d\n", reg_to_temp(value));
+}
+
+static ssize_t store_throttle(struct cbe_pmd_regs __iomem *pmd_regs, const char *buf, size_t size, int pos)
+{
+	u64 reg_value;
+	int temp;
+	u64 new_value;
+	int ret;
+
+	ret = sscanf(buf, "%u", &temp);
+
+	if (ret != 1 || temp < TEMP_MIN || temp > TEMP_MAX)
+		return -EINVAL;
+
+	new_value = temp_to_reg(temp);
+
+	reg_value = in_be64(&pmd_regs->tm_tpr.val);
+
+	/* zero out bits for new value */
+	reg_value &= ~(0xffull << pos);
+	/* set bits to new value */
+	reg_value |= new_value << pos;
+
+	out_be64(&pmd_regs->tm_tpr.val, reg_value);
+	return size;
+}
+
+static ssize_t spu_show_throttle_end(struct sys_device *sysdev, char *buf)
+{
+	return show_throttle(get_pmd_regs(sysdev), buf, 0);
+}
+
+static ssize_t spu_show_throttle_begin(struct sys_device *sysdev, char *buf)
+{
+	return show_throttle(get_pmd_regs(sysdev), buf, 8);
+}
+
+static ssize_t spu_show_throttle_full_stop(struct sys_device *sysdev, char *buf)
+{
+	return show_throttle(get_pmd_regs(sysdev), buf, 16);
+}
+
+static ssize_t spu_store_throttle_end(struct sys_device *sysdev, const char *buf, size_t size)
+{
+	return store_throttle(get_pmd_regs(sysdev), buf, size, 0);
+}
+
+static ssize_t spu_store_throttle_begin(struct sys_device *sysdev, const char *buf, size_t size)
+{
+	return store_throttle(get_pmd_regs(sysdev), buf, size, 8);
+}
+
+static ssize_t spu_store_throttle_full_stop(struct sys_device *sysdev, const char *buf, size_t size)
+{
+	return store_throttle(get_pmd_regs(sysdev), buf, size, 16);
 }
 
 static ssize_t ppe_show_temp(struct sys_device *sysdev, char *buf, int pos)
@@ -82,16 +188,9 @@ static ssize_t ppe_show_temp(struct sys_device *sysdev, char *buf, int pos)
 	pmd_regs = cbe_get_cpu_pmd_regs(sysdev->id);
 	value = in_be64(&pmd_regs->ts_ctsr2);
 
-	/* access the corresponding byte */
-	value >>= pos;
-	/* clear all other bits */
-	value &= 0x3F;
-	/* temp is stored in steps of 2 degrees */
-	value *= 2;
-	/* base temp is 65 degrees */
-	value += 65;
+	value = (value >> pos) & 0x3f;
 
-	return sprintf(buf, "%d\n", (int) value);
+	return sprintf(buf, "%d\n", reg_to_temp(value));
 }
 
 
@@ -108,13 +207,52 @@ static ssize_t ppe_show_temp1(struct sys_device *sysdev, char *buf)
 	return ppe_show_temp(sysdev, buf, 0);
 }
 
+static ssize_t ppe_show_throttle_end(struct sys_device *sysdev, char *buf)
+{
+	return show_throttle(cbe_get_cpu_pmd_regs(sysdev->id), buf, 32);
+}
+
+static ssize_t ppe_show_throttle_begin(struct sys_device *sysdev, char *buf)
+{
+	return show_throttle(cbe_get_cpu_pmd_regs(sysdev->id), buf, 40);
+}
+
+static ssize_t ppe_show_throttle_full_stop(struct sys_device *sysdev, char *buf)
+{
+	return show_throttle(cbe_get_cpu_pmd_regs(sysdev->id), buf, 48);
+}
+
+static ssize_t ppe_store_throttle_end(struct sys_device *sysdev, const char *buf, size_t size)
+{
+	return store_throttle(cbe_get_cpu_pmd_regs(sysdev->id), buf, size, 32);
+}
+
+static ssize_t ppe_store_throttle_begin(struct sys_device *sysdev, const char *buf, size_t size)
+{
+	return store_throttle(cbe_get_cpu_pmd_regs(sysdev->id), buf, size, 40);
+}
+
+static ssize_t ppe_store_throttle_full_stop(struct sys_device *sysdev, const char *buf, size_t size)
+{
+	return store_throttle(cbe_get_cpu_pmd_regs(sysdev->id), buf, size, 48);
+}
+
+
 static struct sysdev_attribute attr_spu_temperature = {
 	.attr = {.name = "temperature", .mode = 0400 },
 	.show = spu_show_temp,
 };
 
+static SYSDEV_PREFIX_ATTR(spu, throttle_end, 0600);
+static SYSDEV_PREFIX_ATTR(spu, throttle_begin, 0600);
+static SYSDEV_PREFIX_ATTR(spu, throttle_full_stop, 0600);
+
+
 static struct attribute *spu_attributes[] = {
 	&attr_spu_temperature.attr,
+	&attr_spu_throttle_end.attr,
+	&attr_spu_throttle_begin.attr,
+	&attr_spu_throttle_full_stop.attr,
 	NULL,
 };
 
@@ -133,9 +271,16 @@ static struct sysdev_attribute attr_ppe_temperature1 = {
 	.show = ppe_show_temp1,
 };
 
+static SYSDEV_PREFIX_ATTR(ppe, throttle_end, 0600);
+static SYSDEV_PREFIX_ATTR(ppe, throttle_begin, 0600);
+static SYSDEV_PREFIX_ATTR(ppe, throttle_full_stop, 0600);
+
 static struct attribute *ppe_attributes[] = {
 	&attr_ppe_temperature0.attr,
 	&attr_ppe_temperature1.attr,
+	&attr_ppe_throttle_end.attr,
+	&attr_ppe_throttle_begin.attr,
+	&attr_ppe_throttle_full_stop.attr,
 	NULL,
 };
 
diff --git a/arch/powerpc/platforms/cell/ras.c b/arch/powerpc/platforms/cell/ras.c
index 0984c7071695..b5ebc916388b 100644
--- a/arch/powerpc/platforms/cell/ras.c
+++ b/arch/powerpc/platforms/cell/ras.c
@@ -3,11 +3,13 @@
 #include <linux/types.h>
 #include <linux/kernel.h>
 #include <linux/smp.h>
+#include <linux/reboot.h>
 
 #include <asm/reg.h>
 #include <asm/io.h>
 #include <asm/prom.h>
 #include <asm/machdep.h>
+#include <asm/rtas.h>
 
 #include "ras.h"
 #include "cbe_regs.h"
@@ -82,6 +84,164 @@ static int cbe_machine_check_handler(struct pt_regs *regs)
 	return 0;
 }
 
+struct ptcal_area {
+	struct list_head list;
+	int nid;
+	int order;
+	struct page *pages;
+};
+
+static LIST_HEAD(ptcal_list);
+
+static int ptcal_start_tok, ptcal_stop_tok;
+
+static int __init cbe_ptcal_enable_on_node(int nid, int order)
+{
+	struct ptcal_area *area;
+	int ret = -ENOMEM;
+	unsigned long addr;
+
+#ifdef CONFIG_CRASH_DUMP
+	rtas_call(ptcal_stop_tok, 1, 1, NULL, nid);
+#endif
+
+	area = kmalloc(sizeof(*area), GFP_KERNEL);
+	if (!area)
+		goto out_err;
+
+	area->nid = nid;
+	area->order = order;
+	area->pages = alloc_pages_node(area->nid, GFP_KERNEL, area->order);
+
+	if (!area->pages)
+		goto out_free_area;
+
+	addr = __pa(page_address(area->pages));
+
+	ret = -EIO;
+	if (rtas_call(ptcal_start_tok, 3, 1, NULL, area->nid,
+				(unsigned int)(addr >> 32),
+				(unsigned int)(addr & 0xffffffff))) {
+		printk(KERN_ERR "%s: error enabling PTCAL on node %d!\n",
+				__FUNCTION__, nid);
+		goto out_free_pages;
+	}
+
+	list_add(&area->list, &ptcal_list);
+
+	return 0;
+
+out_free_pages:
+	__free_pages(area->pages, area->order);
+out_free_area:
+	kfree(area);
+out_err:
+	return ret;
+}
+
+static int __init cbe_ptcal_enable(void)
+{
+	const u32 *size;
+	struct device_node *np;
+	int order, found_mic = 0;
+
+	np = of_find_node_by_path("/rtas");
+	if (!np)
+		return -ENODEV;
+
+	size = get_property(np, "ibm,cbe-ptcal-size", NULL);
+	if (!size)
+		return -ENODEV;
+
+	pr_debug("%s: enabling PTCAL, size = 0x%x\n", __FUNCTION__, *size);
+	order = get_order(*size);
+	of_node_put(np);
+
+	/* support for malta device trees, with be@/mic@ nodes */
+	for_each_node_by_type(np, "mic-tm") {
+		cbe_ptcal_enable_on_node(of_node_to_nid(np), order);
+		found_mic = 1;
+	}
+
+	if (found_mic)
+		return 0;
+
+	/* support for older device tree - use cpu nodes */
+	for_each_node_by_type(np, "cpu") {
+		const u32 *nid = get_property(np, "node-id", NULL);
+		if (!nid) {
+			printk(KERN_ERR "%s: node %s is missing node-id?\n",
+					__FUNCTION__, np->full_name);
+			continue;
+		}
+		cbe_ptcal_enable_on_node(*nid, order);
+		found_mic = 1;
+	}
+
+	return found_mic ? 0 : -ENODEV;
+}
+
+static int cbe_ptcal_disable(void)
+{
+	struct ptcal_area *area, *tmp;
+	int ret = 0;
+
+	pr_debug("%s: disabling PTCAL\n", __FUNCTION__);
+
+	list_for_each_entry_safe(area, tmp, &ptcal_list, list) {
+		/* disable ptcal on this node */
+		if (rtas_call(ptcal_stop_tok, 1, 1, NULL, area->nid)) {
+			printk(KERN_ERR "%s: error disabling PTCAL "
+					"on node %d!\n", __FUNCTION__,
+					area->nid);
+			ret = -EIO;
+			continue;
+		}
+
+		/* ensure we can access the PTCAL area */
+		memset(page_address(area->pages), 0,
+				1 << (area->order + PAGE_SHIFT));
+
+		/* clean up */
+		list_del(&area->list);
+		__free_pages(area->pages, area->order);
+		kfree(area);
+	}
+
+	return ret;
+}
+
+static int cbe_ptcal_notify_reboot(struct notifier_block *nb,
+		unsigned long code, void *data)
+{
+	return cbe_ptcal_disable();
+}
+
+static struct notifier_block cbe_ptcal_reboot_notifier = {
+	.notifier_call = cbe_ptcal_notify_reboot
+};
+
+int __init cbe_ptcal_init(void)
+{
+	int ret;
+	ptcal_start_tok = rtas_token("ibm,cbe-start-ptcal");
+	ptcal_stop_tok = rtas_token("ibm,cbe-stop-ptcal");
+
+	if (ptcal_start_tok == RTAS_UNKNOWN_SERVICE
+			|| ptcal_stop_tok == RTAS_UNKNOWN_SERVICE)
+		return -ENODEV;
+
+	ret = register_reboot_notifier(&cbe_ptcal_reboot_notifier);
+	if (ret) {
+		printk(KERN_ERR "Can't disable PTCAL, so not enabling\n");
+		return ret;
+	}
+
+	return cbe_ptcal_enable();
+}
+
+arch_initcall(cbe_ptcal_init);
+
 void __init cbe_ras_init(void)
 {
 	unsigned long hid0;