[PATCH v3 0/5] ARM64: juno: add SCPI mailbox protocol, clock and CPUFreq support

[PATCH v3 0/5] ARM64: juno: add SCPI mailbox protocol, clock and CPUFreq support

[Sudeep Holla]

Hi,

This patch series adds support for:
  1. SCPI(System Control and Power Interface) mailbox protocol driver.
     It uses ARM MHU mailbox controller driver on Juno but can work with
     any mailbox controllers using standard mailbox APIs
  2. Add support for clocks provided by SCP firmware through the SCPI
     interface
  3. Using the existing arm_big_little cpufreq driver and the newly
     added SCPI clock driver, it also adds support for CPU DVFS on
     ARM64 JUNO development platforms.

The SCPI protocol document is available @[1],[2]

I am not sure which tree would be appropriate to merge this. The
clock and cpufreq drivers depend on the SCPI protocol header.

Changes v2->v3:
	- Minor fix in SCPI driver and added Tixy's reviewed-by tag
	- Updated scpi clock driver to incorporate all the comments from
	  Stephen
	- Added Viresh's ack

Changes v1->v2:
	- Updated the token handling in scpi driver as per Tixy's
	  suggestion along with other review comments
	- Removed multiple drivers in scpi clock as Lorenzo suggested
	- Added free_opp_table in scpi-cpufreq as Viresh suggested
	- Separated the DT binding document
	- Moved SCPI protocol driver to drivers/firmware

Regards,
Sudeep

[1] http://community.arm.com/servlet/JiveServlet/download/8401-45-18326/DUI0922B_scp_message_interface.pdf
[2] https://wiki.linaro.org/ARM/Juno?action=AttachFile&do=get&target=DUI0922B_scp_message_interface.pdf
v1: https://lkml.org/lkml/2015/4/27/232
v2: https://lkml.org/lkml/2015/5/14/470

Sudeep Holla (5):
  Documentation: add DT binding for ARM System Control and Power
    Interface(SCPI) protocol
  firmware: add support for ARM System Control and Power Interface(SCPI)
    protocol
  clk: add support for clocks provided by SCP(System Control Processor)
  clk: scpi: add support for cpufreq virtual device
  cpufreq: arm_big_little: add SCPI interface driver

 Documentation/devicetree/bindings/arm/arm,scpi.txt | 121 ++++
 drivers/clk/Kconfig                                |  10 +
 drivers/clk/Makefile                               |   1 +
 drivers/clk/clk-scpi.c                             | 299 +++++++++
 drivers/cpufreq/Kconfig.arm                        |   9 +
 drivers/cpufreq/Makefile                           |   1 +
 drivers/cpufreq/scpi-cpufreq.c                     | 124 ++++
 drivers/firmware/Kconfig                           |  19 +
 drivers/firmware/Makefile                          |   1 +
 drivers/firmware/arm_scpi.c                        | 709 +++++++++++++++++++++
 include/linux/scpi_protocol.h                      |  62 ++
 11 files changed, 1356 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/arm,scpi.txt
 create mode 100644 drivers/clk/clk-scpi.c
 create mode 100644 drivers/cpufreq/scpi-cpufreq.c
 create mode 100644 drivers/firmware/arm_scpi.c
 create mode 100644 include/linux/scpi_protocol.h

--
1.9.1

[PATCH v3 1/5] Documentation: add DT binding for ARM System Control and Power Interface(SCPI) protocol

[Sudeep Holla]

This patch adds devicetree binding for System Control and Power
Interface (SCPI) Message Protocol used between the Application Cores(AP)
and the System Control Processor(SCP). The MHU peripheral provides a
mechanism for inter-processor communication between SCP's M3 processor
and AP.

SCP offers control and management of the core/cluster power states,
various power domain DVFS including the core/cluster, certain system
clocks configuration, thermal sensors and many others.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
CC: Jassi Brar <jassisinghbrar@gmail.com>
Cc: Liviu Dudau <Liviu.Dudau@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
Cc: devicetree at vger.kernel.org
---
 Documentation/devicetree/bindings/arm/arm,scpi.txt | 121 +++++++++++++++++++++
 1 file changed, 121 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/arm,scpi.txt

diff --git a/Documentation/devicetree/bindings/arm/arm,scpi.txt b/Documentation/devicetree/bindings/arm/arm,scpi.txt
new file mode 100644
index 000000000000..5db235f69e54
--- /dev/null
+++ b/Documentation/devicetree/bindings/arm/arm,scpi.txt
@@ -0,0 +1,121 @@
+System Control and Power Interface (SCPI) Message Protocol
+----------------------------------------------------------
+
+Required properties:
+
+- compatible : should be "arm,scpi"
+- mboxes: List of phandle and mailbox channel specifiers
+- shmem : List of phandle pointing to the shared memory(SHM) area between the
+	  processors using these mailboxes for IPC, one for each mailbox
+
+See Documentation/devicetree/bindings/mailbox/mailbox.txt
+for more details about the generic mailbox controller and
+client driver bindings.
+
+Clock bindings for the clocks based on SCPI Message Protocol
+------------------------------------------------------------
+
+This binding uses the common clock binding[1].
+
+Required properties:
+- compatible : shall be one of the following:
+	"arm,scpi-clocks" - for the container node with all the clocks
+		based on the SCPI protocol
+	"arm,scpi-dvfs" - all the clocks that are variable and index based.
+		These clocks don't provide the full range between the limits
+		but only discrete points within the range. The firmware
+		provides the mapping for each such operating frequency and the
+		index associated with it. The firmware also manages the
+		voltage scaling appropriately with the clock scaling.
+	"arm,scpi-clk" - all the clocks that are variable and provide full
+		range within the specified range. The firmware provides the
+		supported range for each clock.
+
+Required properties for all clocks(all from common clock binding):
+- #clock-cells : should be set to 1 as each of the SCPI clocks have multiple
+	outputs. The clock specifier will be the index to an entry in the list
+	of output clocks.
+- clock-output-names : shall be the corresponding names of the outputs.
+- clock-indices: The identifyng number for the clocks(clock_id) in the node as
+	expected by the firmware. It can be non linear and hence provide the
+	mapping	of identifiers into the clock-output-names array.
+
+[1] Documentation/devicetree/bindings/clock/clock-bindings.txt
+
+Example:
+
+sram: sram at 50000000 {
+	compatible = "arm,juno-sram-ns", "mmio-sram";
+	reg = <0x0 0x50000000 0x0 0x10000>;
+
+	#address-cells = <1>;
+	#size-cells = <1>;
+	ranges = <0 0x0 0x50000000 0x10000>;
+
+	cpu_scp_lpri: scp-shmem at 0 {
+		compatible = "arm,juno-scp-shmem";
+		reg = <0x0 0x200>;
+	};
+
+	cpu_scp_hpri: scp-shmem at 200 {
+		compatible = "arm,juno-scp-shmem";
+		reg = <0x200 0x200>;
+	};
+};
+
+mailbox: mailbox0 at 40000000 {
+	....
+	#mbox-cells = <1>;
+};
+
+scpi_protocol: scpi at 2e000000 {
+	compatible = "arm,scpi";
+	mboxes = <&mailbox 0 &mailbox 1>;
+	shmem = <&cpu_scp_lpri &cpu_scp_hpri>;
+
+	clocks {
+		compatible = "arm,scpi-clocks";
+
+		scpi_dvfs: scpi_clocks at 0 {
+			compatible = "arm,scpi-dvfs";
+			#clock-cells = <1>;
+			clock-indices = <0>, <1>, <2>;
+			clock-output-names = "vbig", "vlittle", "vgpu";
+		};
+		scpi_clk: scpi_clocks at 3 {
+			compatible = "arm,scpi-clk";
+			#clock-cells = <1>;
+			clock-indices = <3>, <4>;
+			clock-output-names = "pxlclk0", "pxlclk1";
+		};
+	};
+};
+
+cpu at 0 {
+	...
+	reg = <0 0>;
+	clocks = <&scpi_dvfs 0>;
+	clock-names = "big";
+};
+
+hdlcd at 7ff60000 {
+	...
+	reg = <0 0x7ff60000 0 0x1000>;
+	clocks = <&scpi_clk 1>;
+	clock-names = "pxlclk";
+};
+
+In the above example, the #clock-cells is set to 1 as required.
+scpi_dvfs has 3 output clocks namely: vbig, vlittle and vgpu with 0, 1
+and 2 as clock-indices. scpi_clk has 2 output clocks namely: pxlclk0 and
+pxlclk1 with 3 and 4 as clock-indices.
+
+The first consumer in the example is cpu at 0 and it has vbig as input clock.
+The index '0' in the clock specifier here points to the first entry in the
+output clocks of scpi_dvfs for which clock_id asrequired by the firmware
+is 0.
+
+Similarly the second example is hdlcd at 7ff60000 and it has pxlclk0 as input
+clock. The index '1' in the clock specifier here points to the second entry
+in the output clocks of scpi_clocks for which clock_id as required by the
+firmware is 4.
-- 
1.9.1

[PATCH v3 1/5] Documentation: add DT binding for ARM System Control and Power Interface(SCPI) protocol

[Mark Rutland]

On Wed, May 27, 2015 at 10:53:14AM +0100, Sudeep Holla wrote:
> This patch adds devicetree binding for System Control and Power
> Interface (SCPI) Message Protocol used between the Application Cores(AP)
> and the System Control Processor(SCP). The MHU peripheral provides a
> mechanism for inter-processor communication between SCP's M3 processor
> and AP.
> 
> SCP offers control and management of the core/cluster power states,
> various power domain DVFS including the core/cluster, certain system
> clocks configuration, thermal sensors and many others.
> 
> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Mark Rutland <mark.rutland@arm.com>
> CC: Jassi Brar <jassisinghbrar@gmail.com>
> Cc: Liviu Dudau <Liviu.Dudau@arm.com>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
> Cc: devicetree at vger.kernel.org
> ---
>  Documentation/devicetree/bindings/arm/arm,scpi.txt | 121 +++++++++++++++++++++
>  1 file changed, 121 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/arm/arm,scpi.txt
> 
> diff --git a/Documentation/devicetree/bindings/arm/arm,scpi.txt b/Documentation/devicetree/bindings/arm/arm,scpi.txt
> new file mode 100644
> index 000000000000..5db235f69e54
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/arm/arm,scpi.txt
> @@ -0,0 +1,121 @@
> +System Control and Power Interface (SCPI) Message Protocol
> +----------------------------------------------------------

Could we refer to the documentation for SCPI similarly to what we do
in the PSCI binding (i.e. give the document number and title so people
can search for it)?

> +
> +Required properties:
> +
> +- compatible : should be "arm,scpi"
> +- mboxes: List of phandle and mailbox channel specifiers

How many, in which order, and what are they used for?

> +- shmem : List of phandle pointing to the shared memory(SHM) area between the
> +	  processors using these mailboxes for IPC, one for each mailbox

When you say "shared memory", what exactly are we referring to? It looks
like we refer to SRAM areas?

> +
> +See Documentation/devicetree/bindings/mailbox/mailbox.txt
> +for more details about the generic mailbox controller and
> +client driver bindings.
> +
> +Clock bindings for the clocks based on SCPI Message Protocol
> +------------------------------------------------------------
> +
> +This binding uses the common clock binding[1].
> +
> +Required properties:
> +- compatible : shall be one of the following:

Nit: s/be/include/

> +	"arm,scpi-clocks" - for the container node with all the clocks
> +		based on the SCPI protocol

Please separate this from the description of its sub-nodes (e.g. have a
sub-nodes heading under "arm,scpi-clocks").

Is there any reason to have this container if we're only going to place
the two clock controllers described below within it? Can't we just place
them directly under the SCPI node?

> +	"arm,scpi-dvfs" - all the clocks that are variable and index based.
> +		These clocks don't provide the full range between the limits
> +		but only discrete points within the range. The firmware
> +		provides the mapping for each such operating frequency and the
> +		index associated with it. The firmware also manages the
> +		voltage scaling appropriately with the clock scaling.

Could this be "arm,scpi-dvfs-clocks"?

> +	"arm,scpi-clk" - all the clocks that are variable and provide full
> +		range within the specified range. The firmware provides the
> +		supported range for each clock.

Likewise "arm,scpi-variable-clocks"?

Using "arm,scpi-clk" makes it sound like there's a single clock output.

> +
> +Required properties for all clocks(all from common clock binding):
> +- #clock-cells : should be set to 1 as each of the SCPI clocks have multiple
> +	outputs. The clock specifier will be the index to an entry in the list
> +	of output clocks.
> +- clock-output-names : shall be the corresponding names of the outputs.
> +- clock-indices: The identifyng number for the clocks(clock_id) in the node as
> +	expected by the firmware. It can be non linear and hence provide the
> +	mapping	of identifiers into the clock-output-names array.
> +
> +[1] Documentation/devicetree/bindings/clock/clock-bindings.txt
> +
> +Example:
> +
> +sram: sram at 50000000 {
> +	compatible = "arm,juno-sram-ns", "mmio-sram";

Nit: undocumented compatible string.

> +	reg = <0x0 0x50000000 0x0 0x10000>;
> +
> +	#address-cells = <1>;
> +	#size-cells = <1>;
> +	ranges = <0 0x0 0x50000000 0x10000>;
> +
> +	cpu_scp_lpri: scp-shmem at 0 {
> +		compatible = "arm,juno-scp-shmem";

Nit: undocumented compatible string.

> +		reg = <0x0 0x200>;
> +	};
> +
> +	cpu_scp_hpri: scp-shmem at 200 {
> +		compatible = "arm,juno-scp-shmem";
> +		reg = <0x200 0x200>;
> +	};
> +};
> +
> +mailbox: mailbox0 at 40000000 {
> +	....
> +	#mbox-cells = <1>;
> +};
> +
> +scpi_protocol: scpi at 2e000000 {
> +	compatible = "arm,scpi";
> +	mboxes = <&mailbox 0 &mailbox 1>;
> +	shmem = <&cpu_scp_lpri &cpu_scp_hpri>;
> +
> +	clocks {
> +		compatible = "arm,scpi-clocks";
> +
> +		scpi_dvfs: scpi_clocks at 0 {
> +			compatible = "arm,scpi-dvfs";
> +			#clock-cells = <1>;
> +			clock-indices = <0>, <1>, <2>;
> +			clock-output-names = "vbig", "vlittle", "vgpu";
> +		};
> +		scpi_clk: scpi_clocks at 3 {
> +			compatible = "arm,scpi-clk";
> +			#clock-cells = <1>;
> +			clock-indices = <3>, <4>;
> +			clock-output-names = "pxlclk0", "pxlclk1";
> +		};
> +	};
> +};
> +
> +cpu at 0 {
> +	...
> +	reg = <0 0>;
> +	clocks = <&scpi_dvfs 0>;
> +	clock-names = "big";

This isn't documented anywhere, and I can't see any code in this series
using it. The Juno dts doesn't seem to have CPU clocks, and nothing in
this series adds them -- have I missed a series?

Also, I'm not keen on using software-defined names (e.g. "big") for CPU
clocks, as it doesn't stictly relate to the hardware. That said, the set
of clocks, their names, and how they related to CPUs is
implementation-specific, which is somewhat painful. I'm not sure how
we cater for that with generic software.

Thanks,
Mark.

[PATCH v3 1/5] Documentation: add DT binding for ARM System Control and Power Interface(SCPI) protocol

[Sudeep Holla]

Hi Mark,

Thanks for the review.

On 27/05/15 14:37, Mark Rutland wrote:
> On Wed, May 27, 2015 at 10:53:14AM +0100, Sudeep Holla wrote:
>> This patch adds devicetree binding for System Control and Power
>> Interface (SCPI) Message Protocol used between the Application Cores(AP)
>> and the System Control Processor(SCP). The MHU peripheral provides a
>> mechanism for inter-processor communication between SCP's M3 processor
>> and AP.
>>
>> SCP offers control and management of the core/cluster power states,
>> various power domain DVFS including the core/cluster, certain system
>> clocks configuration, thermal sensors and many others.
>>
>> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
>> Cc: Rob Herring <robh+dt@kernel.org>
>> Cc: Mark Rutland <mark.rutland@arm.com>
>> CC: Jassi Brar <jassisinghbrar@gmail.com>
>> Cc: Liviu Dudau <Liviu.Dudau@arm.com>
>> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>> Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
>> Cc: devicetree at vger.kernel.org
>> ---
>>   Documentation/devicetree/bindings/arm/arm,scpi.txt | 121 +++++++++++++++++++++
>>   1 file changed, 121 insertions(+)
>>   create mode 100644 Documentation/devicetree/bindings/arm/arm,scpi.txt
>>
>> diff --git a/Documentation/devicetree/bindings/arm/arm,scpi.txt b/Documentation/devicetree/bindings/arm/arm,scpi.txt
>> new file mode 100644
>> index 000000000000..5db235f69e54
>> --- /dev/null
>> +++ b/Documentation/devicetree/bindings/arm/arm,scpi.txt
>> @@ -0,0 +1,121 @@
>> +System Control and Power Interface (SCPI) Message Protocol
>> +----------------------------------------------------------
>
> Could we refer to the documentation for SCPI similarly to what we do
> in the PSCI binding (i.e. give the document number and title so people
> can search for it)?
>

Yes that's the plan but was waiting for a stable URL(currently it's not
in ARM Infocenter and simple google search for document/title doesn't
seem to work). For now I will include the URL too and may be remove or
fix it in future.

>> +
>> +Required properties:
>> +
>> +- compatible : should be "arm,scpi"
>> +- mboxes: List of phandle and mailbox channel specifiers
>
> How many, in which order, and what are they used for?
>

Ok will update.

>> +- shmem : List of phandle pointing to the shared memory(SHM) area between the
>> +	  processors using these mailboxes for IPC, one for each mailbox
>
> When you say "shared memory", what exactly are we referring to? It looks
> like we refer to SRAM areas?
>

It can be any memory reserved for the purpose of this communication via
MHU between the Application Processors and the remote System Control
Processor. On most platforms with MHU it happens to be SRAM like in Juno
and Fijitsu MB86S7X

>> +
>> +See Documentation/devicetree/bindings/mailbox/mailbox.txt
>> +for more details about the generic mailbox controller and
>> +client driver bindings.
>> +
>> +Clock bindings for the clocks based on SCPI Message Protocol
>> +------------------------------------------------------------
>> +
>> +This binding uses the common clock binding[1].
>> +
>> +Required properties:
>> +- compatible : shall be one of the following:
>
> Nit: s/be/include/
>
>> +	"arm,scpi-clocks" - for the container node with all the clocks
>> +		based on the SCPI protocol
>
> Please separate this from the description of its sub-nodes (e.g. have a
> sub-nodes heading under "arm,scpi-clocks").
>

OK, I will update.

> Is there any reason to have this container if we're only going to place
> the two clock controllers described below within it? Can't we just place
> them directly under the SCPI node?
>

SCP also supports power domains, sensors and other misc features. I
wanted to avoid parsing through all the nodes in the clock driver once
they are added. Is that not recommended ?

>> +	"arm,scpi-dvfs" - all the clocks that are variable and index based.
>> +		These clocks don't provide the full range between the limits
>> +		but only discrete points within the range. The firmware
>> +		provides the mapping for each such operating frequency and the
>> +		index associated with it. The firmware also manages the
>> +		voltage scaling appropriately with the clock scaling.
>
> Could this be "arm,scpi-dvfs-clocks"?
>
>> +	"arm,scpi-clk" - all the clocks that are variable and provide full
>> +		range within the specified range. The firmware provides the
>> +		supported range for each clock.
>
> Likewise "arm,scpi-variable-clocks"?
>
> Using "arm,scpi-clk" makes it sound like there's a single clock output.
>
>> +
>> +Required properties for all clocks(all from common clock binding):
>> +- #clock-cells : should be set to 1 as each of the SCPI clocks have multiple
>> +	outputs. The clock specifier will be the index to an entry in the list
>> +	of output clocks.
>> +- clock-output-names : shall be the corresponding names of the outputs.
>> +- clock-indices: The identifyng number for the clocks(clock_id) in the node as
>> +	expected by the firmware. It can be non linear and hence provide the
>> +	mapping	of identifiers into the clock-output-names array.
>> +
>> +[1] Documentation/devicetree/bindings/clock/clock-bindings.txt
>> +
>> +Example:
>> +
>> +sram: sram at 50000000 {
>> +	compatible = "arm,juno-sram-ns", "mmio-sram";
>
> Nit: undocumented compatible string.
>
>> +	reg = <0x0 0x50000000 0x0 0x10000>;
>> +
>> +	#address-cells = <1>;
>> +	#size-cells = <1>;
>> +	ranges = <0 0x0 0x50000000 0x10000>;
>> +
>> +	cpu_scp_lpri: scp-shmem at 0 {
>> +		compatible = "arm,juno-scp-shmem";
>
> Nit: undocumented compatible string.
>

Will fix all the above compatible strings and add missing document.

[...]

>> +cpu at 0 {
>> +	...
>> +	reg = <0 0>;
>> +	clocks = <&scpi_dvfs 0>;
>> +	clock-names = "big";
>
> This isn't documented anywhere, and I can't see any code in this series
> using it. The Juno dts doesn't seem to have CPU clocks, and nothing in
> this series adds them -- have I missed a series?
>

No, you didn't miss anything, I didn't post the DTS changes to avoid
churn when the binding is still under discussion, will post in the next
version.

> Also, I'm not keen on using software-defined names (e.g. "big") for CPU
> clocks, as it doesn't stictly relate to the hardware. That said, the set
> of clocks, their names, and how they related to CPUs is
> implementation-specific, which is somewhat painful. I'm not sure how
> we cater for that with generic software.
>

Sorry for the confusion, it should be "vbig" in the above example. These
names are derived from the SCPI document and yes it may vary with platforms.

Regards,
Sudeep

[PATCH v3 2/5] firmware: add support for ARM System Control and Power Interface(SCPI) protocol

[Sudeep Holla]

This patch adds support for System Control and Power Interface (SCPI)
Message Protocol used between the Application Cores(AP) and the System
Control Processor(SCP). The MHU peripheral provides a mechanism for
inter-processor communication between SCP's M3 processor and AP.

SCP offers control and management of the core/cluster power states,
various power domain DVFS including the core/cluster, certain system
clocks configuration, thermal sensors and many others.

This protocol driver provides interface for all the client drivers using
SCPI to make use of the features offered by the SCP.

Cc: Jassi Brar <jassisinghbrar@gmail.com>
Cc: Liviu Dudau <Liviu.Dudau@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Jon Medhurst (Tixy) <tixy@linaro.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/Kconfig      |  19 ++
 drivers/firmware/Makefile     |   1 +
 drivers/firmware/arm_scpi.c   | 709 ++++++++++++++++++++++++++++++++++++++++++
 include/linux/scpi_protocol.h |  62 ++++
 4 files changed, 791 insertions(+)
 create mode 100644 drivers/firmware/arm_scpi.c
 create mode 100644 include/linux/scpi_protocol.h

diff --git a/drivers/firmware/Kconfig b/drivers/firmware/Kconfig
index 6517132e5d8b..855092d73266 100644
--- a/drivers/firmware/Kconfig
+++ b/drivers/firmware/Kconfig
@@ -136,6 +136,25 @@ config QCOM_SCM
 	bool
 	depends on ARM || ARM64
 
+config ARM_SCPI_PROTOCOL
+	tristate "ARM System Control and Power Interface (SCPI) Message Protocol"
+	depends on ARM_MHU
+	help
+	  System Control and Power Interface (SCPI) Message Protocol is
+	  defined for the purpose of communication between the Application
+	  Cores(AP) and the System Control Processor(SCP). The MHU peripheral
+	  provides a mechanism for inter-processor communication between SCP
+	  and AP.
+
+	  SCP controls most of the power managament on the Application
+	  Processors. It offers control and management of: the core/cluster
+	  power states, various power domain DVFS including the core/cluster,
+	  certain system clocks configuration, thermal sensors and many
+	  others.
+
+	  This protocol library provides interface for all the client drivers
+	  making use of the features offered by the SCP.
+
 source "drivers/firmware/google/Kconfig"
 source "drivers/firmware/efi/Kconfig"
 
diff --git a/drivers/firmware/Makefile b/drivers/firmware/Makefile
index 3fdd3912709a..a91e890423de 100644
--- a/drivers/firmware/Makefile
+++ b/drivers/firmware/Makefile
@@ -1,6 +1,7 @@
 #
 # Makefile for the linux kernel.
 #
+obj-$(CONFIG_ARM_SCPI_PROTOCOL)	+= arm_scpi.o
 obj-$(CONFIG_DMI)		+= dmi_scan.o
 obj-$(CONFIG_DMI_SYSFS)		+= dmi-sysfs.o
 obj-$(CONFIG_EDD)		+= edd.o
diff --git a/drivers/firmware/arm_scpi.c b/drivers/firmware/arm_scpi.c
new file mode 100644
index 000000000000..420a1b57d867
--- /dev/null
+++ b/drivers/firmware/arm_scpi.c
@@ -0,0 +1,709 @@
+/*
+ * System Control and Power Interface (SCPI) Message Protocol driver
+ *
+ * SCPI Message Protocol is used between the System Control Processor(SCP)
+ * and the Application Processors(AP). The Message Handling Unit(MHU)
+ * provides a mechanism for inter-processor communication between SCP's
+ * Cortex M3 and AP.
+ *
+ * SCP offers control and management of the core/cluster power states,
+ * various power domain DVFS including the core/cluster, certain system
+ * clocks configuration, thermal sensors and many others.
+ *
+ * Copyright (C) 2015 ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/bitmap.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/export.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/list.h>
+#include <linux/mailbox_client.h>
+#include <linux/module.h>
+#include <linux/of_address.h>
+#include <linux/of_platform.h>
+#include <linux/printk.h>
+#include <linux/scpi_protocol.h>
+#include <linux/slab.h>
+#include <linux/sort.h>
+#include <linux/spinlock.h>
+
+#define CMD_ID_SHIFT		0
+#define CMD_ID_MASK		0x7f
+#define CMD_TOKEN_ID_SHIFT	8
+#define CMD_TOKEN_ID_MASK	0xff
+#define CMD_DATA_SIZE_SHIFT	16
+#define CMD_DATA_SIZE_MASK	0x1ff
+#define PACK_SCPI_CMD(cmd_id, tx_sz)			\
+	((((cmd_id) & CMD_ID_MASK) << CMD_ID_SHIFT) |	\
+	(((tx_sz) & CMD_DATA_SIZE_MASK) << CMD_DATA_SIZE_SHIFT))
+#define ADD_SCPI_TOKEN(cmd, token)			\
+	((cmd) |= (((token) & CMD_TOKEN_ID_MASK) << CMD_TOKEN_ID_SHIFT))
+
+#define CMD_SIZE(cmd)	(((cmd) >> CMD_DATA_SIZE_SHIFT) & CMD_DATA_SIZE_MASK)
+#define CMD_UNIQ_MASK	(CMD_TOKEN_ID_MASK << CMD_TOKEN_ID_SHIFT | CMD_ID_MASK)
+#define CMD_XTRACT_UNIQ(cmd)	((cmd) & CMD_UNIQ_MASK)
+
+#define SCPI_SLOT		0
+
+#define MAX_DVFS_DOMAINS	8
+#define MAX_DVFS_OPPS		8
+#define DVFS_LATENCY(hdr)	(le32_to_cpu(hdr) >> 16)
+#define DVFS_OPP_COUNT(hdr)	((le32_to_cpu(hdr) >> 8) & 0xff)
+
+#define PROTOCOL_REV_MINOR_BITS	16
+#define PROTOCOL_REV_MINOR_MASK	((1U << PROTOCOL_REV_MINOR_BITS) - 1)
+#define PROTOCOL_REV_MAJOR(x)	((x) >> PROTOCOL_REV_MINOR_BITS)
+#define PROTOCOL_REV_MINOR(x)	((x) & PROTOCOL_REV_MINOR_MASK)
+
+#define FW_REV_MAJOR_BITS	24
+#define FW_REV_MINOR_BITS	16
+#define FW_REV_PATCH_MASK	((1U << FW_REV_MINOR_BITS) - 1)
+#define FW_REV_MINOR_MASK	((1U << FW_REV_MAJOR_BITS) - 1)
+#define FW_REV_MAJOR(x)		((x) >> FW_REV_MAJOR_BITS)
+#define FW_REV_MINOR(x)		(((x) & FW_REV_MINOR_MASK) >> FW_REV_MINOR_BITS)
+#define FW_REV_PATCH(x)		((x) & FW_REV_PATCH_MASK)
+
+#define MAX_RX_TIMEOUT		(msecs_to_jiffies(30))
+
+enum scpi_error_codes {
+	SCPI_SUCCESS = 0, /* Success */
+	SCPI_ERR_PARAM = 1, /* Invalid parameter(s) */
+	SCPI_ERR_ALIGN = 2, /* Invalid alignment */
+	SCPI_ERR_SIZE = 3, /* Invalid size */
+	SCPI_ERR_HANDLER = 4, /* Invalid handler/callback */
+	SCPI_ERR_ACCESS = 5, /* Invalid access/permission denied */
+	SCPI_ERR_RANGE = 6, /* Value out of range */
+	SCPI_ERR_TIMEOUT = 7, /* Timeout has occurred */
+	SCPI_ERR_NOMEM = 8, /* Invalid memory area or pointer */
+	SCPI_ERR_PWRSTATE = 9, /* Invalid power state */
+	SCPI_ERR_SUPPORT = 10, /* Not supported or disabled */
+	SCPI_ERR_DEVICE = 11, /* Device error */
+	SCPI_ERR_BUSY = 12, /* Device busy */
+	SCPI_ERR_MAX
+};
+
+enum scpi_std_cmd {
+	SCPI_CMD_INVALID		= 0x00,
+	SCPI_CMD_SCPI_READY		= 0x01,
+	SCPI_CMD_SCPI_CAPABILITIES	= 0x02,
+	SCPI_CMD_SET_CSS_PWR_STATE	= 0x03,
+	SCPI_CMD_GET_CSS_PWR_STATE	= 0x04,
+	SCPI_CMD_SET_SYS_PWR_STATE	= 0x05,
+	SCPI_CMD_SET_CPU_TIMER		= 0x06,
+	SCPI_CMD_CANCEL_CPU_TIMER	= 0x07,
+	SCPI_CMD_DVFS_CAPABILITIES	= 0x08,
+	SCPI_CMD_GET_DVFS_INFO		= 0x09,
+	SCPI_CMD_SET_DVFS		= 0x0a,
+	SCPI_CMD_GET_DVFS		= 0x0b,
+	SCPI_CMD_GET_DVFS_STAT		= 0x0c,
+	SCPI_CMD_CLOCK_CAPABILITIES	= 0x0d,
+	SCPI_CMD_GET_CLOCK_INFO		= 0x0e,
+	SCPI_CMD_SET_CLOCK_VALUE	= 0x0f,
+	SCPI_CMD_GET_CLOCK_VALUE	= 0x10,
+	SCPI_CMD_PSU_CAPABILITIES	= 0x11,
+	SCPI_CMD_GET_PSU_INFO		= 0x12,
+	SCPI_CMD_SET_PSU		= 0x13,
+	SCPI_CMD_GET_PSU		= 0x14,
+	SCPI_CMD_SENSOR_CAPABILITIES	= 0x15,
+	SCPI_CMD_SENSOR_INFO		= 0x16,
+	SCPI_CMD_SENSOR_VALUE		= 0x17,
+	SCPI_CMD_SENSOR_CFG_PERIODIC	= 0x18,
+	SCPI_CMD_SENSOR_CFG_BOUNDS	= 0x19,
+	SCPI_CMD_SENSOR_ASYNC_VALUE	= 0x1a,
+	SCPI_CMD_SET_DEVICE_PWR_STATE	= 0x1b,
+	SCPI_CMD_GET_DEVICE_PWR_STATE	= 0x1c,
+	SCPI_CMD_COUNT
+};
+
+struct scpi_xfer {
+	u32 slot; /* has to be first element */
+	u32 cmd;
+	u32 status;
+	const void *tx_buf;
+	void *rx_buf;
+	unsigned int tx_len;
+	unsigned int rx_len;
+	struct list_head node;
+	struct completion done;
+};
+
+struct scpi_chan {
+	struct mbox_client cl;
+	struct mbox_chan *chan;
+	void __iomem *tx_payload;
+	void __iomem *rx_payload;
+	struct list_head rx_pending;
+	struct list_head xfers_list;
+	struct scpi_xfer *xfers;
+	spinlock_t rx_lock; /* locking for the rx pending list */
+	struct mutex xfers_lock;
+	u8 token;
+};
+
+struct scpi_drvinfo {
+	u32 protocol_version;
+	u32 firmware_version;
+	int num_chans;
+	atomic_t next_chan;
+	struct scpi_ops *scpi_ops;
+	struct scpi_chan *channels;
+	struct scpi_dvfs_info *dvfs[MAX_DVFS_DOMAINS];
+};
+
+/*
+ * The SCP firmware only executes in little-endian mode, so any buffers
+ * shared through SCPI should have their contents converted to little-endian
+ */
+struct scpi_shared_mem {
+	__le32 command;
+	__le32 status;
+	u8 payload[0];
+} __packed;
+
+struct scp_capabilities {
+	__le32 protocol_version;
+	__le32 event_version;
+	__le32 platform_version;
+	__le32 commands[4];
+} __packed;
+
+struct clk_get_info {
+	__le16 id;
+	__le16 flags;
+	__le32 min_rate;
+	__le32 max_rate;
+	u8 name[20];
+} __packed;
+
+struct clk_get_value {
+	__le32 rate;
+} __packed;
+
+struct clk_set_value {
+	__le16 id;
+	__le16 reserved;
+	__le32 rate;
+} __packed;
+
+struct dvfs_info {
+	__le32 header;
+	struct {
+		__le32 freq;
+		__le32 m_volt;
+	} opps[MAX_DVFS_OPPS];
+} __packed;
+
+struct dvfs_get {
+	u8 index;
+} __packed;
+
+struct dvfs_set {
+	u8 domain;
+	u8 index;
+} __packed;
+
+static struct scpi_drvinfo *scpi_info;
+
+static int scpi_linux_errmap[SCPI_ERR_MAX] = {
+	/* better than switch case as long as return value is continuous */
+	0, /* SCPI_SUCCESS */
+	-EINVAL, /* SCPI_ERR_PARAM */
+	-ENOEXEC, /* SCPI_ERR_ALIGN */
+	-EMSGSIZE, /* SCPI_ERR_SIZE */
+	-EINVAL, /* SCPI_ERR_HANDLER */
+	-EACCES, /* SCPI_ERR_ACCESS */
+	-ERANGE, /* SCPI_ERR_RANGE */
+	-ETIMEDOUT, /* SCPI_ERR_TIMEOUT */
+	-ENOMEM, /* SCPI_ERR_NOMEM */
+	-EINVAL, /* SCPI_ERR_PWRSTATE */
+	-EOPNOTSUPP, /* SCPI_ERR_SUPPORT */
+	-EIO, /* SCPI_ERR_DEVICE */
+	-EBUSY, /* SCPI_ERR_BUSY */
+};
+
+static inline int scpi_to_linux_errno(int errno)
+{
+	if (errno >= SCPI_SUCCESS && errno < SCPI_ERR_MAX)
+		return scpi_linux_errmap[errno];
+	return -EIO;
+}
+
+static void scpi_process_cmd(struct scpi_chan *ch, u32 cmd)
+{
+	unsigned long flags;
+	struct scpi_xfer *t, *match = NULL;
+
+	spin_lock_irqsave(&ch->rx_lock, flags);
+	if (list_empty(&ch->rx_pending)) {
+		spin_unlock_irqrestore(&ch->rx_lock, flags);
+		return;
+	}
+
+	list_for_each_entry(t, &ch->rx_pending, node)
+		if (CMD_XTRACT_UNIQ(t->cmd) == CMD_XTRACT_UNIQ(cmd)) {
+			list_del(&t->node);
+			match = t;
+			break;
+		}
+	/* check if wait_for_completion is in progress or timed-out */
+	if (match && !completion_done(&match->done)) {
+		struct scpi_shared_mem *mem = ch->rx_payload;
+		unsigned int len = min(match->rx_len, CMD_SIZE(cmd));
+
+		match->status = le32_to_cpu(mem->status);
+		memcpy_fromio(match->rx_buf, mem->payload, len);
+		if (match->rx_len > len)
+			memset(match->rx_buf + len, 0, match->rx_len - len);
+		complete(&match->done);
+	}
+	spin_unlock_irqrestore(&ch->rx_lock, flags);
+}
+
+static void scpi_handle_remote_msg(struct mbox_client *c, void *msg)
+{
+	struct scpi_chan *ch = container_of(c, struct scpi_chan, cl);
+	struct scpi_shared_mem *mem = ch->rx_payload;
+	u32 cmd = le32_to_cpu(mem->command);
+
+	scpi_process_cmd(ch, cmd);
+}
+
+static void scpi_tx_prepare(struct mbox_client *c, void *msg)
+{
+	unsigned long flags;
+	struct scpi_xfer *t = msg;
+	struct scpi_chan *ch = container_of(c, struct scpi_chan, cl);
+	struct scpi_shared_mem *mem = (struct scpi_shared_mem *)ch->tx_payload;
+
+	if (t->tx_buf)
+		memcpy_toio(mem->payload, t->tx_buf, t->tx_len);
+	if (t->rx_buf) {
+		if (!(++ch->token))
+			++ch->token;
+		ADD_SCPI_TOKEN(t->cmd, ch->token);
+		spin_lock_irqsave(&ch->rx_lock, flags);
+		list_add_tail(&t->node, &ch->rx_pending);
+		spin_unlock_irqrestore(&ch->rx_lock, flags);
+	}
+	mem->command = cpu_to_le32(t->cmd);
+}
+
+static struct scpi_xfer *get_scpi_xfer(struct scpi_chan *ch)
+{
+	struct scpi_xfer *t;
+
+	mutex_lock(&ch->xfers_lock);
+	if (list_empty(&ch->xfers_list)) {
+		mutex_unlock(&ch->xfers_lock);
+		return NULL;
+	}
+	t = list_first_entry(&ch->xfers_list, struct scpi_xfer, node);
+	list_del(&t->node);
+	mutex_unlock(&ch->xfers_lock);
+	return t;
+}
+
+static void put_scpi_xfer(struct scpi_xfer *t, struct scpi_chan *ch)
+{
+	mutex_lock(&ch->xfers_lock);
+	list_add_tail(&t->node, &ch->xfers_list);
+	mutex_unlock(&ch->xfers_lock);
+}
+
+static int scpi_send_message(u8 cmd, void *tx_buf, unsigned int tx_len,
+			     void *rx_buf, unsigned int rx_len)
+{
+	int ret;
+	u8 chan;
+	struct scpi_xfer *msg;
+	struct scpi_chan *scpi_chan;
+
+	chan = atomic_inc_return(&scpi_info->next_chan) % scpi_info->num_chans;
+	scpi_chan = scpi_info->channels + chan;
+
+	msg = get_scpi_xfer(scpi_chan);
+	if (!msg)
+		return -ENOMEM;
+
+	msg->slot = BIT(SCPI_SLOT);
+	msg->cmd = PACK_SCPI_CMD(cmd, tx_len);
+	msg->tx_buf = tx_buf;
+	msg->tx_len = tx_len;
+	msg->rx_buf = rx_buf;
+	msg->rx_len = rx_len;
+	init_completion(&msg->done);
+
+	ret = mbox_send_message(scpi_chan->chan, msg);
+	if (ret < 0 || !rx_buf)
+		goto out;
+
+	if (!wait_for_completion_timeout(&msg->done, MAX_RX_TIMEOUT))
+		ret = -ETIMEDOUT;
+	else
+		/* first status word */
+		ret = le32_to_cpu(msg->status);
+out:
+	if (ret < 0 && rx_buf) /* remove entry from the list if timed-out */
+		scpi_process_cmd(scpi_chan, msg->cmd);
+
+	put_scpi_xfer(msg, scpi_chan);
+	/* SCPI error codes > 0, translate them to Linux scale*/
+	return ret > 0 ? scpi_to_linux_errno(ret) : ret;
+}
+
+static u32 scpi_get_version(void)
+{
+	return scpi_info->protocol_version;
+}
+
+static int
+scpi_clk_get_range(u16 clk_id, unsigned long *min, unsigned long *max)
+{
+	int ret;
+	struct clk_get_info clk;
+	__le16 le_clk_id = cpu_to_le16(clk_id);
+
+	ret = scpi_send_message(SCPI_CMD_GET_CLOCK_INFO, &le_clk_id,
+				sizeof(le_clk_id), &clk, sizeof(clk));
+	if (!ret) {
+		*min = le32_to_cpu(clk.min_rate);
+		*max = le32_to_cpu(clk.max_rate);
+	}
+	return ret;
+}
+
+static unsigned long scpi_clk_get_val(u16 clk_id)
+{
+	int ret;
+	struct clk_get_value clk;
+	__le16 le_clk_id = cpu_to_le16(clk_id);
+
+	ret = scpi_send_message(SCPI_CMD_GET_CLOCK_VALUE, &le_clk_id,
+				sizeof(le_clk_id), &clk, sizeof(clk));
+	return ret ? ret : le32_to_cpu(clk.rate);
+}
+
+static int scpi_clk_set_val(u16 clk_id, unsigned long rate)
+{
+	int stat;
+	struct clk_set_value clk = {
+		.id = cpu_to_le16(clk_id),
+		.rate = cpu_to_le32(rate)
+	};
+
+	return scpi_send_message(SCPI_CMD_SET_CLOCK_VALUE, &clk, sizeof(clk),
+				 &stat, sizeof(stat));
+}
+
+static int scpi_dvfs_get_idx(u8 domain)
+{
+	int ret;
+	struct dvfs_get dvfs;
+
+	ret = scpi_send_message(SCPI_CMD_GET_DVFS, &domain, sizeof(domain),
+				&dvfs, sizeof(dvfs));
+	return ret ? ret : dvfs.index;
+}
+
+static int scpi_dvfs_set_idx(u8 domain, u8 index)
+{
+	int stat;
+	struct dvfs_set dvfs = {domain, index};
+
+	return scpi_send_message(SCPI_CMD_SET_DVFS, &dvfs, sizeof(dvfs),
+				 &stat, sizeof(stat));
+}
+
+static int opp_cmp_func(const void *opp1, const void *opp2)
+{
+	const struct scpi_opp *t1 = opp1, *t2 = opp2;
+
+	return t1->freq - t2->freq;
+}
+
+static struct scpi_dvfs_info *scpi_dvfs_get_info(u8 domain)
+{
+	struct scpi_dvfs_info *info;
+	struct scpi_opp *opp;
+	struct dvfs_info buf;
+	int ret, i;
+
+	if (domain >= MAX_DVFS_DOMAINS)
+		return ERR_PTR(-EINVAL);
+
+	if (scpi_info->dvfs[domain])	/* data already populated */
+		return scpi_info->dvfs[domain];
+
+	ret = scpi_send_message(SCPI_CMD_GET_DVFS_INFO, &domain, sizeof(domain),
+				&buf, sizeof(buf));
+
+	if (ret)
+		return ERR_PTR(ret);
+
+	info = kmalloc(sizeof(*info), GFP_KERNEL);
+	if (!info)
+		return ERR_PTR(-ENOMEM);
+
+	info->count = DVFS_OPP_COUNT(buf.header);
+	info->latency = DVFS_LATENCY(buf.header) * 1000; /* uS to nS */
+
+	info->opps = kcalloc(info->count, sizeof(*opp), GFP_KERNEL);
+	if (!info->opps) {
+		kfree(info);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	for (i = 0, opp = info->opps; i < info->count; i++, opp++) {
+		opp->freq = le32_to_cpu(buf.opps[i].freq);
+		opp->m_volt = le32_to_cpu(buf.opps[i].m_volt);
+	}
+
+	sort(info->opps, info->count, sizeof(*opp), opp_cmp_func, NULL);
+
+	scpi_info->dvfs[domain] = info;
+	return info;
+}
+
+static struct scpi_ops scpi_ops = {
+	.get_version = scpi_get_version,
+	.clk_get_range = scpi_clk_get_range,
+	.clk_get_val = scpi_clk_get_val,
+	.clk_set_val = scpi_clk_set_val,
+	.dvfs_get_idx = scpi_dvfs_get_idx,
+	.dvfs_set_idx = scpi_dvfs_set_idx,
+	.dvfs_get_info = scpi_dvfs_get_info,
+};
+
+struct scpi_ops *get_scpi_ops(void)
+{
+	return scpi_info ? scpi_info->scpi_ops : NULL;
+}
+EXPORT_SYMBOL_GPL(get_scpi_ops);
+
+static int scpi_init_versions(struct scpi_drvinfo *info)
+{
+	int ret;
+	struct scp_capabilities caps;
+
+	ret = scpi_send_message(SCPI_CMD_SCPI_CAPABILITIES, NULL, 0,
+				&caps, sizeof(caps));
+	if (!ret) {
+		info->protocol_version = le32_to_cpu(caps.protocol_version);
+		info->firmware_version = le32_to_cpu(caps.platform_version);
+	}
+	return ret;
+}
+
+static ssize_t protocol_version_show(struct device *dev,
+				     struct device_attribute *attr, char *buf)
+{
+	struct scpi_drvinfo *scpi_info = dev_get_drvdata(dev);
+
+	return sprintf(buf, "%d.%d\n",
+		       PROTOCOL_REV_MAJOR(scpi_info->protocol_version),
+		       PROTOCOL_REV_MINOR(scpi_info->protocol_version));
+}
+static DEVICE_ATTR_RO(protocol_version);
+
+static ssize_t firmware_version_show(struct device *dev,
+				     struct device_attribute *attr, char *buf)
+{
+	struct scpi_drvinfo *scpi_info = dev_get_drvdata(dev);
+
+	return sprintf(buf, "%d.%d.%d\n",
+		       FW_REV_MAJOR(scpi_info->firmware_version),
+		       FW_REV_MINOR(scpi_info->firmware_version),
+		       FW_REV_PATCH(scpi_info->firmware_version));
+}
+static DEVICE_ATTR_RO(firmware_version);
+
+static struct attribute *versions_attrs[] = {
+	&dev_attr_firmware_version.attr,
+	&dev_attr_protocol_version.attr,
+	NULL,
+};
+ATTRIBUTE_GROUPS(versions);
+
+static void
+scpi_free_channels(struct device *dev, struct scpi_chan *pchan, int count)
+{
+	int i;
+
+	for (i = 0; i < count && pchan->chan; i++, pchan++) {
+		mbox_free_channel(pchan->chan);
+		devm_kfree(dev, pchan->xfers);
+		devm_iounmap(dev, pchan->rx_payload);
+	}
+}
+
+static int scpi_remove(struct platform_device *pdev)
+{
+	int i;
+	struct device *dev = &pdev->dev;
+	struct scpi_drvinfo *info = platform_get_drvdata(pdev);
+
+	scpi_info = NULL; /* stop exporting SCPI ops through get_scpi_ops */
+
+	of_platform_depopulate(dev);
+	sysfs_remove_groups(&dev->kobj, versions_groups);
+	scpi_free_channels(dev, info->channels, info->num_chans);
+	platform_set_drvdata(pdev, NULL);
+
+	for (i = 0; i < MAX_DVFS_DOMAINS && info->dvfs[i]; i++) {
+		kfree(info->dvfs[i]->opps);
+		kfree(info->dvfs[i]);
+	}
+	devm_kfree(dev, info->channels);
+	devm_kfree(dev, info);
+
+	return 0;
+}
+
+#define MAX_SCPI_XFERS		10
+static int scpi_alloc_xfer_list(struct device *dev, struct scpi_chan *ch)
+{
+	int i;
+	struct scpi_xfer *xfers;
+
+	xfers = devm_kzalloc(dev, MAX_SCPI_XFERS * sizeof(*xfers), GFP_KERNEL);
+	if (!xfers)
+		return -ENOMEM;
+
+	ch->xfers = xfers;
+	for (i = 0; i < MAX_SCPI_XFERS; i++, xfers++)
+		list_add_tail(&xfers->node, &ch->xfers_list);
+	return 0;
+}
+
+static int scpi_probe(struct platform_device *pdev)
+{
+	int count, idx, ret;
+	struct resource res;
+	struct scpi_chan *scpi_chan;
+	struct device *dev = &pdev->dev;
+	struct device_node *np = dev->of_node;
+
+	scpi_info = devm_kzalloc(dev, sizeof(*scpi_info), GFP_KERNEL);
+	if (!scpi_info)
+		return -ENOMEM;
+
+	count = of_count_phandle_with_args(np, "mboxes", "#mbox-cells");
+	if (count < 0) {
+		dev_err(dev, "no mboxes property in '%s'\n", np->full_name);
+		return -ENODEV;
+	}
+
+	scpi_chan = devm_kcalloc(dev, count, sizeof(*scpi_chan), GFP_KERNEL);
+	if (!scpi_chan)
+		return -ENOMEM;
+
+	for (idx = 0; idx < count; idx++) {
+		resource_size_t size;
+		struct scpi_chan *pchan = scpi_chan + idx;
+		struct mbox_client *cl = &pchan->cl;
+		struct device_node *shmem = of_parse_phandle(np, "shmem", idx);
+
+		if (of_address_to_resource(shmem, 0, &res)) {
+			dev_err(dev, "failed to get SCPI payload mem resource\n");
+			ret = -EINVAL;
+			goto err;
+		}
+
+		size = resource_size(&res);
+		pchan->rx_payload = devm_ioremap(dev, res.start, size);
+		if (!pchan->rx_payload) {
+			dev_err(dev, "failed to ioremap SCPI payload\n");
+			ret = -EADDRNOTAVAIL;
+			goto err;
+		}
+		pchan->tx_payload = pchan->rx_payload + (size >> 1);
+
+		cl->dev = dev;
+		cl->rx_callback = scpi_handle_remote_msg;
+		cl->tx_prepare = scpi_tx_prepare;
+		cl->tx_block = true;
+		cl->tx_tout = 50;
+		cl->knows_txdone = false; /* controller can ack */
+
+		INIT_LIST_HEAD(&pchan->rx_pending);
+		INIT_LIST_HEAD(&pchan->xfers_list);
+		spin_lock_init(&pchan->rx_lock);
+		mutex_init(&pchan->xfers_lock);
+
+		ret = scpi_alloc_xfer_list(dev, pchan);
+		if (!ret) {
+			pchan->chan = mbox_request_channel(cl, idx);
+			if (!IS_ERR(pchan->chan))
+				continue;
+			ret = -EPROBE_DEFER;
+			dev_err(dev, "failed to acquire channel#%d\n", idx);
+		}
+err:
+		scpi_free_channels(dev, scpi_chan, idx);
+		scpi_info = NULL;
+		return ret;
+	}
+
+	scpi_info->channels = scpi_chan;
+	scpi_info->num_chans = count;
+	platform_set_drvdata(pdev, scpi_info);
+
+	ret = scpi_init_versions(scpi_info);
+	if (ret) {
+		dev_err(dev, "incorrect or no SCP firmware found\n");
+		scpi_remove(pdev);
+		return ret;
+	}
+
+	_dev_info(dev, "SCP Protocol %d.%d Firmware %d.%d.%d version\n",
+		  PROTOCOL_REV_MAJOR(scpi_info->protocol_version),
+		  PROTOCOL_REV_MINOR(scpi_info->protocol_version),
+		  FW_REV_MAJOR(scpi_info->firmware_version),
+		  FW_REV_MINOR(scpi_info->firmware_version),
+		  FW_REV_PATCH(scpi_info->firmware_version));
+	scpi_info->scpi_ops = &scpi_ops;
+
+	ret = sysfs_create_groups(&dev->kobj, versions_groups);
+	if (ret)
+		dev_err(dev, "unable to create sysfs version group\n");
+
+	return of_platform_populate(dev->of_node, NULL, NULL, dev);
+}
+
+static const struct of_device_id scpi_of_match[] = {
+	{.compatible = "arm,scpi"},
+	{},
+};
+
+MODULE_DEVICE_TABLE(of, scpi_of_match);
+
+static struct platform_driver scpi_driver = {
+	.driver = {
+		.name = "scpi_protocol",
+		.of_match_table = scpi_of_match,
+	},
+	.probe = scpi_probe,
+	.remove = scpi_remove,
+};
+module_platform_driver(scpi_driver);
+
+MODULE_AUTHOR("Sudeep Holla <sudeep.holla@arm.com>");
+MODULE_DESCRIPTION("ARM SCPI mailbox protocol driver");
+MODULE_LICENSE("GPL v2");
diff --git a/include/linux/scpi_protocol.h b/include/linux/scpi_protocol.h
new file mode 100644
index 000000000000..da8631e59b79
--- /dev/null
+++ b/include/linux/scpi_protocol.h
@@ -0,0 +1,62 @@
+/*
+ * SCPI Message Protocol driver header
+ *
+ * Copyright (C) 2014 ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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, see <http://www.gnu.org/licenses/>.
+ */
+#include <linux/types.h>
+
+struct scpi_opp {
+	u32 freq;
+	u32 m_volt;
+} __packed;
+
+struct scpi_dvfs_info {
+	unsigned int count;
+	unsigned int latency; /* in nanoseconds */
+	struct scpi_opp *opps;
+};
+
+/**
+ * struct scpi_ops - represents the various operations provided
+ *	by SCP through SCPI message protocol
+ * @get_version: returns the major and minor revision on the SCPI
+ *	message protocol
+ * @clk_get_range: gets clock range limit(min - max in Hz)
+ * @clk_get_val: gets clock value(in Hz)
+ * @clk_set_val: sets the clock value, setting to 0 will disable the
+ *	clock (if supported)
+ * @dvfs_get_idx: gets the Operating Point of the given power domain.
+ *	OPP is an index to the list return by @dvfs_get_info
+ * @dvfs_set_idx: sets the Operating Point of the given power domain.
+ *	OPP is an index to the list return by @dvfs_get_info
+ * @dvfs_get_info: returns the DVFS capabilities of the given power
+ *	domain. It includes the OPP list and the latency information
+ */
+struct scpi_ops {
+	u32 (*get_version)(void);
+	int (*clk_get_range)(u16, unsigned long *, unsigned long *);
+	unsigned long (*clk_get_val)(u16);
+	int (*clk_set_val)(u16, unsigned long);
+	int (*dvfs_get_idx)(u8);
+	int (*dvfs_set_idx)(u8, u8);
+	struct scpi_dvfs_info *(*dvfs_get_info)(u8);
+};
+
+#if defined(CONFIG_ARM_SCPI_PROTOCOL) ||	\
+	defined(CONFIG_ARM_SCPI_PROTOCOL_MODULE)
+struct scpi_ops *get_scpi_ops(void);
+#else
+static inline struct scpi_ops *get_scpi_ops(void) { return NULL; }
+#endif
-- 
1.9.1

[PATCH v3 3/5] clk: add support for clocks provided by SCP(System Control Processor)

[Sudeep Holla]

On some ARM based systems, a separate Cortex-M based System Control
Processor(SCP) provides the overall power, clock, reset and system
control. System Control and Power Interface(SCPI) Message Protocol
is defined for the communication between the Application Cores(AP)
and the SCP.

This patch adds support for the clocks provided by SCP using SCPI
protocol.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Cc: Mike Turquette <mturquette@linaro.org>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Cc: Liviu Dudau <Liviu.Dudau@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
Cc: linux-clk at vger.kernel.org
---
 drivers/clk/Kconfig    |  10 ++
 drivers/clk/Makefile   |   1 +
 drivers/clk/clk-scpi.c | 287 +++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 298 insertions(+)
 create mode 100644 drivers/clk/clk-scpi.c

diff --git a/drivers/clk/Kconfig b/drivers/clk/Kconfig
index 9897f353bf1a..0fe8daefc105 100644
--- a/drivers/clk/Kconfig
+++ b/drivers/clk/Kconfig
@@ -59,6 +59,16 @@ config COMMON_CLK_RK808
 	  clocked at 32KHz each. Clkout1 is always on, Clkout2 can off
 	  by control register.
 
+config COMMON_CLK_SCPI
+        tristate "Clock driver controlled via SCPI interface"
+        depends on ARM_SCPI_PROTOCOL || COMPILE_TEST
+        ---help---
+          This driver provides support for clocks that are controlled
+          by firmware that implements the SCPI interface.
+
+	  This driver uses SCPI Message Protocol to interact with the
+	  firmware providing all the clock controls.
+
 config COMMON_CLK_SI5351
 	tristate "Clock driver for SiLabs 5351A/B/C"
 	depends on I2C
diff --git a/drivers/clk/Makefile b/drivers/clk/Makefile
index 3d00c25382c5..442ab6ebd5b1 100644
--- a/drivers/clk/Makefile
+++ b/drivers/clk/Makefile
@@ -36,6 +36,7 @@ obj-$(CONFIG_COMMON_CLK_PALMAS)		+= clk-palmas.o
 obj-$(CONFIG_CLK_QORIQ)			+= clk-qoriq.o
 obj-$(CONFIG_COMMON_CLK_RK808)		+= clk-rk808.o
 obj-$(CONFIG_COMMON_CLK_S2MPS11)	+= clk-s2mps11.o
+obj-$(CONFIG_COMMON_CLK_SCPI)           += clk-scpi.o
 obj-$(CONFIG_COMMON_CLK_SI5351)		+= clk-si5351.o
 obj-$(CONFIG_COMMON_CLK_SI570)		+= clk-si570.o
 obj-$(CONFIG_CLK_TWL6040)		+= clk-twl6040.o
diff --git a/drivers/clk/clk-scpi.c b/drivers/clk/clk-scpi.c
new file mode 100644
index 000000000000..55bfa708a5b4
--- /dev/null
+++ b/drivers/clk/clk-scpi.c
@@ -0,0 +1,287 @@
+/*
+ * System Control and Power Interface (SCPI) Protocol based clock driver
+ *
+ * Copyright (C) 2015 ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/clk-provider.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/of.h>
+#include <linux/module.h>
+#include <linux/of_platform.h>
+#include <linux/scpi_protocol.h>
+
+struct scpi_clk {
+	u32 id;
+	const char *name;
+	struct clk_hw hw;
+	struct scpi_dvfs_info *info;
+	unsigned long rate_min;
+	unsigned long rate_max;
+};
+
+#define to_scpi_clk(clk) container_of(clk, struct scpi_clk, hw)
+
+static struct scpi_ops *scpi_ops;
+
+static unsigned long scpi_clk_recalc_rate(struct clk_hw *hw,
+					  unsigned long parent_rate)
+{
+	struct scpi_clk *clk = to_scpi_clk(hw);
+
+	return scpi_ops->clk_get_val(clk->id);
+}
+
+static long scpi_clk_round_rate(struct clk_hw *hw, unsigned long rate,
+				unsigned long *parent_rate)
+{
+	struct scpi_clk *clk = to_scpi_clk(hw);
+
+	if (WARN_ON(clk->rate_min && rate < clk->rate_min))
+		rate = clk->rate_min;
+	if (WARN_ON(clk->rate_max && rate > clk->rate_max))
+		rate = clk->rate_max;
+
+	return rate;
+}
+
+static int scpi_clk_set_rate(struct clk_hw *hw, unsigned long rate,
+			     unsigned long parent_rate)
+{
+	struct scpi_clk *clk = to_scpi_clk(hw);
+
+	return scpi_ops->clk_set_val(clk->id, rate);
+}
+
+static const struct clk_ops scpi_clk_ops = {
+	.recalc_rate = scpi_clk_recalc_rate,
+	.round_rate = scpi_clk_round_rate,
+	.set_rate = scpi_clk_set_rate,
+};
+
+/* find closest match to given frequency in OPP table */
+static int __scpi_dvfs_round_rate(struct scpi_clk *clk, unsigned long rate)
+{
+	int idx;
+	u32 fmin = 0, fmax = ~0, ftmp;
+	struct scpi_opp *opp = clk->info->opps;
+
+	for (idx = 0; idx < clk->info->count; idx++, opp++) {
+		ftmp = opp->freq;
+		if (ftmp >= (u32)rate) {
+			if (ftmp <= fmax)
+				fmax = ftmp;
+			break;
+		} else if (ftmp >= fmin) {
+			fmin = ftmp;
+		}
+	}
+	return fmax != ~0 ? fmax : fmin;
+}
+
+static unsigned long scpi_dvfs_recalc_rate(struct clk_hw *hw,
+					   unsigned long parent_rate)
+{
+	struct scpi_clk *clk = to_scpi_clk(hw);
+	int idx = scpi_ops->dvfs_get_idx(clk->id);
+	struct scpi_opp *opp;
+
+	if (idx < 0)
+		return 0;
+
+	opp = clk->info->opps + idx;
+	return opp->freq;
+}
+
+static long scpi_dvfs_round_rate(struct clk_hw *hw, unsigned long rate,
+				 unsigned long *parent_rate)
+{
+	struct scpi_clk *clk = to_scpi_clk(hw);
+
+	return __scpi_dvfs_round_rate(clk, rate);
+}
+
+static int __scpi_find_dvfs_index(struct scpi_clk *clk, unsigned long rate)
+{
+	int idx, max_opp = clk->info->count;
+	struct scpi_opp *opp = clk->info->opps;
+
+	for (idx = 0; idx < max_opp; idx++, opp++)
+		if (opp->freq == rate)
+			return idx;
+	return -EINVAL;
+}
+
+static int scpi_dvfs_set_rate(struct clk_hw *hw, unsigned long rate,
+			      unsigned long parent_rate)
+{
+	struct scpi_clk *clk = to_scpi_clk(hw);
+	int ret = __scpi_find_dvfs_index(clk, rate);
+
+	if (ret < 0)
+		return ret;
+	return scpi_ops->dvfs_set_idx(clk->id, (u8)ret);
+}
+
+static const struct clk_ops scpi_dvfs_ops = {
+	.recalc_rate = scpi_dvfs_recalc_rate,
+	.round_rate = scpi_dvfs_round_rate,
+	.set_rate = scpi_dvfs_set_rate,
+};
+
+static const struct of_device_id scpi_clk_match[] = {
+	{ .compatible = "arm,scpi-dvfs", .data = &scpi_dvfs_ops, },
+	{ .compatible = "arm,scpi-clk", .data = &scpi_clk_ops, },
+	{}
+};
+
+static struct clk *
+scpi_clk_ops_init(struct device *dev, struct device_node *np,
+		  struct scpi_clk *sclk)
+{
+	struct clk_init_data init;
+	const struct of_device_id *match = of_match_node(scpi_clk_match, np);
+
+	init.name = sclk->name;
+	init.flags = CLK_IS_ROOT;
+	init.num_parents = 0;
+	init.ops = match->data;
+	sclk->hw.init = &init;
+
+	if (init.ops == &scpi_dvfs_ops) {
+		struct scpi_dvfs_info *info = scpi_ops->dvfs_get_info(sclk->id);
+
+		if (IS_ERR(info))
+			return NULL;
+		sclk->info = info;
+	} else if (init.ops == &scpi_clk_ops) {
+		int ret = scpi_ops->clk_get_range(sclk->id, &sclk->rate_min,
+						  &sclk->rate_max);
+		if (ret || !sclk->rate_max)
+			return NULL;
+	} else {
+		return NULL;
+	}
+
+	return devm_clk_register(dev, &sclk->hw);
+}
+
+static int scpi_clk_add(struct device *dev, struct device_node *np)
+{
+	struct clk **clks;
+	int idx, count;
+	struct clk_onecell_data *clk_data;
+
+	count = of_property_count_strings(np, "clock-output-names");
+	if (count < 0) {
+		dev_err(dev, "%s: invalid clock output count\n", np->name);
+		return -EINVAL;
+	}
+
+	clk_data = devm_kmalloc(dev, sizeof(*clk_data), GFP_KERNEL);
+	if (!clk_data)
+		return -ENOMEM;
+
+	clks = devm_kcalloc(dev, count, sizeof(*clks), GFP_KERNEL);
+	if (!clks)
+		return -ENOMEM;
+
+	for (idx = 0; idx < count; idx++) {
+		struct scpi_clk *sclk;
+		u32 val;
+
+		sclk = devm_kzalloc(dev, sizeof(*sclk), GFP_KERNEL);
+		if (!sclk)
+			return -ENOMEM;
+
+		if (of_property_read_string_index(np, "clock-output-names",
+						  idx, &sclk->name)) {
+			dev_err(dev, "invalid clock name @ %s\n", np->name);
+			return -EINVAL;
+		}
+
+		if (of_property_read_u32_index(np, "clock-indices",
+					       idx, &val)) {
+			dev_err(dev, "invalid clock index @ %s\n", np->name);
+			return -EINVAL;
+		}
+
+		sclk->id = val;
+
+		clks[idx] = scpi_clk_ops_init(dev, np, sclk);
+		if (IS_ERR_OR_NULL(clks[idx]))
+			dev_err(dev, "failed to register clock '%s'\n",
+				sclk->name);
+		else
+			dev_dbg(dev, "Registered clock '%s'\n", sclk->name);
+	}
+
+	clk_data->clks = clks;
+	clk_data->clk_num = idx;
+	of_clk_add_provider(np, of_clk_src_onecell_get, clk_data);
+
+	return 0;
+}
+
+static int scpi_clocks_remove(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *child, *np = dev->of_node;
+
+	for_each_available_child_of_node(np, child)
+		of_clk_del_provider(np);
+	scpi_ops = NULL;
+	return 0;
+}
+
+static int scpi_clocks_probe(struct platform_device *pdev)
+{
+	int ret;
+	struct device *dev = &pdev->dev;
+	struct device_node *child, *np = dev->of_node;
+
+	scpi_ops = get_scpi_ops();
+	if (!scpi_ops)
+		return -ENXIO;
+
+	for_each_available_child_of_node(np, child) {
+		ret = scpi_clk_add(dev, child);
+		if (ret) {
+			scpi_clocks_remove(pdev);
+			return ret;
+		}
+	}
+	return 0;
+}
+
+static const struct of_device_id scpi_clocks_ids[] = {
+	{ .compatible = "arm,scpi-clocks", },
+	{}
+};
+
+static struct platform_driver scpi_clocks_driver = {
+	.driver	= {
+		.name = "scpi_clocks",
+		.of_match_table = scpi_clocks_ids,
+	},
+	.probe = scpi_clocks_probe,
+	.remove = scpi_clocks_remove,
+};
+module_platform_driver(scpi_clocks_driver);
+
+MODULE_AUTHOR("Sudeep Holla <sudeep.holla@arm.com>");
+MODULE_DESCRIPTION("ARM SCPI clock driver");
+MODULE_LICENSE("GPL v2");
-- 
1.9.1

[PATCH v3 3/5] clk: add support for clocks provided by SCP(System Control Processor)

[Stephen Boyd]

On 05/27, Sudeep Holla wrote:
> +
> +#include <linux/clk-provider.h>
> +#include <linux/device.h>
> +#include <linux/err.h>
> +#include <linux/of.h>
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/scpi_protocol.h>
> +
> +struct scpi_clk {
> +	u32 id;
> +	const char *name;
> +	struct clk_hw hw;
> +	struct scpi_dvfs_info *info;
> +	unsigned long rate_min;
> +	unsigned long rate_max;
> +};
> +
> +#define to_scpi_clk(clk) container_of(clk, struct scpi_clk, hw)
> +
> +static struct scpi_ops *scpi_ops;

Why do we need this singleton? Can we put this pointer into scpi_clk?

> +
> +static unsigned long scpi_clk_recalc_rate(struct clk_hw *hw,
> +					  unsigned long parent_rate)
> +{
> +	struct scpi_clk *clk = to_scpi_clk(hw);
> +
> +	return scpi_ops->clk_get_val(clk->id);
> +}
> +
> +static long scpi_clk_round_rate(struct clk_hw *hw, unsigned long rate,
> +				unsigned long *parent_rate)
> +{
> +	struct scpi_clk *clk = to_scpi_clk(hw);
> +
> +	if (WARN_ON(clk->rate_min && rate < clk->rate_min))
> +		rate = clk->rate_min;
> +	if (WARN_ON(clk->rate_max && rate > clk->rate_max))
> +		rate = clk->rate_max;
> +
> +	return rate;
> +}

Hm.. this seems really generic. It might be better to support a
way to tell the framework to limit the min/max rate that's
accepted for a clk. That could be done later though.

> +
> +static int scpi_clk_set_rate(struct clk_hw *hw, unsigned long rate,
> +			     unsigned long parent_rate)
> +{
> +	struct scpi_clk *clk = to_scpi_clk(hw);
> +
> +	return scpi_ops->clk_set_val(clk->id, rate);
> +}
> +
[..]
> +
> +static int scpi_clk_add(struct device *dev, struct device_node *np)
> +{
> +	struct clk **clks;
> +	int idx, count;
> +	struct clk_onecell_data *clk_data;
> +
> +	count = of_property_count_strings(np, "clock-output-names");
> +	if (count < 0) {
> +		dev_err(dev, "%s: invalid clock output count\n", np->name);
> +		return -EINVAL;
> +	}
> +
> +	clk_data = devm_kmalloc(dev, sizeof(*clk_data), GFP_KERNEL);
> +	if (!clk_data)
> +		return -ENOMEM;
> +
> +	clks = devm_kcalloc(dev, count, sizeof(*clks), GFP_KERNEL);
> +	if (!clks)
> +		return -ENOMEM;
> +
> +	for (idx = 0; idx < count; idx++) {
> +		struct scpi_clk *sclk;
> +		u32 val;
> +
> +		sclk = devm_kzalloc(dev, sizeof(*sclk), GFP_KERNEL);
> +		if (!sclk)
> +			return -ENOMEM;
> +
> +		if (of_property_read_string_index(np, "clock-output-names",
> +						  idx, &sclk->name)) {
> +			dev_err(dev, "invalid clock name @ %s\n", np->name);
> +			return -EINVAL;
> +		}
> +
> +		if (of_property_read_u32_index(np, "clock-indices",
> +					       idx, &val)) {
> +			dev_err(dev, "invalid clock index @ %s\n", np->name);
> +			return -EINVAL;
> +		}
> +
> +		sclk->id = val;
> +
> +		clks[idx] = scpi_clk_ops_init(dev, np, sclk);
> +		if (IS_ERR_OR_NULL(clks[idx]))
> +			dev_err(dev, "failed to register clock '%s'\n",
> +				sclk->name);
> +		else
> +			dev_dbg(dev, "Registered clock '%s'\n", sclk->name);
> +	}
> +
> +	clk_data->clks = clks;
> +	clk_data->clk_num = idx;
> +	of_clk_add_provider(np, of_clk_src_onecell_get, clk_data);

And if of_clk_add_provider() fails?

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

[PATCH v3 3/5] clk: add support for clocks provided by SCP(System Control Processor)

[Sudeep Holla]

On 04/06/15 21:20, Stephen Boyd wrote:
> On 05/27, Sudeep Holla wrote:
>> +
>> +#include <linux/clk-provider.h>
>> +#include <linux/device.h>
>> +#include <linux/err.h>
>> +#include <linux/of.h>
>> +#include <linux/module.h>
>> +#include <linux/of_platform.h>
>> +#include <linux/scpi_protocol.h>
>> +
>> +struct scpi_clk {
>> +	u32 id;
>> +	const char *name;
>> +	struct clk_hw hw;
>> +	struct scpi_dvfs_info *info;
>> +	unsigned long rate_min;
>> +	unsigned long rate_max;
>> +};
>> +
>> +#define to_scpi_clk(clk) container_of(clk, struct scpi_clk, hw)
>> +
>> +static struct scpi_ops *scpi_ops;
>
> Why do we need this singleton? Can we put this pointer into scpi_clk?
>

Yes I will move it.

>> +
>> +static unsigned long scpi_clk_recalc_rate(struct clk_hw *hw,
>> +					  unsigned long parent_rate)
>> +{
>> +	struct scpi_clk *clk = to_scpi_clk(hw);
>> +
>> +	return scpi_ops->clk_get_val(clk->id);
>> +}
>> +
>> +static long scpi_clk_round_rate(struct clk_hw *hw, unsigned long rate,
>> +				unsigned long *parent_rate)
>> +{
>> +	struct scpi_clk *clk = to_scpi_clk(hw);
>> +
>> +	if (WARN_ON(clk->rate_min && rate < clk->rate_min))
>> +		rate = clk->rate_min;
>> +	if (WARN_ON(clk->rate_max && rate > clk->rate_max))
>> +		rate = clk->rate_max;
>> +
>> +	return rate;
>> +}
>
> Hm.. this seems really generic. It might be better to support a
> way to tell the framework to limit the min/max rate that's
> accepted for a clk. That could be done later though.
>

True, framework have some boundary checks in place. I will check if
I can use it with minimum changes to the core. If not, we can take this
up later as you suggested.

>> +
>> +static int scpi_clk_set_rate(struct clk_hw *hw, unsigned long rate,
>> +			     unsigned long parent_rate)
>> +{
>> +	struct scpi_clk *clk = to_scpi_clk(hw);
>> +
>> +	return scpi_ops->clk_set_val(clk->id, rate);
>> +}
>> +
> [..]
>> +
>> +static int scpi_clk_add(struct device *dev, struct device_node *np)
>> +{
>> +	struct clk **clks;
>> +	int idx, count;
>> +	struct clk_onecell_data *clk_data;
>> +
>> +	count = of_property_count_strings(np, "clock-output-names");
>> +	if (count < 0) {
>> +		dev_err(dev, "%s: invalid clock output count\n", np->name);
>> +		return -EINVAL;
>> +	}
>> +
>> +	clk_data = devm_kmalloc(dev, sizeof(*clk_data), GFP_KERNEL);
>> +	if (!clk_data)
>> +		return -ENOMEM;
>> +
>> +	clks = devm_kcalloc(dev, count, sizeof(*clks), GFP_KERNEL);
>> +	if (!clks)
>> +		return -ENOMEM;
>> +
>> +	for (idx = 0; idx < count; idx++) {
>> +		struct scpi_clk *sclk;
>> +		u32 val;
>> +
>> +		sclk = devm_kzalloc(dev, sizeof(*sclk), GFP_KERNEL);
>> +		if (!sclk)
>> +			return -ENOMEM;
>> +
>> +		if (of_property_read_string_index(np, "clock-output-names",
>> +						  idx, &sclk->name)) {
>> +			dev_err(dev, "invalid clock name @ %s\n", np->name);
>> +			return -EINVAL;
>> +		}
>> +
>> +		if (of_property_read_u32_index(np, "clock-indices",
>> +					       idx, &val)) {
>> +			dev_err(dev, "invalid clock index @ %s\n", np->name);
>> +			return -EINVAL;
>> +		}
>> +
>> +		sclk->id = val;
>> +
>> +		clks[idx] = scpi_clk_ops_init(dev, np, sclk);
>> +		if (IS_ERR_OR_NULL(clks[idx]))
>> +			dev_err(dev, "failed to register clock '%s'\n",
>> +				sclk->name);
>> +		else
>> +			dev_dbg(dev, "Registered clock '%s'\n", sclk->name);
>> +	}
>> +
>> +	clk_data->clks = clks;
>> +	clk_data->clk_num = idx;
>> +	of_clk_add_provider(np, of_clk_src_onecell_get, clk_data);
>
> And if of_clk_add_provider() fails?
>

Ah, my bad, will fix it.

Regards,
Sudeep

[PATCH v3 3/5] clk: add support for clocks provided by SCP(System Control Processor)

[Sudeep Holla]

On 05/06/15 10:36, Sudeep Holla wrote:
>
>
> On 04/06/15 21:20, Stephen Boyd wrote:
>> On 05/27, Sudeep Holla wrote:

[...]

>>> +
>>> +static unsigned long scpi_clk_recalc_rate(struct clk_hw *hw,
>>> +					  unsigned long parent_rate)
>>> +{
>>> +	struct scpi_clk *clk = to_scpi_clk(hw);
>>> +
>>> +	return scpi_ops->clk_get_val(clk->id);
>>> +}
>>> +
>>> +static long scpi_clk_round_rate(struct clk_hw *hw, unsigned long rate,
>>> +				unsigned long *parent_rate)
>>> +{
>>> +	struct scpi_clk *clk = to_scpi_clk(hw);
>>> +
>>> +	if (WARN_ON(clk->rate_min && rate < clk->rate_min))
>>> +		rate = clk->rate_min;
>>> +	if (WARN_ON(clk->rate_max && rate > clk->rate_max))
>>> +		rate = clk->rate_max;
>>> +
>>> +	return rate;
>>> +}
>>
>> Hm.. this seems really generic. It might be better to support a
>> way to tell the framework to limit the min/max rate that's
>> accepted for a clk. That could be done later though.
>>
>
> True, framework have some boundary checks in place. I will check if
> I can use it with minimum changes to the core. If not, we can take this
> up later as you suggested.
>

I found that the framework already provides clk_set_rate_range for this
purpose. Sorry for missing this earlier(seems like that's added quite
recently in v4.0). I think I still need to retain round_rate as the core
framework insists.

Regards,
Sudeep

[PATCH v3 3/5] clk: add support for clocks provided by SCP(System Control Processor)

[Stephen Boyd]

On 06/05, Sudeep Holla wrote:
> 
> 
> On 05/06/15 10:36, Sudeep Holla wrote:
> >
> >
> >On 04/06/15 21:20, Stephen Boyd wrote:
> >>On 05/27, Sudeep Holla wrote:
> 
> [...]
> 
> >>>+
> >>>+static unsigned long scpi_clk_recalc_rate(struct clk_hw *hw,
> >>>+					  unsigned long parent_rate)
> >>>+{
> >>>+	struct scpi_clk *clk = to_scpi_clk(hw);
> >>>+
> >>>+	return scpi_ops->clk_get_val(clk->id);
> >>>+}
> >>>+
> >>>+static long scpi_clk_round_rate(struct clk_hw *hw, unsigned long rate,
> >>>+				unsigned long *parent_rate)
> >>>+{
> >>>+	struct scpi_clk *clk = to_scpi_clk(hw);
> >>>+
> >>>+	if (WARN_ON(clk->rate_min && rate < clk->rate_min))
> >>>+		rate = clk->rate_min;
> >>>+	if (WARN_ON(clk->rate_max && rate > clk->rate_max))
> >>>+		rate = clk->rate_max;
> >>>+
> >>>+	return rate;
> >>>+}
> >>
> >>Hm.. this seems really generic. It might be better to support a
> >>way to tell the framework to limit the min/max rate that's
> >>accepted for a clk. That could be done later though.
> >>
> >
> >True, framework have some boundary checks in place. I will check if
> >I can use it with minimum changes to the core. If not, we can take this
> >up later as you suggested.
> >
> 
> I found that the framework already provides clk_set_rate_range for this
> purpose. Sorry for missing this earlier(seems like that's added quite
> recently in v4.0). I think I still need to retain round_rate as the core
> framework insists.

Sure, or use determine_rate if you want to limit the min/max from
the clk provider itself.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

[PATCH v3 4/5] clk: scpi: add support for cpufreq virtual device

[Sudeep Holla]

The clocks for the CPUs are provided by SCP and are managed by this
clock driver. So the cpufreq device needs to be added only after the
clock get registered and removed when this driver is unloaded.

This patch manages the cpufreq virtual device based on the clock
availability.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Cc: Mike Turquette <mturquette@linaro.org>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Cc: Liviu Dudau <Liviu.Dudau@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
Cc: linux-clk at vger.kernel.org
---
 drivers/clk/clk-scpi.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/drivers/clk/clk-scpi.c b/drivers/clk/clk-scpi.c
index 55bfa708a5b4..04068072f8d1 100644
--- a/drivers/clk/clk-scpi.c
+++ b/drivers/clk/clk-scpi.c
@@ -36,6 +36,7 @@ struct scpi_clk {
 #define to_scpi_clk(clk) container_of(clk, struct scpi_clk, hw)
 
 static struct scpi_ops *scpi_ops;
+static struct platform_device *cpufreq_dev;
 
 static unsigned long scpi_clk_recalc_rate(struct clk_hw *hw,
 					  unsigned long parent_rate)
@@ -241,6 +242,11 @@ static int scpi_clocks_remove(struct platform_device *pdev)
 	struct device *dev = &pdev->dev;
 	struct device_node *child, *np = dev->of_node;
 
+	if (cpufreq_dev) {
+		platform_device_unregister(cpufreq_dev);
+		cpufreq_dev = NULL;
+	}
+
 	for_each_available_child_of_node(np, child)
 		of_clk_del_provider(np);
 	scpi_ops = NULL;
@@ -264,6 +270,12 @@ static int scpi_clocks_probe(struct platform_device *pdev)
 			return ret;
 		}
 	}
+	/* Add the virtual cpufreq device */
+	cpufreq_dev = platform_device_register_simple("scpi-cpufreq",
+						      -1, NULL, 0);
+	if (!cpufreq_dev)
+		pr_warn("unable to register cpufreq device");
+
 	return 0;
 }
 
-- 
1.9.1

[PATCH v3 4/5] clk: scpi: add support for cpufreq virtual device

[Stephen Boyd]

On 05/27, Sudeep Holla wrote:
> The clocks for the CPUs are provided by SCP and are managed by this
> clock driver. So the cpufreq device needs to be added only after the
> clock get registered and removed when this driver is unloaded.
> 
> This patch manages the cpufreq virtual device based on the clock
> availability.
> 
> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
> Cc: Mike Turquette <mturquette@linaro.org>
> Cc: Stephen Boyd <sboyd@codeaurora.org>
> Cc: Liviu Dudau <Liviu.Dudau@arm.com>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
> Cc: linux-clk at vger.kernel.org
> ---

Acked-by: Stephen Boyd <sboyd@codeaurora.org>


I guess there's no better place for this right now.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

[PATCH v3 4/5] clk: scpi: add support for cpufreq virtual device

[Sudeep Holla]

On 04/06/15 22:19, Stephen Boyd wrote:
> On 05/27, Sudeep Holla wrote:
>> The clocks for the CPUs are provided by SCP and are managed by this
>> clock driver. So the cpufreq device needs to be added only after the
>> clock get registered and removed when this driver is unloaded.
>>
>> This patch manages the cpufreq virtual device based on the clock
>> availability.
>>
>> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
>> Cc: Mike Turquette <mturquette@linaro.org>
>> Cc: Stephen Boyd <sboyd@codeaurora.org>
>> Cc: Liviu Dudau <Liviu.Dudau@arm.com>
>> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>> Cc: Jon Medhurst (Tixy) <tixy@linaro.org>
>> Cc: linux-clk at vger.kernel.org
>> ---
>
> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
>
>

Thanks

> I guess there's no better place for this right now.
>

True

Regards,
Sudeep

[PATCH v3 5/5] cpufreq: arm_big_little: add SCPI interface driver

[Sudeep Holla]

On some ARM based systems, a separate Cortex-M based System Control
Processor(SCP) provides the overall power, clock, reset and system
control including CPU DVFS. SCPI Message Protocol is used to
communicate with the SCPI.

This patch adds a interface driver for adding OPPs and registering
the arm_big_little cpufreq driver for such systems.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: linux-pm at vger.kernel.org
---
 drivers/cpufreq/Kconfig.arm    |   9 +++
 drivers/cpufreq/Makefile       |   1 +
 drivers/cpufreq/scpi-cpufreq.c | 124 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 134 insertions(+)
 create mode 100644 drivers/cpufreq/scpi-cpufreq.c

diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm
index 611cb09239eb..6419b17b89dd 100644
--- a/drivers/cpufreq/Kconfig.arm
+++ b/drivers/cpufreq/Kconfig.arm
@@ -24,6 +24,15 @@ config ARM_VEXPRESS_SPC_CPUFREQ
           This add the CPUfreq driver support for Versatile Express
 	  big.LITTLE platforms using SPC for power management.
 
+config ARM_SCPI_CPUFREQ
+        tristate "SCPI based CPUfreq driver"
+	depends on ARM_BIG_LITTLE_CPUFREQ && ARM_SCPI_PROTOCOL
+        help
+	  This adds the CPUfreq driver support for ARM big.LITTLE platforms
+	  using SCPI protocol for CPU power management.
+
+	  This driver uses SCPI Message Protocol driver to interact with the
+	  firmware providing the CPU DVFS functionality.
 
 config ARM_EXYNOS_CPUFREQ
 	tristate "SAMSUNG EXYNOS CPUfreq Driver"
diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
index cdce92ae2e8b..02fc9f849d4b 100644
--- a/drivers/cpufreq/Makefile
+++ b/drivers/cpufreq/Makefile
@@ -79,6 +79,7 @@ obj-$(CONFIG_ARM_SA1110_CPUFREQ)	+= sa1110-cpufreq.o
 obj-$(CONFIG_ARM_SPEAR_CPUFREQ)		+= spear-cpufreq.o
 obj-$(CONFIG_ARM_TEGRA_CPUFREQ)		+= tegra-cpufreq.o
 obj-$(CONFIG_ARM_VEXPRESS_SPC_CPUFREQ)	+= vexpress-spc-cpufreq.o
+obj-$(CONFIG_ARM_SCPI_CPUFREQ)		+= scpi-cpufreq.o
 
 ##################################################################################
 # PowerPC platform drivers
diff --git a/drivers/cpufreq/scpi-cpufreq.c b/drivers/cpufreq/scpi-cpufreq.c
new file mode 100644
index 000000000000..2c3b16fd3a01
--- /dev/null
+++ b/drivers/cpufreq/scpi-cpufreq.c
@@ -0,0 +1,124 @@
+/*
+ * System Control and Power Interface (SCPI) based CPUFreq Interface driver
+ *
+ * It provides necessary ops to arm_big_little cpufreq driver.
+ *
+ * Copyright (C) 2015 ARM Ltd.
+ * Sudeep Holla <sudeep.holla@arm.com>
+ *
+ * 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 "as is" WITHOUT ANY WARRANTY of any
+ * kind, whether express or implied; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/cpufreq.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/scpi_protocol.h>
+#include <linux/types.h>
+
+#include "arm_big_little.h"
+
+static struct scpi_ops *scpi_ops;
+
+static struct scpi_dvfs_info *scpi_get_dvfs_info(struct device *cpu_dev)
+{
+	u8 domain = topology_physical_package_id(cpu_dev->id);
+
+	if (domain < 0)
+		return ERR_PTR(-EINVAL);
+	return scpi_ops->dvfs_get_info(domain);
+}
+
+static int scpi_opp_table_ops(struct device *cpu_dev, bool remove)
+{
+	int idx, ret = 0;
+	struct scpi_opp *opp;
+	struct scpi_dvfs_info *info = scpi_get_dvfs_info(cpu_dev);
+
+	if (IS_ERR(info))
+		return PTR_ERR(info);
+
+	if (!info->opps)
+		return -EIO;
+
+	for (opp = info->opps, idx = 0; idx < info->count; idx++, opp++) {
+		if (remove)
+			dev_pm_opp_remove(cpu_dev, opp->freq);
+		else
+			ret = dev_pm_opp_add(cpu_dev, opp->freq,
+					     opp->m_volt * 1000);
+		if (ret) {
+			dev_warn(cpu_dev, "failed to add opp %uHz %umV\n",
+				 opp->freq, opp->m_volt);
+			while (idx-- > 0)
+				dev_pm_opp_remove(cpu_dev, (--opp)->freq);
+			return ret;
+		}
+	}
+	return ret;
+}
+
+static int scpi_get_transition_latency(struct device *cpu_dev)
+{
+	struct scpi_dvfs_info *info = scpi_get_dvfs_info(cpu_dev);
+
+	if (IS_ERR(info))
+		return PTR_ERR(info);
+	return info->latency;
+}
+
+static int scpi_init_opp_table(struct device *cpu_dev)
+{
+	return scpi_opp_table_ops(cpu_dev, false);
+}
+
+static void scpi_free_opp_table(struct device *cpu_dev)
+{
+	scpi_opp_table_ops(cpu_dev, true);
+}
+
+static struct cpufreq_arm_bL_ops scpi_cpufreq_ops = {
+	.name	= "scpi",
+	.get_transition_latency = scpi_get_transition_latency,
+	.init_opp_table = scpi_init_opp_table,
+	.free_opp_table = scpi_free_opp_table,
+};
+
+static int scpi_cpufreq_probe(struct platform_device *pdev)
+{
+	scpi_ops = get_scpi_ops();
+	if (!scpi_ops)
+		return -EIO;
+
+	return bL_cpufreq_register(&scpi_cpufreq_ops);
+}
+
+static int scpi_cpufreq_remove(struct platform_device *pdev)
+{
+	bL_cpufreq_unregister(&scpi_cpufreq_ops);
+	scpi_ops = NULL;
+	return 0;
+}
+
+static struct platform_driver scpi_cpufreq_platdrv = {
+	.driver = {
+		.name	= "scpi-cpufreq",
+		.owner	= THIS_MODULE,
+	},
+	.probe		= scpi_cpufreq_probe,
+	.remove		= scpi_cpufreq_remove,
+};
+module_platform_driver(scpi_cpufreq_platdrv);
+
+MODULE_AUTHOR("Sudeep Holla <sudeep.holla@arm.com>");
+MODULE_DESCRIPTION("ARM SCPI CPUFreq interface driver");
+MODULE_LICENSE("GPL v2");
-- 
1.9.1