cpufreq: powerpc: move cpufreq driver to drivers/cpufreq

Move cpufreq driver of powerpc platform to drivers/cpufreq.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

5 files changed, 1827 insertions, 0 deletions

diff --git a/drivers/cpufreq/Kconfig.powerpc b/drivers/cpufreq/Kconfig.powerpc
index 88f629e5a39..c48f83c93a9 100644
--- a/drivers/cpufreq/Kconfig.powerpc
+++ b/drivers/cpufreq/Kconfig.powerpc
@@ -33,3 +33,29 @@ config PPC_CORENET_CPUFREQ
 	  This adds the CPUFreq driver support for Freescale e500mc,
 	  e5500 and e6500 series SoCs which are capable of changing
 	  the CPU's frequency dynamically.
+
+config CPU_FREQ_PMAC
+	bool "Support for Apple PowerBooks"
+	depends on ADB_PMU && PPC32
+	select CPU_FREQ_TABLE
+	help
+	  This adds support for frequency switching on Apple PowerBooks,
+	  this currently includes some models of iBook & Titanium
+	  PowerBook.
+
+config CPU_FREQ_PMAC64
+	bool "Support for some Apple G5s"
+	depends on PPC_PMAC && PPC64
+	select CPU_FREQ_TABLE
+	help
+	  This adds support for frequency switching on Apple iMac G5,
+	  and some of the more recent desktop G5 machines as well.
+
+config PPC_PASEMI_CPUFREQ
+	bool "Support for PA Semi PWRficient"
+	depends on PPC_PASEMI
+	select CPU_FREQ_TABLE
+	default y
+	help
+	  This adds the support for frequency switching on PA Semi
+	  PWRficient processors.
diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
index c956094ccc3..2de2af2ca3b 100644
--- a/drivers/cpufreq/Makefile
+++ b/drivers/cpufreq/Makefile
@@ -80,6 +80,9 @@ ppc-cbe-cpufreq-y			+= ppc_cbe_cpufreq_pervasive.o ppc_cbe_cpufreq.o
 obj-$(CONFIG_CPU_FREQ_CBE_PMI)		+= ppc_cbe_cpufreq_pmi.o
 obj-$(CONFIG_CPU_FREQ_MAPLE)		+= maple-cpufreq.o
 obj-$(CONFIG_PPC_CORENET_CPUFREQ)   += ppc-corenet-cpufreq.o
+obj-$(CONFIG_CPU_FREQ_PMAC)		+= pmac32-cpufreq.o
+obj-$(CONFIG_CPU_FREQ_PMAC64)		+= pmac64-cpufreq.o
+obj-$(CONFIG_PPC_PASEMI_CPUFREQ)	+= pasemi-cpufreq.o
 
 ##################################################################################
 # Other platform drivers
diff --git a/drivers/cpufreq/pasemi-cpufreq.c b/drivers/cpufreq/pasemi-cpufreq.c
new file mode 100644
index 00000000000..b704da40406
--- /dev/null
+++ b/drivers/cpufreq/pasemi-cpufreq.c
@@ -0,0 +1,331 @@
+/*
+ * Copyright (C) 2007 PA Semi, Inc
+ *
+ * Authors: Egor Martovetsky <egor@pasemi.com>
+ *	    Olof Johansson <olof@lixom.net>
+ *
+ * Maintained by: Olof Johansson <olof@lixom.net>
+ *
+ * Based on arch/powerpc/platforms/cell/cbe_cpufreq.c:
+ * (C) Copyright IBM Deutschland Entwicklung GmbH 2005
+ *
+ * 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, 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., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+#include <linux/cpufreq.h>
+#include <linux/timer.h>
+#include <linux/module.h>
+
+#include <asm/hw_irq.h>
+#include <asm/io.h>
+#include <asm/prom.h>
+#include <asm/time.h>
+#include <asm/smp.h>
+
+#define SDCASR_REG		0x0100
+#define SDCASR_REG_STRIDE	0x1000
+#define SDCPWR_CFGA0_REG	0x0100
+#define SDCPWR_PWST0_REG	0x0000
+#define SDCPWR_GIZTIME_REG	0x0440
+
+/* SDCPWR_GIZTIME_REG fields */
+#define SDCPWR_GIZTIME_GR	0x80000000
+#define SDCPWR_GIZTIME_LONGLOCK	0x000000ff
+
+/* Offset of ASR registers from SDC base */
+#define SDCASR_OFFSET		0x120000
+
+static void __iomem *sdcpwr_mapbase;
+static void __iomem *sdcasr_mapbase;
+
+static DEFINE_MUTEX(pas_switch_mutex);
+
+/* Current astate, is used when waking up from power savings on
+ * one core, in case the other core has switched states during
+ * the idle time.
+ */
+static int current_astate;
+
+/* We support 5(A0-A4) power states excluding turbo(A5-A6) modes */
+static struct cpufreq_frequency_table pas_freqs[] = {
+	{0,	0},
+	{1,	0},
+	{2,	0},
+	{3,	0},
+	{4,	0},
+	{0,	CPUFREQ_TABLE_END},
+};
+
+static struct freq_attr *pas_cpu_freqs_attr[] = {
+	&cpufreq_freq_attr_scaling_available_freqs,
+	NULL,
+};
+
+/*
+ * hardware specific functions
+ */
+
+static int get_astate_freq(int astate)
+{
+	u32 ret;
+	ret = in_le32(sdcpwr_mapbase + SDCPWR_CFGA0_REG + (astate * 0x10));
+
+	return ret & 0x3f;
+}
+
+static int get_cur_astate(int cpu)
+{
+	u32 ret;
+
+	ret = in_le32(sdcpwr_mapbase + SDCPWR_PWST0_REG);
+	ret = (ret >> (cpu * 4)) & 0x7;
+
+	return ret;
+}
+
+static int get_gizmo_latency(void)
+{
+	u32 giztime, ret;
+
+	giztime = in_le32(sdcpwr_mapbase + SDCPWR_GIZTIME_REG);
+
+	/* just provide the upper bound */
+	if (giztime & SDCPWR_GIZTIME_GR)
+		ret = (giztime & SDCPWR_GIZTIME_LONGLOCK) * 128000;
+	else
+		ret = (giztime & SDCPWR_GIZTIME_LONGLOCK) * 1000;
+
+	return ret;
+}
+
+static void set_astate(int cpu, unsigned int astate)
+{
+	unsigned long flags;
+
+	/* Return if called before init has run */
+	if (unlikely(!sdcasr_mapbase))
+		return;
+
+	local_irq_save(flags);
+
+	out_le32(sdcasr_mapbase + SDCASR_REG + SDCASR_REG_STRIDE*cpu, astate);
+
+	local_irq_restore(flags);
+}
+
+int check_astate(void)
+{
+	return get_cur_astate(hard_smp_processor_id());
+}
+
+void restore_astate(int cpu)
+{
+	set_astate(cpu, current_astate);
+}
+
+/*
+ * cpufreq functions
+ */
+
+static int pas_cpufreq_cpu_init(struct cpufreq_policy *policy)
+{
+	const u32 *max_freqp;
+	u32 max_freq;
+	int i, cur_astate;
+	struct resource res;
+	struct device_node *cpu, *dn;
+	int err = -ENODEV;
+
+	cpu = of_get_cpu_node(policy->cpu, NULL);
+
+	if (!cpu)
+		goto out;
+
+	dn = of_find_compatible_node(NULL, NULL, "1682m-sdc");
+	if (!dn)
+		dn = of_find_compatible_node(NULL, NULL,
+					     "pasemi,pwrficient-sdc");
+	if (!dn)
+		goto out;
+	err = of_address_to_resource(dn, 0, &res);
+	of_node_put(dn);
+	if (err)
+		goto out;
+	sdcasr_mapbase = ioremap(res.start + SDCASR_OFFSET, 0x2000);
+	if (!sdcasr_mapbase) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	dn = of_find_compatible_node(NULL, NULL, "1682m-gizmo");
+	if (!dn)
+		dn = of_find_compatible_node(NULL, NULL,
+					     "pasemi,pwrficient-gizmo");
+	if (!dn) {
+		err = -ENODEV;
+		goto out_unmap_sdcasr;
+	}
+	err = of_address_to_resource(dn, 0, &res);
+	of_node_put(dn);
+	if (err)
+		goto out_unmap_sdcasr;
+	sdcpwr_mapbase = ioremap(res.start, 0x1000);
+	if (!sdcpwr_mapbase) {
+		err = -EINVAL;
+		goto out_unmap_sdcasr;
+	}
+
+	pr_debug("init cpufreq on CPU %d\n", policy->cpu);
+
+	max_freqp = of_get_property(cpu, "clock-frequency", NULL);
+	if (!max_freqp) {
+		err = -EINVAL;
+		goto out_unmap_sdcpwr;
+	}
+
+	/* we need the freq in kHz */
+	max_freq = *max_freqp / 1000;
+
+	pr_debug("max clock-frequency is at %u kHz\n", max_freq);
+	pr_debug("initializing frequency table\n");
+
+	/* initialize frequency table */
+	for (i=0; pas_freqs[i].frequency!=CPUFREQ_TABLE_END; i++) {
+		pas_freqs[i].frequency =
+			get_astate_freq(pas_freqs[i].driver_data) * 100000;
+		pr_debug("%d: %d\n", i, pas_freqs[i].frequency);
+	}
+
+	policy->cpuinfo.transition_latency = get_gizmo_latency();
+
+	cur_astate = get_cur_astate(policy->cpu);
+	pr_debug("current astate is at %d\n",cur_astate);
+
+	policy->cur = pas_freqs[cur_astate].frequency;
+	cpumask_copy(policy->cpus, cpu_online_mask);
+
+	ppc_proc_freq = policy->cur * 1000ul;
+
+	cpufreq_frequency_table_get_attr(pas_freqs, policy->cpu);
+
+	/* this ensures that policy->cpuinfo_min and policy->cpuinfo_max
+	 * are set correctly
+	 */
+	return cpufreq_frequency_table_cpuinfo(policy, pas_freqs);
+
+out_unmap_sdcpwr:
+	iounmap(sdcpwr_mapbase);
+
+out_unmap_sdcasr:
+	iounmap(sdcasr_mapbase);
+out:
+	return err;
+}
+
+static int pas_cpufreq_cpu_exit(struct cpufreq_policy *policy)
+{
+	/*
+	 * We don't support CPU hotplug. Don't unmap after the system
+	 * has already made it to a running state.
+	 */
+	if (system_state != SYSTEM_BOOTING)
+		return 0;
+
+	if (sdcasr_mapbase)
+		iounmap(sdcasr_mapbase);
+	if (sdcpwr_mapbase)
+		iounmap(sdcpwr_mapbase);
+
+	cpufreq_frequency_table_put_attr(policy->cpu);
+	return 0;
+}
+
+static int pas_cpufreq_verify(struct cpufreq_policy *policy)
+{
+	return cpufreq_frequency_table_verify(policy, pas_freqs);
+}
+
+static int pas_cpufreq_target(struct cpufreq_policy *policy,
+			      unsigned int target_freq,
+			      unsigned int relation)
+{
+	struct cpufreq_freqs freqs;
+	int pas_astate_new;
+	int i;
+
+	cpufreq_frequency_table_target(policy,
+				       pas_freqs,
+				       target_freq,
+				       relation,
+				       &pas_astate_new);
+
+	freqs.old = policy->cur;
+	freqs.new = pas_freqs[pas_astate_new].frequency;
+
+	mutex_lock(&pas_switch_mutex);
+	cpufreq_notify_transition(policy, &freqs, CPUFREQ_PRECHANGE);
+
+	pr_debug("setting frequency for cpu %d to %d kHz, 1/%d of max frequency\n",
+		 policy->cpu,
+		 pas_freqs[pas_astate_new].frequency,
+		 pas_freqs[pas_astate_new].driver_data);
+
+	current_astate = pas_astate_new;
+
+	for_each_online_cpu(i)
+		set_astate(i, pas_astate_new);
+
+	cpufreq_notify_transition(policy, &freqs, CPUFREQ_POSTCHANGE);
+	mutex_unlock(&pas_switch_mutex);
+
+	ppc_proc_freq = freqs.new * 1000ul;
+	return 0;
+}
+
+static struct cpufreq_driver pas_cpufreq_driver = {
+	.name		= "pas-cpufreq",
+	.owner		= THIS_MODULE,
+	.flags		= CPUFREQ_CONST_LOOPS,
+	.init		= pas_cpufreq_cpu_init,
+	.exit		= pas_cpufreq_cpu_exit,
+	.verify		= pas_cpufreq_verify,
+	.target		= pas_cpufreq_target,
+	.attr		= pas_cpu_freqs_attr,
+};
+
+/*
+ * module init and destoy
+ */
+
+static int __init pas_cpufreq_init(void)
+{
+	if (!of_machine_is_compatible("PA6T-1682M") &&
+	    !of_machine_is_compatible("pasemi,pwrficient"))
+		return -ENODEV;
+
+	return cpufreq_register_driver(&pas_cpufreq_driver);
+}
+
+static void __exit pas_cpufreq_exit(void)
+{
+	cpufreq_unregister_driver(&pas_cpufreq_driver);
+}
+
+module_init(pas_cpufreq_init);
+module_exit(pas_cpufreq_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Egor Martovetsky <egor@pasemi.com>, Olof Johansson <olof@lixom.net>");
diff --git a/drivers/cpufreq/pmac32-cpufreq.c b/drivers/cpufreq/pmac32-cpufreq.c
new file mode 100644
index 00000000000..3104fad8248
--- /dev/null
+++ b/drivers/cpufreq/pmac32-cpufreq.c
@@ -0,0 +1,721 @@
+/*
+ *  Copyright (C) 2002 - 2005 Benjamin Herrenschmidt <benh@kernel.crashing.org>
+ *  Copyright (C) 2004        John Steele Scott <toojays@toojays.net>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * TODO: Need a big cleanup here. Basically, we need to have different
+ * cpufreq_driver structures for the different type of HW instead of the
+ * current mess. We also need to better deal with the detection of the
+ * type of machine.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/sched.h>
+#include <linux/adb.h>
+#include <linux/pmu.h>
+#include <linux/cpufreq.h>
+#include <linux/init.h>
+#include <linux/device.h>
+#include <linux/hardirq.h>
+#include <asm/prom.h>
+#include <asm/machdep.h>
+#include <asm/irq.h>
+#include <asm/pmac_feature.h>
+#include <asm/mmu_context.h>
+#include <asm/sections.h>
+#include <asm/cputable.h>
+#include <asm/time.h>
+#include <asm/mpic.h>
+#include <asm/keylargo.h>
+#include <asm/switch_to.h>
+
+/* WARNING !!! This will cause calibrate_delay() to be called,
+ * but this is an __init function ! So you MUST go edit
+ * init/main.c to make it non-init before enabling DEBUG_FREQ
+ */
+#undef DEBUG_FREQ
+
+extern void low_choose_7447a_dfs(int dfs);
+extern void low_choose_750fx_pll(int pll);
+extern void low_sleep_handler(void);
+
+/*
+ * Currently, PowerMac cpufreq supports only high & low frequencies
+ * that are set by the firmware
+ */
+static unsigned int low_freq;
+static unsigned int hi_freq;
+static unsigned int cur_freq;
+static unsigned int sleep_freq;
+static unsigned long transition_latency;
+
+/*
+ * Different models uses different mechanisms to switch the frequency
+ */
+static int (*set_speed_proc)(int low_speed);
+static unsigned int (*get_speed_proc)(void);
+
+/*
+ * Some definitions used by the various speedprocs
+ */
+static u32 voltage_gpio;
+static u32 frequency_gpio;
+static u32 slew_done_gpio;
+static int no_schedule;
+static int has_cpu_l2lve;
+static int is_pmu_based;
+
+/* There are only two frequency states for each processor. Values
+ * are in kHz for the time being.
+ */
+#define CPUFREQ_HIGH                  0
+#define CPUFREQ_LOW                   1
+
+static struct cpufreq_frequency_table pmac_cpu_freqs[] = {
+	{CPUFREQ_HIGH, 		0},
+	{CPUFREQ_LOW,		0},
+	{0,			CPUFREQ_TABLE_END},
+};
+
+static struct freq_attr* pmac_cpu_freqs_attr[] = {
+	&cpufreq_freq_attr_scaling_available_freqs,
+	NULL,
+};
+
+static inline void local_delay(unsigned long ms)
+{
+	if (no_schedule)
+		mdelay(ms);
+	else
+		msleep(ms);
+}
+
+#ifdef DEBUG_FREQ
+static inline void debug_calc_bogomips(void)
+{
+	/* This will cause a recalc of bogomips and display the
+	 * result. We backup/restore the value to avoid affecting the
+	 * core cpufreq framework's own calculation.
+	 */
+	unsigned long save_lpj = loops_per_jiffy;
+	calibrate_delay();
+	loops_per_jiffy = save_lpj;
+}
+#endif /* DEBUG_FREQ */
+
+/* Switch CPU speed under 750FX CPU control
+ */
+static int cpu_750fx_cpu_speed(int low_speed)
+{
+	u32 hid2;
+
+	if (low_speed == 0) {
+		/* ramping up, set voltage first */
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, voltage_gpio, 0x05);
+		/* Make sure we sleep for at least 1ms */
+		local_delay(10);
+
+		/* tweak L2 for high voltage */
+		if (has_cpu_l2lve) {
+			hid2 = mfspr(SPRN_HID2);
+			hid2 &= ~0x2000;
+			mtspr(SPRN_HID2, hid2);
+		}
+	}
+#ifdef CONFIG_6xx
+	low_choose_750fx_pll(low_speed);
+#endif
+	if (low_speed == 1) {
+		/* tweak L2 for low voltage */
+		if (has_cpu_l2lve) {
+			hid2 = mfspr(SPRN_HID2);
+			hid2 |= 0x2000;
+			mtspr(SPRN_HID2, hid2);
+		}
+
+		/* ramping down, set voltage last */
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, voltage_gpio, 0x04);
+		local_delay(10);
+	}
+
+	return 0;
+}
+
+static unsigned int cpu_750fx_get_cpu_speed(void)
+{
+	if (mfspr(SPRN_HID1) & HID1_PS)
+		return low_freq;
+	else
+		return hi_freq;
+}
+
+/* Switch CPU speed using DFS */
+static int dfs_set_cpu_speed(int low_speed)
+{
+	if (low_speed == 0) {
+		/* ramping up, set voltage first */
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, voltage_gpio, 0x05);
+		/* Make sure we sleep for at least 1ms */
+		local_delay(1);
+	}
+
+	/* set frequency */
+#ifdef CONFIG_6xx
+	low_choose_7447a_dfs(low_speed);
+#endif
+	udelay(100);
+
+	if (low_speed == 1) {
+		/* ramping down, set voltage last */
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, voltage_gpio, 0x04);
+		local_delay(1);
+	}
+
+	return 0;
+}
+
+static unsigned int dfs_get_cpu_speed(void)
+{
+	if (mfspr(SPRN_HID1) & HID1_DFS)
+		return low_freq;
+	else
+		return hi_freq;
+}
+
+
+/* Switch CPU speed using slewing GPIOs
+ */
+static int gpios_set_cpu_speed(int low_speed)
+{
+	int gpio, timeout = 0;
+
+	/* If ramping up, set voltage first */
+	if (low_speed == 0) {
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, voltage_gpio, 0x05);
+		/* Delay is way too big but it's ok, we schedule */
+		local_delay(10);
+	}
+
+	/* Set frequency */
+	gpio = 	pmac_call_feature(PMAC_FTR_READ_GPIO, NULL, frequency_gpio, 0);
+	if (low_speed == ((gpio & 0x01) == 0))
+		goto skip;
+
+	pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, frequency_gpio,
+			  low_speed ? 0x04 : 0x05);
+	udelay(200);
+	do {
+		if (++timeout > 100)
+			break;
+		local_delay(1);
+		gpio = pmac_call_feature(PMAC_FTR_READ_GPIO, NULL, slew_done_gpio, 0);
+	} while((gpio & 0x02) == 0);
+ skip:
+	/* If ramping down, set voltage last */
+	if (low_speed == 1) {
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, voltage_gpio, 0x04);
+		/* Delay is way too big but it's ok, we schedule */
+		local_delay(10);
+	}
+
+#ifdef DEBUG_FREQ
+	debug_calc_bogomips();
+#endif
+
+	return 0;
+}
+
+/* Switch CPU speed under PMU control
+ */
+static int pmu_set_cpu_speed(int low_speed)
+{
+	struct adb_request req;
+	unsigned long save_l2cr;
+	unsigned long save_l3cr;
+	unsigned int pic_prio;
+	unsigned long flags;
+
+	preempt_disable();
+
+#ifdef DEBUG_FREQ
+	printk(KERN_DEBUG "HID1, before: %x\n", mfspr(SPRN_HID1));
+#endif
+	pmu_suspend();
+
+	/* Disable all interrupt sources on openpic */
+ 	pic_prio = mpic_cpu_get_priority();
+	mpic_cpu_set_priority(0xf);
+
+	/* Make sure the decrementer won't interrupt us */
+	asm volatile("mtdec %0" : : "r" (0x7fffffff));
+	/* Make sure any pending DEC interrupt occurring while we did
+	 * the above didn't re-enable the DEC */
+	mb();
+	asm volatile("mtdec %0" : : "r" (0x7fffffff));
+
+	/* We can now disable MSR_EE */
+	local_irq_save(flags);
+
+	/* Giveup the FPU & vec */
+	enable_kernel_fp();
+
+#ifdef CONFIG_ALTIVEC
+	if (cpu_has_feature(CPU_FTR_ALTIVEC))
+		enable_kernel_altivec();
+#endif /* CONFIG_ALTIVEC */
+
+	/* Save & disable L2 and L3 caches */
+	save_l3cr = _get_L3CR();	/* (returns -1 if not available) */
+	save_l2cr = _get_L2CR();	/* (returns -1 if not available) */
+
+	/* Send the new speed command. My assumption is that this command
+	 * will cause PLL_CFG[0..3] to be changed next time CPU goes to sleep
+	 */
+	pmu_request(&req, NULL, 6, PMU_CPU_SPEED, 'W', 'O', 'O', 'F', low_speed);
+	while (!req.complete)
+		pmu_poll();
+
+	/* Prepare the northbridge for the speed transition */
+	pmac_call_feature(PMAC_FTR_SLEEP_STATE,NULL,1,1);
+
+	/* Call low level code to backup CPU state and recover from
+	 * hardware reset
+	 */
+	low_sleep_handler();
+
+	/* Restore the northbridge */
+	pmac_call_feature(PMAC_FTR_SLEEP_STATE,NULL,1,0);
+
+	/* Restore L2 cache */
+	if (save_l2cr != 0xffffffff && (save_l2cr & L2CR_L2E) != 0)
+ 		_set_L2CR(save_l2cr);
+	/* Restore L3 cache */
+	if (save_l3cr != 0xffffffff && (save_l3cr & L3CR_L3E) != 0)
+ 		_set_L3CR(save_l3cr);
+
+	/* Restore userland MMU context */
+	switch_mmu_context(NULL, current->active_mm);
+
+#ifdef DEBUG_FREQ
+	printk(KERN_DEBUG "HID1, after: %x\n", mfspr(SPRN_HID1));
+#endif
+
+	/* Restore low level PMU operations */
+	pmu_unlock();
+
+	/*
+	 * Restore decrementer; we'll take a decrementer interrupt
+	 * as soon as interrupts are re-enabled and the generic
+	 * clockevents code will reprogram it with the right value.
+	 */
+	set_dec(1);
+
+	/* Restore interrupts */
+ 	mpic_cpu_set_priority(pic_prio);
+
+	/* Let interrupts flow again ... */
+	local_irq_restore(flags);
+
+#ifdef DEBUG_FREQ
+	debug_calc_bogomips();
+#endif
+
+	pmu_resume();
+
+	preempt_enable();
+
+	return 0;
+}
+
+static int do_set_cpu_speed(struct cpufreq_policy *policy, int speed_mode,
+		int notify)
+{
+	struct cpufreq_freqs freqs;
+	unsigned long l3cr;
+	static unsigned long prev_l3cr;
+
+	freqs.old = cur_freq;
+	freqs.new = (speed_mode == CPUFREQ_HIGH) ? hi_freq : low_freq;
+
+	if (freqs.old == freqs.new)
+		return 0;
+
+	if (notify)
+		cpufreq_notify_transition(policy, &freqs, CPUFREQ_PRECHANGE);
+	if (speed_mode == CPUFREQ_LOW &&
+	    cpu_has_feature(CPU_FTR_L3CR)) {
+		l3cr = _get_L3CR();
+		if (l3cr & L3CR_L3E) {
+			prev_l3cr = l3cr;
+			_set_L3CR(0);
+		}
+	}
+	set_speed_proc(speed_mode == CPUFREQ_LOW);
+	if (speed_mode == CPUFREQ_HIGH &&
+	    cpu_has_feature(CPU_FTR_L3CR)) {
+		l3cr = _get_L3CR();
+		if ((prev_l3cr & L3CR_L3E) && l3cr != prev_l3cr)
+			_set_L3CR(prev_l3cr);
+	}
+	if (notify)
+		cpufreq_notify_transition(policy, &freqs, CPUFREQ_POSTCHANGE);
+	cur_freq = (speed_mode == CPUFREQ_HIGH) ? hi_freq : low_freq;
+
+	return 0;
+}
+
+static unsigned int pmac_cpufreq_get_speed(unsigned int cpu)
+{
+	return cur_freq;
+}
+
+static int pmac_cpufreq_verify(struct cpufreq_policy *policy)
+{
+	return cpufreq_frequency_table_verify(policy, pmac_cpu_freqs);
+}
+
+static int pmac_cpufreq_target(	struct cpufreq_policy *policy,
+					unsigned int target_freq,
+					unsigned int relation)
+{
+	unsigned int    newstate = 0;
+	int		rc;
+
+	if (cpufreq_frequency_table_target(policy, pmac_cpu_freqs,
+			target_freq, relation, &newstate))
+		return -EINVAL;
+
+	rc = do_set_cpu_speed(policy, newstate, 1);
+
+	ppc_proc_freq = cur_freq * 1000ul;
+	return rc;
+}
+
+static int pmac_cpufreq_cpu_init(struct cpufreq_policy *policy)
+{
+	if (policy->cpu != 0)
+		return -ENODEV;
+
+	policy->cpuinfo.transition_latency	= transition_latency;
+	policy->cur = cur_freq;
+
+	cpufreq_frequency_table_get_attr(pmac_cpu_freqs, policy->cpu);
+	return cpufreq_frequency_table_cpuinfo(policy, pmac_cpu_freqs);
+}
+
+static u32 read_gpio(struct device_node *np)
+{
+	const u32 *reg = of_get_property(np, "reg", NULL);
+	u32 offset;
+
+	if (reg == NULL)
+		return 0;
+	/* That works for all keylargos but shall be fixed properly
+	 * some day... The problem is that it seems we can't rely
+	 * on the "reg" property of the GPIO nodes, they are either
+	 * relative to the base of KeyLargo or to the base of the
+	 * GPIO space, and the device-tree doesn't help.
+	 */
+	offset = *reg;
+	if (offset < KEYLARGO_GPIO_LEVELS0)
+		offset += KEYLARGO_GPIO_LEVELS0;
+	return offset;
+}
+
+static int pmac_cpufreq_suspend(struct cpufreq_policy *policy)
+{
+	/* Ok, this could be made a bit smarter, but let's be robust for now. We
+	 * always force a speed change to high speed before sleep, to make sure
+	 * we have appropriate voltage and/or bus speed for the wakeup process,
+	 * and to make sure our loops_per_jiffies are "good enough", that is will
+	 * not cause too short delays if we sleep in low speed and wake in high
+	 * speed..
+	 */
+	no_schedule = 1;
+	sleep_freq = cur_freq;
+	if (cur_freq == low_freq && !is_pmu_based)
+		do_set_cpu_speed(policy, CPUFREQ_HIGH, 0);
+	return 0;
+}
+
+static int pmac_cpufreq_resume(struct cpufreq_policy *policy)
+{
+	/* If we resume, first check if we have a get() function */
+	if (get_speed_proc)
+		cur_freq = get_speed_proc();
+	else
+		cur_freq = 0;
+
+	/* We don't, hrm... we don't really know our speed here, best
+	 * is that we force a switch to whatever it was, which is
+	 * probably high speed due to our suspend() routine
+	 */
+	do_set_cpu_speed(policy, sleep_freq == low_freq ?
+			 CPUFREQ_LOW : CPUFREQ_HIGH, 0);
+
+	ppc_proc_freq = cur_freq * 1000ul;
+
+	no_schedule = 0;
+	return 0;
+}
+
+static struct cpufreq_driver pmac_cpufreq_driver = {
+	.verify 	= pmac_cpufreq_verify,
+	.target 	= pmac_cpufreq_target,
+	.get		= pmac_cpufreq_get_speed,
+	.init		= pmac_cpufreq_cpu_init,
+	.suspend	= pmac_cpufreq_suspend,
+	.resume		= pmac_cpufreq_resume,
+	.flags		= CPUFREQ_PM_NO_WARN,
+	.attr		= pmac_cpu_freqs_attr,
+	.name		= "powermac",
+	.owner		= THIS_MODULE,
+};
+
+
+static int pmac_cpufreq_init_MacRISC3(struct device_node *cpunode)
+{
+	struct device_node *volt_gpio_np = of_find_node_by_name(NULL,
+								"voltage-gpio");
+	struct device_node *freq_gpio_np = of_find_node_by_name(NULL,
+								"frequency-gpio");
+	struct device_node *slew_done_gpio_np = of_find_node_by_name(NULL,
+								     "slewing-done");
+	const u32 *value;
+
+	/*
+	 * Check to see if it's GPIO driven or PMU only
+	 *
+	 * The way we extract the GPIO address is slightly hackish, but it
+	 * works well enough for now. We need to abstract the whole GPIO
+	 * stuff sooner or later anyway
+	 */
+
+	if (volt_gpio_np)
+		voltage_gpio = read_gpio(volt_gpio_np);
+	if (freq_gpio_np)
+		frequency_gpio = read_gpio(freq_gpio_np);
+	if (slew_done_gpio_np)
+		slew_done_gpio = read_gpio(slew_done_gpio_np);
+
+	/* If we use the frequency GPIOs, calculate the min/max speeds based
+	 * on the bus frequencies
+	 */
+	if (frequency_gpio && slew_done_gpio) {
+		int lenp, rc;
+		const u32 *freqs, *ratio;
+
+		freqs = of_get_property(cpunode, "bus-frequencies", &lenp);
+		lenp /= sizeof(u32);
+		if (freqs == NULL || lenp != 2) {
+			printk(KERN_ERR "cpufreq: bus-frequencies incorrect or missing\n");
+			return 1;
+		}
+		ratio = of_get_property(cpunode, "processor-to-bus-ratio*2",
+						NULL);
+		if (ratio == NULL) {
+			printk(KERN_ERR "cpufreq: processor-to-bus-ratio*2 missing\n");
+			return 1;
+		}
+
+		/* Get the min/max bus frequencies */
+		low_freq = min(freqs[0], freqs[1]);
+		hi_freq = max(freqs[0], freqs[1]);
+
+		/* Grrrr.. It _seems_ that the device-tree is lying on the low bus
+		 * frequency, it claims it to be around 84Mhz on some models while
+		 * it appears to be approx. 101Mhz on all. Let's hack around here...
+		 * fortunately, we don't need to be too precise
+		 */
+		if (low_freq < 98000000)
+			low_freq = 101000000;
+
+		/* Convert those to CPU core clocks */
+		low_freq = (low_freq * (*ratio)) / 2000;
+		hi_freq = (hi_freq * (*ratio)) / 2000;
+
+		/* Now we get the frequencies, we read the GPIO to see what is out current
+		 * speed
+		 */
+		rc = pmac_call_feature(PMAC_FTR_READ_GPIO, NULL, frequency_gpio, 0);
+		cur_freq = (rc & 0x01) ? hi_freq : low_freq;
+
+		set_speed_proc = gpios_set_cpu_speed;
+		return 1;
+	}
+
+	/* If we use the PMU, look for the min & max frequencies in the
+	 * device-tree
+	 */
+	value = of_get_property(cpunode, "min-clock-frequency", NULL);
+	if (!value)
+		return 1;
+	low_freq = (*value) / 1000;
+	/* The PowerBook G4 12" (PowerBook6,1) has an error in the device-tree
+	 * here */
+	if (low_freq < 100000)
+		low_freq *= 10;
+
+	value = of_get_property(cpunode, "max-clock-frequency", NULL);
+	if (!value)
+		return 1;
+	hi_freq = (*value) / 1000;
+	set_speed_proc = pmu_set_cpu_speed;
+	is_pmu_based = 1;
+
+	return 0;
+}
+
+static int pmac_cpufreq_init_7447A(struct device_node *cpunode)
+{
+	struct device_node *volt_gpio_np;
+
+	if (of_get_property(cpunode, "dynamic-power-step", NULL) == NULL)
+		return 1;
+
+	volt_gpio_np = of_find_node_by_name(NULL, "cpu-vcore-select");
+	if (volt_gpio_np)
+		voltage_gpio = read_gpio(volt_gpio_np);
+	if (!voltage_gpio){
+		printk(KERN_ERR "cpufreq: missing cpu-vcore-select gpio\n");
+		return 1;
+	}
+
+	/* OF only reports the high frequency */
+	hi_freq = cur_freq;
+	low_freq = cur_freq/2;
+
+	/* Read actual frequency from CPU */
+	cur_freq = dfs_get_cpu_speed();
+	set_speed_proc = dfs_set_cpu_speed;
+	get_speed_proc = dfs_get_cpu_speed;
+
+	return 0;
+}
+
+static int pmac_cpufreq_init_750FX(struct device_node *cpunode)
+{
+	struct device_node *volt_gpio_np;
+	u32 pvr;
+	const u32 *value;
+
+	if (of_get_property(cpunode, "dynamic-power-step", NULL) == NULL)
+		return 1;
+
+	hi_freq = cur_freq;
+	value = of_get_property(cpunode, "reduced-clock-frequency", NULL);
+	if (!value)
+		return 1;
+	low_freq = (*value) / 1000;
+
+	volt_gpio_np = of_find_node_by_name(NULL, "cpu-vcore-select");
+	if (volt_gpio_np)
+		voltage_gpio = read_gpio(volt_gpio_np);
+
+	pvr = mfspr(SPRN_PVR);
+	has_cpu_l2lve = !((pvr & 0xf00) == 0x100);
+
+	set_speed_proc = cpu_750fx_cpu_speed;
+	get_speed_proc = cpu_750fx_get_cpu_speed;
+	cur_freq = cpu_750fx_get_cpu_speed();
+
+	return 0;
+}
+
+/* Currently, we support the following machines:
+ *
+ *  - Titanium PowerBook 1Ghz (PMU based, 667Mhz & 1Ghz)
+ *  - Titanium PowerBook 800 (PMU based, 667Mhz & 800Mhz)
+ *  - Titanium PowerBook 400 (PMU based, 300Mhz & 400Mhz)
+ *  - Titanium PowerBook 500 (PMU based, 300Mhz & 500Mhz)
+ *  - iBook2 500/600 (PMU based, 400Mhz & 500/600Mhz)
+ *  - iBook2 700 (CPU based, 400Mhz & 700Mhz, support low voltage)
+ *  - Recent MacRISC3 laptops
+ *  - All new machines with 7447A CPUs
+ */
+static int __init pmac_cpufreq_setup(void)
+{
+	struct device_node	*cpunode;
+	const u32		*value;
+
+	if (strstr(cmd_line, "nocpufreq"))
+		return 0;
+
+	/* Assume only one CPU */
+	cpunode = of_find_node_by_type(NULL, "cpu");
+	if (!cpunode)
+		goto out;
+
+	/* Get current cpu clock freq */
+	value = of_get_property(cpunode, "clock-frequency", NULL);
+	if (!value)
+		goto out;
+	cur_freq = (*value) / 1000;
+	transition_latency = CPUFREQ_ETERNAL;
+
+	/*  Check for 7447A based MacRISC3 */
+	if (of_machine_is_compatible("MacRISC3") &&
+	    of_get_property(cpunode, "dynamic-power-step", NULL) &&
+	    PVR_VER(mfspr(SPRN_PVR)) == 0x8003) {
+		pmac_cpufreq_init_7447A(cpunode);
+		transition_latency = 8000000;
+	/* Check for other MacRISC3 machines */
+	} else if (of_machine_is_compatible("PowerBook3,4") ||
+		   of_machine_is_compatible("PowerBook3,5") ||
+		   of_machine_is_compatible("MacRISC3")) {
+		pmac_cpufreq_init_MacRISC3(cpunode);
+	/* Else check for iBook2 500/600 */
+	} else if (of_machine_is_compatible("PowerBook4,1")) {
+		hi_freq = cur_freq;
+		low_freq = 400000;
+		set_speed_proc = pmu_set_cpu_speed;
+		is_pmu_based = 1;
+	}
+	/* Else check for TiPb 550 */
+	else if (of_machine_is_compatible("PowerBook3,3") && cur_freq == 550000) {
+		hi_freq = cur_freq;
+		low_freq = 500000;
+		set_speed_proc = pmu_set_cpu_speed;
+		is_pmu_based = 1;
+	}
+	/* Else check for TiPb 400 & 500 */
+	else if (of_machine_is_compatible("PowerBook3,2")) {
+		/* We only know about the 400 MHz and the 500Mhz model
+		 * they both have 300 MHz as low frequency
+		 */
+		if (cur_freq < 350000 || cur_freq > 550000)
+			goto out;
+		hi_freq = cur_freq;
+		low_freq = 300000;
+		set_speed_proc = pmu_set_cpu_speed;
+		is_pmu_based = 1;
+	}
+	/* Else check for 750FX */
+	else if (PVR_VER(mfspr(SPRN_PVR)) == 0x7000)
+		pmac_cpufreq_init_750FX(cpunode);
+out:
+	of_node_put(cpunode);
+	if (set_speed_proc == NULL)
+		return -ENODEV;
+
+	pmac_cpu_freqs[CPUFREQ_LOW].frequency = low_freq;
+	pmac_cpu_freqs[CPUFREQ_HIGH].frequency = hi_freq;
+	ppc_proc_freq = cur_freq * 1000ul;
+
+	printk(KERN_INFO "Registering PowerMac CPU frequency driver\n");
+	printk(KERN_INFO "Low: %d Mhz, High: %d Mhz, Boot: %d Mhz\n",
+	       low_freq/1000, hi_freq/1000, cur_freq/1000);
+
+	return cpufreq_register_driver(&pmac_cpufreq_driver);
+}
+
+module_init(pmac_cpufreq_setup);
+
diff --git a/drivers/cpufreq/pmac64-cpufreq.c b/drivers/cpufreq/pmac64-cpufreq.c
new file mode 100644
index 00000000000..7ba423431cf
--- /dev/null
+++ b/drivers/cpufreq/pmac64-cpufreq.c
@@ -0,0 +1,746 @@
+/*
+ *  Copyright (C) 2002 - 2005 Benjamin Herrenschmidt <benh@kernel.crashing.org>
+ *  and                       Markus Demleitner <msdemlei@cl.uni-heidelberg.de>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This driver adds basic cpufreq support for SMU & 970FX based G5 Macs,
+ * that is iMac G5 and latest single CPU desktop.
+ */
+
+#undef DEBUG
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/sched.h>
+#include <linux/cpufreq.h>
+#include <linux/init.h>
+#include <linux/completion.h>
+#include <linux/mutex.h>
+#include <asm/prom.h>
+#include <asm/machdep.h>
+#include <asm/irq.h>
+#include <asm/sections.h>
+#include <asm/cputable.h>
+#include <asm/time.h>
+#include <asm/smu.h>
+#include <asm/pmac_pfunc.h>
+
+#define DBG(fmt...) pr_debug(fmt)
+
+/* see 970FX user manual */
+
+#define SCOM_PCR 0x0aa001			/* PCR scom addr */
+
+#define PCR_HILO_SELECT		0x80000000U	/* 1 = PCR, 0 = PCRH */
+#define PCR_SPEED_FULL		0x00000000U	/* 1:1 speed value */
+#define PCR_SPEED_HALF		0x00020000U	/* 1:2 speed value */
+#define PCR_SPEED_QUARTER	0x00040000U	/* 1:4 speed value */
+#define PCR_SPEED_MASK		0x000e0000U	/* speed mask */
+#define PCR_SPEED_SHIFT		17
+#define PCR_FREQ_REQ_VALID	0x00010000U	/* freq request valid */
+#define PCR_VOLT_REQ_VALID	0x00008000U	/* volt request valid */
+#define PCR_TARGET_TIME_MASK	0x00006000U	/* target time */
+#define PCR_STATLAT_MASK	0x00001f00U	/* STATLAT value */
+#define PCR_SNOOPLAT_MASK	0x000000f0U	/* SNOOPLAT value */
+#define PCR_SNOOPACC_MASK	0x0000000fU	/* SNOOPACC value */
+
+#define SCOM_PSR 0x408001			/* PSR scom addr */
+/* warning: PSR is a 64 bits register */
+#define PSR_CMD_RECEIVED	0x2000000000000000U   /* command received */
+#define PSR_CMD_COMPLETED	0x1000000000000000U   /* command completed */
+#define PSR_CUR_SPEED_MASK	0x0300000000000000U   /* current speed */
+#define PSR_CUR_SPEED_SHIFT	(56)
+
+/*
+ * The G5 only supports two frequencies (Quarter speed is not supported)
+ */
+#define CPUFREQ_HIGH                  0
+#define CPUFREQ_LOW                   1
+
+static struct cpufreq_frequency_table g5_cpu_freqs[] = {
+	{CPUFREQ_HIGH, 		0},
+	{CPUFREQ_LOW,		0},
+	{0,			CPUFREQ_TABLE_END},
+};
+
+static struct freq_attr* g5_cpu_freqs_attr[] = {
+	&cpufreq_freq_attr_scaling_available_freqs,
+	NULL,
+};
+
+/* Power mode data is an array of the 32 bits PCR values to use for
+ * the various frequencies, retrieved from the device-tree
+ */
+static int g5_pmode_cur;
+
+static void (*g5_switch_volt)(int speed_mode);
+static int (*g5_switch_freq)(int speed_mode);
+static int (*g5_query_freq)(void);
+
+static DEFINE_MUTEX(g5_switch_mutex);
+
+static unsigned long transition_latency;
+
+#ifdef CONFIG_PMAC_SMU
+
+static const u32 *g5_pmode_data;
+static int g5_pmode_max;
+
+static struct smu_sdbp_fvt *g5_fvt_table;	/* table of op. points */
+static int g5_fvt_count;			/* number of op. points */
+static int g5_fvt_cur;				/* current op. point */
+
+/*
+ * SMU based voltage switching for Neo2 platforms
+ */
+
+static void g5_smu_switch_volt(int speed_mode)
+{
+	struct smu_simple_cmd	cmd;
+
+	DECLARE_COMPLETION_ONSTACK(comp);
+	smu_queue_simple(&cmd, SMU_CMD_POWER_COMMAND, 8, smu_done_complete,
+			 &comp, 'V', 'S', 'L', 'E', 'W',
+			 0xff, g5_fvt_cur+1, speed_mode);
+	wait_for_completion(&comp);
+}
+
+/*
+ * Platform function based voltage/vdnap switching for Neo2
+ */
+
+static struct pmf_function *pfunc_set_vdnap0;
+static struct pmf_function *pfunc_vdnap0_complete;
+
+static void g5_vdnap_switch_volt(int speed_mode)
+{
+	struct pmf_args args;
+	u32 slew, done = 0;
+	unsigned long timeout;
+
+	slew = (speed_mode == CPUFREQ_LOW) ? 1 : 0;
+	args.count = 1;
+	args.u[0].p = &slew;
+
+	pmf_call_one(pfunc_set_vdnap0, &args);
+
+	/* It's an irq GPIO so we should be able to just block here,
+	 * I'll do that later after I've properly tested the IRQ code for
+	 * platform functions
+	 */
+	timeout = jiffies + HZ/10;
+	while(!time_after(jiffies, timeout)) {
+		args.count = 1;
+		args.u[0].p = &done;
+		pmf_call_one(pfunc_vdnap0_complete, &args);
+		if (done)
+			break;
+		msleep(1);
+	}
+	if (done == 0)
+		printk(KERN_WARNING "cpufreq: Timeout in clock slewing !\n");
+}
+
+
+/*
+ * SCOM based frequency switching for 970FX rev3
+ */
+static int g5_scom_switch_freq(int speed_mode)
+{
+	unsigned long flags;
+	int to;
+
+	/* If frequency is going up, first ramp up the voltage */
+	if (speed_mode < g5_pmode_cur)
+		g5_switch_volt(speed_mode);
+
+	local_irq_save(flags);
+
+	/* Clear PCR high */
+	scom970_write(SCOM_PCR, 0);
+	/* Clear PCR low */
+       	scom970_write(SCOM_PCR, PCR_HILO_SELECT | 0);
+	/* Set PCR low */
+	scom970_write(SCOM_PCR, PCR_HILO_SELECT |
+		      g5_pmode_data[speed_mode]);
+
+	/* Wait for completion */
+	for (to = 0; to < 10; to++) {
+		unsigned long psr = scom970_read(SCOM_PSR);
+
+		if ((psr & PSR_CMD_RECEIVED) == 0 &&
+		    (((psr >> PSR_CUR_SPEED_SHIFT) ^
+		      (g5_pmode_data[speed_mode] >> PCR_SPEED_SHIFT)) & 0x3)
+		    == 0)
+			break;
+		if (psr & PSR_CMD_COMPLETED)
+			break;
+		udelay(100);
+	}
+
+	local_irq_restore(flags);
+
+	/* If frequency is going down, last ramp the voltage */
+	if (speed_mode > g5_pmode_cur)
+		g5_switch_volt(speed_mode);
+
+	g5_pmode_cur = speed_mode;
+	ppc_proc_freq = g5_cpu_freqs[speed_mode].frequency * 1000ul;
+
+	return 0;
+}
+
+static int g5_scom_query_freq(void)
+{
+	unsigned long psr = scom970_read(SCOM_PSR);
+	int i;
+
+	for (i = 0; i <= g5_pmode_max; i++)
+		if ((((psr >> PSR_CUR_SPEED_SHIFT) ^
+		      (g5_pmode_data[i] >> PCR_SPEED_SHIFT)) & 0x3) == 0)
+			break;
+	return i;
+}
+
+/*
+ * Fake voltage switching for platforms with missing support
+ */
+
+static void g5_dummy_switch_volt(int speed_mode)
+{
+}
+
+#endif /* CONFIG_PMAC_SMU */
+
+/*
+ * Platform function based voltage switching for PowerMac7,2 & 7,3
+ */
+
+static struct pmf_function *pfunc_cpu0_volt_high;
+static struct pmf_function *pfunc_cpu0_volt_low;
+static struct pmf_function *pfunc_cpu1_volt_high;
+static struct pmf_function *pfunc_cpu1_volt_low;
+
+static void g5_pfunc_switch_volt(int speed_mode)
+{
+	if (speed_mode == CPUFREQ_HIGH) {
+		if (pfunc_cpu0_volt_high)
+			pmf_call_one(pfunc_cpu0_volt_high, NULL);
+		if (pfunc_cpu1_volt_high)
+			pmf_call_one(pfunc_cpu1_volt_high, NULL);
+	} else {
+		if (pfunc_cpu0_volt_low)
+			pmf_call_one(pfunc_cpu0_volt_low, NULL);
+		if (pfunc_cpu1_volt_low)
+			pmf_call_one(pfunc_cpu1_volt_low, NULL);
+	}
+	msleep(10); /* should be faster , to fix */
+}
+
+/*
+ * Platform function based frequency switching for PowerMac7,2 & 7,3
+ */
+
+static struct pmf_function *pfunc_cpu_setfreq_high;
+static struct pmf_function *pfunc_cpu_setfreq_low;
+static struct pmf_function *pfunc_cpu_getfreq;
+static struct pmf_function *pfunc_slewing_done;
+
+static int g5_pfunc_switch_freq(int speed_mode)
+{
+	struct pmf_args args;
+	u32 done = 0;
+	unsigned long timeout;
+	int rc;
+
+	DBG("g5_pfunc_switch_freq(%d)\n", speed_mode);
+
+	/* If frequency is going up, first ramp up the voltage */
+	if (speed_mode < g5_pmode_cur)
+		g5_switch_volt(speed_mode);
+
+	/* Do it */
+	if (speed_mode == CPUFREQ_HIGH)
+		rc = pmf_call_one(pfunc_cpu_setfreq_high, NULL);
+	else
+		rc = pmf_call_one(pfunc_cpu_setfreq_low, NULL);
+
+	if (rc)
+		printk(KERN_WARNING "cpufreq: pfunc switch error %d\n", rc);
+
+	/* It's an irq GPIO so we should be able to just block here,
+	 * I'll do that later after I've properly tested the IRQ code for
+	 * platform functions
+	 */
+	timeout = jiffies + HZ/10;
+	while(!time_after(jiffies, timeout)) {
+		args.count = 1;
+		args.u[0].p = &done;
+		pmf_call_one(pfunc_slewing_done, &args);
+		if (done)
+			break;
+		msleep(1);
+	}
+	if (done == 0)
+		printk(KERN_WARNING "cpufreq: Timeout in clock slewing !\n");
+
+	/* If frequency is going down, last ramp the voltage */
+	if (speed_mode > g5_pmode_cur)
+		g5_switch_volt(speed_mode);
+
+	g5_pmode_cur = speed_mode;
+	ppc_proc_freq = g5_cpu_freqs[speed_mode].frequency * 1000ul;
+
+	return 0;
+}
+
+static int g5_pfunc_query_freq(void)
+{
+	struct pmf_args args;
+	u32 val = 0;
+
+	args.count = 1;
+	args.u[0].p = &val;
+	pmf_call_one(pfunc_cpu_getfreq, &args);
+	return val ? CPUFREQ_HIGH : CPUFREQ_LOW;
+}
+
+
+/*
+ * Common interface to the cpufreq core
+ */
+
+static int g5_cpufreq_verify(struct cpufreq_policy *policy)
+{
+	return cpufreq_frequency_table_verify(policy, g5_cpu_freqs);
+}
+
+static int g5_cpufreq_target(struct cpufreq_policy *policy,
+	unsigned int target_freq, unsigned int relation)
+{
+	unsigned int newstate = 0;
+	struct cpufreq_freqs freqs;
+	int rc;
+
+	if (cpufreq_frequency_table_target(policy, g5_cpu_freqs,
+			target_freq, relation, &newstate))
+		return -EINVAL;
+
+	if (g5_pmode_cur == newstate)
+		return 0;
+
+	mutex_lock(&g5_switch_mutex);
+
+	freqs.old = g5_cpu_freqs[g5_pmode_cur].frequency;
+	freqs.new = g5_cpu_freqs[newstate].frequency;
+
+	cpufreq_notify_transition(policy, &freqs, CPUFREQ_PRECHANGE);
+	rc = g5_switch_freq(newstate);
+	cpufreq_notify_transition(policy, &freqs, CPUFREQ_POSTCHANGE);
+
+	mutex_unlock(&g5_switch_mutex);
+
+	return rc;
+}
+
+static unsigned int g5_cpufreq_get_speed(unsigned int cpu)
+{
+	return g5_cpu_freqs[g5_pmode_cur].frequency;
+}
+
+static int g5_cpufreq_cpu_init(struct cpufreq_policy *policy)
+{
+	policy->cpuinfo.transition_latency = transition_latency;
+	policy->cur = g5_cpu_freqs[g5_query_freq()].frequency;
+	/* secondary CPUs are tied to the primary one by the
+	 * cpufreq core if in the secondary policy we tell it that
+	 * it actually must be one policy together with all others. */
+	cpumask_copy(policy->cpus, cpu_online_mask);
+	cpufreq_frequency_table_get_attr(g5_cpu_freqs, policy->cpu);
+
+	return cpufreq_frequency_table_cpuinfo(policy,
+		g5_cpu_freqs);
+}
+
+
+static struct cpufreq_driver g5_cpufreq_driver = {
+	.name		= "powermac",
+	.owner		= THIS_MODULE,
+	.flags		= CPUFREQ_CONST_LOOPS,
+	.init		= g5_cpufreq_cpu_init,
+	.verify		= g5_cpufreq_verify,
+	.target		= g5_cpufreq_target,
+	.get		= g5_cpufreq_get_speed,
+	.attr 		= g5_cpu_freqs_attr,
+};
+
+
+#ifdef CONFIG_PMAC_SMU
+
+static int __init g5_neo2_cpufreq_init(struct device_node *cpus)
+{
+	struct device_node *cpunode;
+	unsigned int psize, ssize;
+	unsigned long max_freq;
+	char *freq_method, *volt_method;
+	const u32 *valp;
+	u32 pvr_hi;
+	int use_volts_vdnap = 0;
+	int use_volts_smu = 0;
+	int rc = -ENODEV;
+
+	/* Check supported platforms */
+	if (of_machine_is_compatible("PowerMac8,1") ||
+	    of_machine_is_compatible("PowerMac8,2") ||
+	    of_machine_is_compatible("PowerMac9,1"))
+		use_volts_smu = 1;
+	else if (of_machine_is_compatible("PowerMac11,2"))
+		use_volts_vdnap = 1;
+	else
+		return -ENODEV;
+
+	/* Get first CPU node */
+	for (cpunode = NULL;
+	     (cpunode = of_get_next_child(cpus, cpunode)) != NULL;) {
+		const u32 *reg = of_get_property(cpunode, "reg", NULL);
+		if (reg == NULL || (*reg) != 0)
+			continue;
+		if (!strcmp(cpunode->type, "cpu"))
+			break;
+	}
+	if (cpunode == NULL) {
+		printk(KERN_ERR "cpufreq: Can't find any CPU 0 node\n");
+		return -ENODEV;
+	}
+
+	/* Check 970FX for now */
+	valp = of_get_property(cpunode, "cpu-version", NULL);
+	if (!valp) {
+		DBG("No cpu-version property !\n");
+		goto bail_noprops;
+	}
+	pvr_hi = (*valp) >> 16;
+	if (pvr_hi != 0x3c && pvr_hi != 0x44) {
+		printk(KERN_ERR "cpufreq: Unsupported CPU version\n");
+		goto bail_noprops;
+	}
+
+	/* Look for the powertune data in the device-tree */
+	g5_pmode_data = of_get_property(cpunode, "power-mode-data",&psize);
+	if (!g5_pmode_data) {
+		DBG("No power-mode-data !\n");
+		goto bail_noprops;
+	}
+	g5_pmode_max = psize / sizeof(u32) - 1;
+
+	if (use_volts_smu) {
+		const struct smu_sdbp_header *shdr;
+
+		/* Look for the FVT table */
+		shdr = smu_get_sdb_partition(SMU_SDB_FVT_ID, NULL);
+		if (!shdr)
+			goto bail_noprops;
+		g5_fvt_table = (struct smu_sdbp_fvt *)&shdr[1];
+		ssize = (shdr->len * sizeof(u32)) -
+			sizeof(struct smu_sdbp_header);
+		g5_fvt_count = ssize / sizeof(struct smu_sdbp_fvt);
+		g5_fvt_cur = 0;
+
+		/* Sanity checking */
+		if (g5_fvt_count < 1 || g5_pmode_max < 1)
+			goto bail_noprops;
+
+		g5_switch_volt = g5_smu_switch_volt;
+		volt_method = "SMU";
+	} else if (use_volts_vdnap) {
+		struct device_node *root;
+
+		root = of_find_node_by_path("/");
+		if (root == NULL) {
+			printk(KERN_ERR "cpufreq: Can't find root of "
+			       "device tree\n");
+			goto bail_noprops;
+		}
+		pfunc_set_vdnap0 = pmf_find_function(root, "set-vdnap0");
+		pfunc_vdnap0_complete =
+			pmf_find_function(root, "slewing-done");
+		if (pfunc_set_vdnap0 == NULL ||
+		    pfunc_vdnap0_complete == NULL) {
+			printk(KERN_ERR "cpufreq: Can't find required "
+			       "platform function\n");
+			goto bail_noprops;
+		}
+
+		g5_switch_volt = g5_vdnap_switch_volt;
+		volt_method = "GPIO";
+	} else {
+		g5_switch_volt = g5_dummy_switch_volt;
+		volt_method = "none";
+	}
+
+	/*
+	 * From what I see, clock-frequency is always the maximal frequency.
+	 * The current driver can not slew sysclk yet, so we really only deal
+	 * with powertune steps for now. We also only implement full freq and
+	 * half freq in this version. So far, I haven't yet seen a machine
+	 * supporting anything else.
+	 */
+	valp = of_get_property(cpunode, "clock-frequency", NULL);
+	if (!valp)
+		return -ENODEV;
+	max_freq = (*valp)/1000;
+	g5_cpu_freqs[0].frequency = max_freq;
+	g5_cpu_freqs[1].frequency = max_freq/2;
+
+	/* Set callbacks */
+	transition_latency = 12000;
+	g5_switch_freq = g5_scom_switch_freq;
+	g5_query_freq = g5_scom_query_freq;
+	freq_method = "SCOM";
+
+	/* Force apply current frequency to make sure everything is in
+	 * sync (voltage is right for example). Firmware may leave us with
+	 * a strange setting ...
+	 */
+	g5_switch_volt(CPUFREQ_HIGH);
+	msleep(10);
+	g5_pmode_cur = -1;
+	g5_switch_freq(g5_query_freq());
+
+	printk(KERN_INFO "Registering G5 CPU frequency driver\n");
+	printk(KERN_INFO "Frequency method: %s, Voltage method: %s\n",
+	       freq_method, volt_method);
+	printk(KERN_INFO "Low: %d Mhz, High: %d Mhz, Cur: %d MHz\n",
+		g5_cpu_freqs[1].frequency/1000,
+		g5_cpu_freqs[0].frequency/1000,
+		g5_cpu_freqs[g5_pmode_cur].frequency/1000);
+
+	rc = cpufreq_register_driver(&g5_cpufreq_driver);
+
+	/* We keep the CPU node on hold... hopefully, Apple G5 don't have
+	 * hotplug CPU with a dynamic device-tree ...
+	 */
+	return rc;
+
+ bail_noprops:
+	of_node_put(cpunode);
+
+	return rc;
+}
+
+#endif /* CONFIG_PMAC_SMU */
+
+
+static int __init g5_pm72_cpufreq_init(struct device_node *cpus)
+{
+	struct device_node *cpuid = NULL, *hwclock = NULL, *cpunode = NULL;
+	const u8 *eeprom = NULL;
+	const u32 *valp;
+	u64 max_freq, min_freq, ih, il;
+	int has_volt = 1, rc = 0;
+
+	DBG("cpufreq: Initializing for PowerMac7,2, PowerMac7,3 and"
+	    " RackMac3,1...\n");
+
+	/* Get first CPU node */
+	for (cpunode = NULL;
+	     (cpunode = of_get_next_child(cpus, cpunode)) != NULL;) {
+		if (!strcmp(cpunode->type, "cpu"))
+			break;
+	}
+	if (cpunode == NULL) {
+		printk(KERN_ERR "cpufreq: Can't find any CPU node\n");
+		return -ENODEV;
+	}
+
+	/* Lookup the cpuid eeprom node */
+        cpuid = of_find_node_by_path("/u3@0,f8000000/i2c@f8001000/cpuid@a0");
+	if (cpuid != NULL)
+		eeprom = of_get_property(cpuid, "cpuid", NULL);
+	if (eeprom == NULL) {
+		printk(KERN_ERR "cpufreq: Can't find cpuid EEPROM !\n");
+		rc = -ENODEV;
+		goto bail;
+	}
+
+	/* Lookup the i2c hwclock */
+	for (hwclock = NULL;
+	     (hwclock = of_find_node_by_name(hwclock, "i2c-hwclock")) != NULL;){
+		const char *loc = of_get_property(hwclock,
+				"hwctrl-location", NULL);
+		if (loc == NULL)
+			continue;
+		if (strcmp(loc, "CPU CLOCK"))
+			continue;
+		if (!of_get_property(hwclock, "platform-get-frequency", NULL))
+			continue;
+		break;
+	}
+	if (hwclock == NULL) {
+		printk(KERN_ERR "cpufreq: Can't find i2c clock chip !\n");
+		rc = -ENODEV;
+		goto bail;
+	}
+
+	DBG("cpufreq: i2c clock chip found: %s\n", hwclock->full_name);
+
+	/* Now get all the platform functions */
+	pfunc_cpu_getfreq =
+		pmf_find_function(hwclock, "get-frequency");
+	pfunc_cpu_setfreq_high =
+		pmf_find_function(hwclock, "set-frequency-high");
+	pfunc_cpu_setfreq_low =
+		pmf_find_function(hwclock, "set-frequency-low");
+	pfunc_slewing_done =
+		pmf_find_function(hwclock, "slewing-done");
+	pfunc_cpu0_volt_high =
+		pmf_find_function(hwclock, "set-voltage-high-0");
+	pfunc_cpu0_volt_low =
+		pmf_find_function(hwclock, "set-voltage-low-0");
+	pfunc_cpu1_volt_high =
+		pmf_find_function(hwclock, "set-voltage-high-1");
+	pfunc_cpu1_volt_low =
+		pmf_find_function(hwclock, "set-voltage-low-1");
+
+	/* Check we have minimum requirements */
+	if (pfunc_cpu_getfreq == NULL || pfunc_cpu_setfreq_high == NULL ||
+	    pfunc_cpu_setfreq_low == NULL || pfunc_slewing_done == NULL) {
+		printk(KERN_ERR "cpufreq: Can't find platform functions !\n");
+		rc = -ENODEV;
+		goto bail;
+	}
+
+	/* Check that we have complete sets */
+	if (pfunc_cpu0_volt_high == NULL || pfunc_cpu0_volt_low == NULL) {
+		pmf_put_function(pfunc_cpu0_volt_high);
+		pmf_put_function(pfunc_cpu0_volt_low);
+		pfunc_cpu0_volt_high = pfunc_cpu0_volt_low = NULL;
+		has_volt = 0;
+	}
+	if (!has_volt ||
+	    pfunc_cpu1_volt_high == NULL || pfunc_cpu1_volt_low == NULL) {
+		pmf_put_function(pfunc_cpu1_volt_high);
+		pmf_put_function(pfunc_cpu1_volt_low);
+		pfunc_cpu1_volt_high = pfunc_cpu1_volt_low = NULL;
+	}
+
+	/* Note: The device tree also contains a "platform-set-values"
+	 * function for which I haven't quite figured out the usage. It
+	 * might have to be called on init and/or wakeup, I'm not too sure
+	 * but things seem to work fine without it so far ...
+	 */
+
+	/* Get max frequency from device-tree */
+	valp = of_get_property(cpunode, "clock-frequency", NULL);
+	if (!valp) {
+		printk(KERN_ERR "cpufreq: Can't find CPU frequency !\n");
+		rc = -ENODEV;
+		goto bail;
+	}
+
+	max_freq = (*valp)/1000;
+
+	/* Now calculate reduced frequency by using the cpuid input freq
+	 * ratio. This requires 64 bits math unless we are willing to lose
+	 * some precision
+	 */
+	ih = *((u32 *)(eeprom + 0x10));
+	il = *((u32 *)(eeprom + 0x20));
+
+	/* Check for machines with no useful settings */
+	if (il == ih) {
+		printk(KERN_WARNING "cpufreq: No low frequency mode available"
+		       " on this model !\n");
+		rc = -ENODEV;
+		goto bail;
+	}
+
+	min_freq = 0;
+	if (ih != 0 && il != 0)
+		min_freq = (max_freq * il) / ih;
+
+	/* Sanity check */
+	if (min_freq >= max_freq || min_freq < 1000) {
+		printk(KERN_ERR "cpufreq: Can't calculate low frequency !\n");
+		rc = -ENXIO;
+		goto bail;
+	}
+	g5_cpu_freqs[0].frequency = max_freq;
+	g5_cpu_freqs[1].frequency = min_freq;
+
+	/* Set callbacks */
+	transition_latency = CPUFREQ_ETERNAL;
+	g5_switch_volt = g5_pfunc_switch_volt;
+	g5_switch_freq = g5_pfunc_switch_freq;
+	g5_query_freq = g5_pfunc_query_freq;
+
+	/* Force apply current frequency to make sure everything is in
+	 * sync (voltage is right for example). Firmware may leave us with
+	 * a strange setting ...
+	 */
+	g5_switch_volt(CPUFREQ_HIGH);
+	msleep(10);
+	g5_pmode_cur = -1;
+	g5_switch_freq(g5_query_freq());
+
+	printk(KERN_INFO "Registering G5 CPU frequency driver\n");
+	printk(KERN_INFO "Frequency method: i2c/pfunc, "
+	       "Voltage method: %s\n", has_volt ? "i2c/pfunc" : "none");
+	printk(KERN_INFO "Low: %d Mhz, High: %d Mhz, Cur: %d MHz\n",
+		g5_cpu_freqs[1].frequency/1000,
+		g5_cpu_freqs[0].frequency/1000,
+		g5_cpu_freqs[g5_pmode_cur].frequency/1000);
+
+	rc = cpufreq_register_driver(&g5_cpufreq_driver);
+ bail:
+	if (rc != 0) {
+		pmf_put_function(pfunc_cpu_getfreq);
+		pmf_put_function(pfunc_cpu_setfreq_high);
+		pmf_put_function(pfunc_cpu_setfreq_low);
+		pmf_put_function(pfunc_slewing_done);
+		pmf_put_function(pfunc_cpu0_volt_high);
+		pmf_put_function(pfunc_cpu0_volt_low);
+		pmf_put_function(pfunc_cpu1_volt_high);
+		pmf_put_function(pfunc_cpu1_volt_low);
+	}
+	of_node_put(hwclock);
+	of_node_put(cpuid);
+	of_node_put(cpunode);
+
+	return rc;
+}
+
+static int __init g5_cpufreq_init(void)
+{
+	struct device_node *cpus;
+	int rc = 0;
+
+	cpus = of_find_node_by_path("/cpus");
+	if (cpus == NULL) {
+		DBG("No /cpus node !\n");
+		return -ENODEV;
+	}
+
+	if (of_machine_is_compatible("PowerMac7,2") ||
+	    of_machine_is_compatible("PowerMac7,3") ||
+	    of_machine_is_compatible("RackMac3,1"))
+		rc = g5_pm72_cpufreq_init(cpus);
+#ifdef CONFIG_PMAC_SMU
+	else
+		rc = g5_neo2_cpufreq_init(cpus);
+#endif /* CONFIG_PMAC_SMU */
+
+	of_node_put(cpus);
+	return rc;
+}
+
+module_init(g5_cpufreq_init);
+
+
+MODULE_LICENSE("GPL");