Xen devel - Dec 2012 - [PATCH v2 4/6] ARM: psci: add support for PSCI invocations from the kernel

This patch adds support for the Power State Coordination Interface
defined by ARM, allowing Linux to request CPU-centric power-management
operations from firmware implementing the PSCI protocol.

Signed-off-by: Will Deacon <will.deacon@arm.com>
---
 arch/arm/Kconfig            |  10 +++
 arch/arm/include/asm/psci.h |  36 ++++++++
 arch/arm/kernel/Makefile    |   1 +
 arch/arm/kernel/psci.c      | 214 ++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 261 insertions(+)
 create mode 100644 arch/arm/include/asm/psci.h
 create mode 100644 arch/arm/kernel/psci.c

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index e4017ea..2a04375 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -1611,6 +1611,16 @@ config HOTPLUG_CPU
 	  Say Y here to experiment with turning CPUs off and on.  CPUs
 	  can be controlled through /sys/devices/system/cpu.
 
+config ARM_PSCI
+	bool "Support for the ARM Power State Coordination Interface (PSCI)"
+	depends on CPU_V7
+	help
+	  Say Y here if you want Linux to communicate with system firmware
+	  implementing the PSCI specification for CPU-centric power
+	  management operations described in ARM document number ARM DEN
+	  0022A ("Power State Coordination Interface System Software on
+	  ARM processors").
+
 config LOCAL_TIMERS
 	bool "Use local timer interrupts"
 	depends on SMP
diff --git a/arch/arm/include/asm/psci.h b/arch/arm/include/asm/psci.h
new file mode 100644
index 0000000..ce0dbe7
--- /dev/null
+++ b/arch/arm/include/asm/psci.h
@@ -0,0 +1,36 @@
+/*
+ * 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 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.
+ *
+ * Copyright (C) 2012 ARM Limited
+ */
+
+#ifndef __ASM_ARM_PSCI_H
+#define __ASM_ARM_PSCI_H
+
+#define PSCI_POWER_STATE_TYPE_STANDBY		0
+#define PSCI_POWER_STATE_TYPE_POWER_DOWN	1
+
+struct psci_power_state {
+	u16	id;
+	u8	type;
+	u8	affinity_level;
+};
+
+struct psci_operations {
+	int (*cpu_suspend)(struct psci_power_state state,
+			   unsigned long entry_point);
+	int (*cpu_off)(struct psci_power_state state);
+	int (*cpu_on)(unsigned long cpuid, unsigned long entry_point);
+	int (*migrate)(unsigned long cpuid);
+};
+
+extern struct psci_operations psci_ops;
+
+#endif /* __ASM_ARM_PSCI_H */
diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile
index 5bbec7b..5f3338e 100644
--- a/arch/arm/kernel/Makefile
+++ b/arch/arm/kernel/Makefile
@@ -82,5 +82,6 @@ obj-$(CONFIG_DEBUG_LL)	+= debug.o
 obj-$(CONFIG_EARLY_PRINTK)	+= early_printk.o
 
 obj-$(CONFIG_ARM_VIRT_EXT)	+= hyp-stub.o
+obj-$(CONFIG_ARM_PSCI)		+= psci.o
 
 extra-y := $(head-y) vmlinux.lds
diff --git a/arch/arm/kernel/psci.c b/arch/arm/kernel/psci.c
new file mode 100644
index 0000000..e7fe909
--- /dev/null
+++ b/arch/arm/kernel/psci.c
@@ -0,0 +1,214 @@
+/*
+ * 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 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.
+ *
+ * Copyright (C) 2012 ARM Limited
+ *
+ * Author: Will Deacon <will.deacon@arm.com>
+ */
+
+#define pr_fmt(fmt) "psci: " fmt
+
+#include <linux/init.h>
+#include <linux/of.h>
+
+#include <asm/compiler.h>
+#include <asm/errno.h>
+#include <asm/opcodes-sec.h>
+#include <asm/opcodes-virt.h>
+#include <asm/psci.h>
+
+struct psci_operations psci_ops;
+
+static int (*invoke_psci_fn)(u32, u32, u32, u32);
+
+enum psci_function {
+	PSCI_FN_CPU_SUSPEND,
+	PSCI_FN_CPU_ON,
+	PSCI_FN_CPU_OFF,
+	PSCI_FN_MIGRATE,
+	PSCI_FN_MAX,
+};
+
+static u32 psci_function_id[PSCI_FN_MAX];
+
+#define PSCI_RET_SUCCESS		0
+#define PSCI_RET_EOPNOTSUPP		-1
+#define PSCI_RET_EINVAL			-2
+#define PSCI_RET_EPERM			-3
+
+static int psci_to_linux_errno(int errno)
+{
+	switch (errno) {
+	case PSCI_RET_SUCCESS:
+		return 0;
+	case PSCI_RET_EOPNOTSUPP:
+		return -EOPNOTSUPP;
+	case PSCI_RET_EINVAL:
+		return -EINVAL;
+	case PSCI_RET_EPERM:
+		return -EPERM;
+	};
+
+	return -EINVAL;
+}
+
+#define PSCI_POWER_STATE_ID_MASK	0xffff
+#define PSCI_POWER_STATE_ID_SHIFT	0
+#define PSCI_POWER_STATE_TYPE_MASK	0x1
+#define PSCI_POWER_STATE_TYPE_SHIFT	16
+#define PSCI_POWER_STATE_AFFL_MASK	0x3
+#define PSCI_POWER_STATE_AFFL_SHIFT	24
+
+static u32 psci_power_state_pack(struct psci_power_state state)
+{
+	return	((state.id & PSCI_POWER_STATE_ID_MASK)
+			<< PSCI_POWER_STATE_ID_SHIFT)	|
+		((state.type & PSCI_POWER_STATE_TYPE_MASK)
+			<< PSCI_POWER_STATE_TYPE_SHIFT)	|
+		((state.affinity_level & PSCI_POWER_STATE_AFFL_MASK)
+			<< PSCI_POWER_STATE_AFFL_SHIFT);
+}
+
+/*
+ * The following two functions are invoked via the invoke_psci_fn pointer
+ * and will not be inlined, allowing us to piggyback on the AAPCS.
+ */
+static int __invoke_psci_fn_hvc(u32 function_id, u32 arg0, u32 arg1, u32 arg2)
+{
+	asm volatile(
+			__asmeq("%0", "r0")
+			__asmeq("%1", "r1")
+			__asmeq("%2", "r2")
+			__asmeq("%3", "r3")
+			__HVC(0)
+		: "+r" (function_id)
+		: "r" (arg0), "r" (arg1), "r" (arg2));
+
+	return function_id;
+}
+
+static int __invoke_psci_fn_smc(u32 function_id, u32 arg0, u32 arg1, u32 arg2)
+{
+	asm volatile(
+			__asmeq("%0", "r0")
+			__asmeq("%1", "r1")
+			__asmeq("%2", "r2")
+			__asmeq("%3", "r3")
+			__SMC(0)
+		: "+r" (function_id)
+		: "r" (arg0), "r" (arg1), "r" (arg2));
+
+	return function_id;
+}
+
+static int psci_cpu_suspend(struct psci_power_state state,
+			    unsigned long entry_point)
+{
+	int err;
+	u32 fn, power_state;
+
+	fn = psci_function_id[PSCI_FN_CPU_SUSPEND];
+	power_state = psci_power_state_pack(state);
+	err = invoke_psci_fn(fn, power_state, (u32)entry_point, 0);
+	return psci_to_linux_errno(err);
+}
+
+static int psci_cpu_off(struct psci_power_state state)
+{
+	int err;
+	u32 fn, power_state;
+
+	fn = psci_function_id[PSCI_FN_CPU_OFF];
+	power_state = psci_power_state_pack(state);
+	err = invoke_psci_fn(fn, power_state, 0, 0);
+	return psci_to_linux_errno(err);
+}
+
+static int psci_cpu_on(unsigned long cpuid, unsigned long entry_point)
+{
+	int err;
+	u32 fn;
+
+	fn = psci_function_id[PSCI_FN_CPU_ON];
+	err = invoke_psci_fn(fn, cpuid, entry_point, 0);
+	return psci_to_linux_errno(err);
+}
+
+static int psci_migrate(unsigned long cpuid)
+{
+	int err;
+	u32 fn;
+
+	fn = psci_function_id[PSCI_FN_MIGRATE];
+	err = invoke_psci_fn(fn, cpuid, 0, 0);
+	return psci_to_linux_errno(err);
+}
+
+static const struct of_device_id psci_of_match[] __initconst = {
+	{ .compatible = "arm,psci",	},
+	{},
+};
+
+static int __init psci_init(void)
+{
+	struct device_node *np;
+	const char *method;
+	u32 base, id;
+
+	np = of_find_matching_node(NULL, psci_of_match);
+	if (!np)
+		return 0;
+
+	pr_info("probing function IDs from device-tree\n");
+
+	if (of_property_read_u32(np, "function-base", &base)) {
+		pr_warning("missing \"function-base\" property\n");
+		goto out_put_node;
+	}
+
+	if (of_property_read_string(np, "method", &method)) {
+		pr_warning("missing \"method\" property\n");
+		goto out_put_node;
+	}
+
+	if (!strcmp("hvc", method)) {
+		invoke_psci_fn = __invoke_psci_fn_hvc;
+	} else if (!strcmp("smc", method)) {
+		invoke_psci_fn = __invoke_psci_fn_smc;
+	} else {
+		pr_warning("invalid \"method\" property: %s\n", method);
+		goto out_put_node;
+	}
+
+	if (!of_property_read_u32(np, "cpu_suspend", &id)) {
+		psci_function_id[PSCI_FN_CPU_SUSPEND] = base + id;
+		psci_ops.cpu_suspend = psci_cpu_suspend;
+	}
+
+	if (!of_property_read_u32(np, "cpu_off", &id)) {
+		psci_function_id[PSCI_FN_CPU_OFF] = base + id;
+		psci_ops.cpu_off = psci_cpu_off;
+	}
+
+	if (!of_property_read_u32(np, "cpu_on", &id)) {
+		psci_function_id[PSCI_FN_CPU_ON] = base + id;
+		psci_ops.cpu_on = psci_cpu_on;
+	}
+
+	if (!of_property_read_u32(np, "migrate", &id)) {
+		psci_function_id[PSCI_FN_MIGRATE] = base + id;
+		psci_ops.migrate = psci_migrate;
+	}
+
+out_put_node:
+	of_node_put(np);
+	return 0;
+}
+early_initcall(psci_init);
-- 
1.8.0

On Mon, 17 Dec 2012, Will Deacon wrote:
> This patch adds support for the Power State Coordination Interface
> defined by ARM, allowing Linux to request CPU-centric power-management
> operations from firmware implementing the PSCI protocol.
> 
> Signed-off-by: Will Deacon <will.deacon@arm.com>
[...]
> +/*
> + * The following two functions are invoked via the invoke_psci_fn pointer
> + * and will not be inlined, allowing us to piggyback on the AAPCS.
> + */
To make sure the code is always in sync with the intent, you could mark 
those with  noinline as well.
> +static int __invoke_psci_fn_hvc(u32 function_id, u32 arg0, u32 arg1, u32
arg2)
> +{
> +	asm volatile(
> +			__asmeq("%0", "r0")
> +			__asmeq("%1", "r1")
> +			__asmeq("%2", "r2")
> +			__asmeq("%3", "r3")
> +			__HVC(0)
> +		: "+r" (function_id)
> +		: "r" (arg0), "r" (arg1), "r" (arg2));
> +
> +	return function_id;
> +}
> +
> +static int __invoke_psci_fn_smc(u32 function_id, u32 arg0, u32 arg1, u32
arg2)
> +{
> +	asm volatile(
> +			__asmeq("%0", "r0")
> +			__asmeq("%1", "r1")
> +			__asmeq("%2", "r2")
> +			__asmeq("%3", "r3")
> +			__SMC(0)
> +		: "+r" (function_id)
> +		: "r" (arg0), "r" (arg1), "r" (arg2));
> +
> +	return function_id;
> +}
> +
> +static int psci_cpu_suspend(struct psci_power_state state,
> +			    unsigned long entry_point)
> +{
> +	int err;
> +	u32 fn, power_state;
> +
> +	fn = psci_function_id[PSCI_FN_CPU_SUSPEND];
> +	power_state = psci_power_state_pack(state);
> +	err = invoke_psci_fn(fn, power_state, (u32)entry_point, 0);
Why do you need the u32 cast here?
> +	return psci_to_linux_errno(err);
> +}
> +
> +static int psci_cpu_off(struct psci_power_state state)
> +{
> +	int err;
> +	u32 fn, power_state;
> +
> +	fn = psci_function_id[PSCI_FN_CPU_OFF];
> +	power_state = psci_power_state_pack(state);
> +	err = invoke_psci_fn(fn, power_state, 0, 0);
> +	return psci_to_linux_errno(err);
> +}
> +
> +static int psci_cpu_on(unsigned long cpuid, unsigned long entry_point)
> +{
> +	int err;
> +	u32 fn;
> +
> +	fn = psci_function_id[PSCI_FN_CPU_ON];
> +	err = invoke_psci_fn(fn, cpuid, entry_point, 0);
> +	return psci_to_linux_errno(err);
> +}
> +
> +static int psci_migrate(unsigned long cpuid)
> +{
> +	int err;
> +	u32 fn;
> +
> +	fn = psci_function_id[PSCI_FN_MIGRATE];
> +	err = invoke_psci_fn(fn, cpuid, 0, 0);
> +	return psci_to_linux_errno(err);
> +}
> +
> +static const struct of_device_id psci_of_match[] __initconst = {
> +	{ .compatible = "arm,psci",	},
> +	{},
> +};
> +
> +static int __init psci_init(void)
> +{
> +	struct device_node *np;
> +	const char *method;
> +	u32 base, id;
> +
> +	np = of_find_matching_node(NULL, psci_of_match);
> +	if (!np)
> +		return 0;
> +
> +	pr_info("probing function IDs from device-tree\n");
Having "probing function IDs from device-tree" in the middle of a
kernel
log isn''t very informative.  Better make this more useful or remove it.
> +
> +	if (of_property_read_u32(np, "function-base", &base)) {
> +		pr_warning("missing \"function-base\" property\n");
Same thing here: this lacks context in a kernel log.
And so on for the other occurrences.


Nicolas

On Mon, Dec 17, 2012 at 08:51:27PM +0000, Nicolas Pitre
wrote:
> On Mon, 17 Dec 2012, Will Deacon wrote:
> 
> > This patch adds support for the Power State Coordination Interface
> > defined by ARM, allowing Linux to request CPU-centric power-management
> > operations from firmware implementing the PSCI protocol.
> > 
> > Signed-off-by: Will Deacon <will.deacon@arm.com>
> 
> [...]
> 
> > +/*
> > + * The following two functions are invoked via the invoke_psci_fn
pointer
> > + * and will not be inlined, allowing us to piggyback on the AAPCS.
> > + */
> 
> To make sure the code is always in sync with the intent, you could mark 
> those with  noinline as well.
Can do.
> > +static int __invoke_psci_fn_hvc(u32 function_id, u32 arg0, u32 arg1,
u32 arg2)
> > +{
> > +	asm volatile(
> > +			__asmeq("%0", "r0")
> > +			__asmeq("%1", "r1")
> > +			__asmeq("%2", "r2")
> > +			__asmeq("%3", "r3")
> > +			__HVC(0)
> > +		: "+r" (function_id)
> > +		: "r" (arg0), "r" (arg1), "r"
(arg2));
> > +
> > +	return function_id;
> > +}
> > +
> > +static int __invoke_psci_fn_smc(u32 function_id, u32 arg0, u32 arg1,
u32 arg2)
> > +{
> > +	asm volatile(
> > +			__asmeq("%0", "r0")
> > +			__asmeq("%1", "r1")
> > +			__asmeq("%2", "r2")
> > +			__asmeq("%3", "r3")
> > +			__SMC(0)
> > +		: "+r" (function_id)
> > +		: "r" (arg0), "r" (arg1), "r"
(arg2));
> > +
> > +	return function_id;
> > +}
> > +
> > +static int psci_cpu_suspend(struct psci_power_state state,
> > +			    unsigned long entry_point)
> > +{
> > +	int err;
> > +	u32 fn, power_state;
> > +
> > +	fn = psci_function_id[PSCI_FN_CPU_SUSPEND];
> > +	power_state = psci_power_state_pack(state);
> > +	err = invoke_psci_fn(fn, power_state, (u32)entry_point, 0);
> 
> Why do you need the u32 cast here?
That''s a hangover from when entry_point was a void *. I''ll fix
that, thanks.
> > +static int __init psci_init(void)
> > +{
> > +	struct device_node *np;
> > +	const char *method;
> > +	u32 base, id;
> > +
> > +	np = of_find_matching_node(NULL, psci_of_match);
> > +	if (!np)
> > +		return 0;
> > +
> > +	pr_info("probing function IDs from device-tree\n");
> 
> Having "probing function IDs from device-tree" in the middle of a
kernel
> log isn''t very informative.  Better make this more useful or
remove it.
> 
> > +
> > +	if (of_property_read_u32(np, "function-base", &base))
{
> > +		pr_warning("missing \"function-base\"
property\n");
> 
> Same thing here: this lacks context in a kernel log.
> And so on for the other occurrences.
Actually, these are all prefixed with "psci: " thanks to the pr_fmt
definition at the top of the file. I can remove them if you like, but then
it''s not obvious which parts of the PSCI code are available from
looking at
a kernel boot log.

Cheers for the review,

Will

On Tue, 18 Dec 2012, Will Deacon wrote:
> On Mon, Dec 17, 2012 at 08:51:27PM +0000, Nicolas Pitre wrote:
> > On Mon, 17 Dec 2012, Will Deacon wrote:
> > > +static int psci_cpu_suspend(struct psci_power_state state,
> > > +			    unsigned long entry_point)
> > > +{
> > > +	int err;
> > > +	u32 fn, power_state;
> > > +
> > > +	fn = psci_function_id[PSCI_FN_CPU_SUSPEND];
> > > +	power_state = psci_power_state_pack(state);
> > > +	err = invoke_psci_fn(fn, power_state, (u32)entry_point, 0);
> > 
> > Why do you need the u32 cast here?
> 
> That''s a hangover from when entry_point was a void *.
I''ll fix that, thanks.
Hopefully you didn''t pass virtual pointers to the PSCI call, did you? 
:-)
> > > +static int __init psci_init(void)
> > > +{
> > > +	struct device_node *np;
> > > +	const char *method;
> > > +	u32 base, id;
> > > +
> > > +	np = of_find_matching_node(NULL, psci_of_match);
> > > +	if (!np)
> > > +		return 0;
> > > +
> > > +	pr_info("probing function IDs from device-tree\n");
> > 
> > Having "probing function IDs from device-tree" in the middle
of a kernel
> > log isn''t very informative.  Better make this more useful or
remove it.
> > 
> > > +
> > > +	if (of_property_read_u32(np, "function-base",
&base)) {
> > > +		pr_warning("missing \"function-base\"
property\n");
> > 
> > Same thing here: this lacks context in a kernel log.
> > And so on for the other occurrences.
> 
> Actually, these are all prefixed with "psci: " thanks to the
pr_fmt
> definition at the top of the file.
Ah, goodie!  No more issue then.


Nicolas

On Tue, Dec 18, 2012 at 09:59:45PM +0000, Nicolas Pitre
wrote:
> On Tue, 18 Dec 2012, Will Deacon wrote:
> > On Mon, Dec 17, 2012 at 08:51:27PM +0000, Nicolas Pitre wrote:
> > > On Mon, 17 Dec 2012, Will Deacon wrote:
> > > > +static int psci_cpu_suspend(struct psci_power_state state,
> > > > +			    unsigned long entry_point)
> > > > +{
> > > > +	int err;
> > > > +	u32 fn, power_state;
> > > > +
> > > > +	fn = psci_function_id[PSCI_FN_CPU_SUSPEND];
> > > > +	power_state = psci_power_state_pack(state);
> > > > +	err = invoke_psci_fn(fn, power_state, (u32)entry_point,
0);
> > > 
> > > Why do you need the u32 cast here?
> > 
> > That''s a hangover from when entry_point was a void *.
I''ll fix that, thanks.
> 
> Hopefully you didn''t pass virtual pointers to the PSCI call, did
you?  :-)
...and I''d have gotten away with it if it wasn''t for those
meddling kids!

It was also made worse by Marc''s code working first time too (after I
blamed
the firmware like any sane kernel hacker would do :)

Will