[PATCH v2 0/9] firmware: Add initial support for Arm FF-A

[PATCH v2 0/9] firmware: Add initial support for Arm FF-A

[Sudeep Holla]

Hi all,

Let me start stating this is just initial implementation to check on
the idea of providing more in-kernel and userspace support. Lot of things
are still work in progress, I am posting just to get the early feedback
before building lot of things on this idea. Consider this more as RFC
though not tagged explicity(just to avoid it being ignored :))

Arm Firmware Framework for Armv8-A specification[1] describes a software
architecture that provides mechanism to utilise the virtualization
extension to isolate software images and describes interfaces that
standardize communication between the various software images. This
includes communication between images in the Secure and Normal world.

The main idea here is to create FFA device to establish any communication
with a partition(secure or normal world VM).

If it is a partition managed by hypervisor, then we will register chardev
associated with each of those partition FFA device.

/dev/arm_ffa:

e3a48fa5-dc54-4a8b-898b-bdc4dfeeb7b8
49f65057-d002-4ae2-b4ee-d31c7940a13d

For in-kernel usage(mostly communication with secure partitions), only
in-kernel APIs are accessible(no userspace). There may be a need to
provide userspace access instead of in-kernel, it is not yet support
in this series as we need way to identify those and I am not sure if
that belong to DT.

--
Regards,
Sudeep

[1] https://developer.arm.com/documentation/den0077/latest

v1->v2:
	- Moved userspace code to a separate unit, will move to separate
	  module. Still working on minimizing initcall dependencies and
	  exported functions to reuse some of the code.
	- Fixed couple of minor issues pointed out
	- Dropped ASYNC send message as I haven't been able to test

Sudeep Holla (8):
  dt-bindings: Arm: Extend FF-A binding to support in-kernel usage of
    partitions
  arm64: smccc: Add support for SMCCCv1.2 input/output registers
  firmware: arm_ffa: Add initial FFA bus support for device enumeration
  firmware: arm_ffa: Add initial Arm FFA driver support
  firmware: arm_ffa: Add support for SMCCC as transport to FFA driver
  firmware: arm_ffa: Setup in-kernel users of FFA partitions
  firmware: arm_ffa: Setup and register all the KVM managed partitions
  firmware: arm_ffa: Add support for MEM_* interfaces

Will Deacon (1):
  dt-bindings: Arm: Add Firmware Framework for Armv8-A (FF-A) binding

 .../devicetree/bindings/arm/arm,ffa-hyp.yaml  | 102 +++
 .../devicetree/bindings/arm/arm,ffa.yaml      |  58 ++
 .../reserved-memory/arm,ffa-memory.yaml       |  71 ++
 arch/arm64/kernel/asm-offsets.c               |   4 +
 arch/arm64/kernel/smccc-call.S                |  22 +
 drivers/firmware/Kconfig                      |   1 +
 drivers/firmware/Makefile                     |   1 +
 drivers/firmware/arm_ffa/Kconfig              |  21 +
 drivers/firmware/arm_ffa/Makefile             |   6 +
 drivers/firmware/arm_ffa/bus.c                | 199 +++++
 drivers/firmware/arm_ffa/common.h             |  35 +
 drivers/firmware/arm_ffa/driver.c             | 737 ++++++++++++++++++
 drivers/firmware/arm_ffa/hyp_partitions.c     | 132 ++++
 drivers/firmware/arm_ffa/smccc.c              |  54 ++
 include/linux/arm-smccc.h                     |  50 ++
 include/linux/arm_ffa.h                       | 310 ++++++++
 include/uapi/linux/arm_ffa.h                  |  67 ++
 17 files changed, 1870 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/arm,ffa-hyp.yaml
 create mode 100644 Documentation/devicetree/bindings/arm/arm,ffa.yaml
 create mode 100644 Documentation/devicetree/bindings/reserved-memory/arm,ffa-memory.yaml
 create mode 100644 drivers/firmware/arm_ffa/Kconfig
 create mode 100644 drivers/firmware/arm_ffa/Makefile
 create mode 100644 drivers/firmware/arm_ffa/bus.c
 create mode 100644 drivers/firmware/arm_ffa/common.h
 create mode 100644 drivers/firmware/arm_ffa/driver.c
 create mode 100644 drivers/firmware/arm_ffa/hyp_partitions.c
 create mode 100644 drivers/firmware/arm_ffa/smccc.c
 create mode 100644 include/linux/arm_ffa.h
 create mode 100644 include/uapi/linux/arm_ffa.h

--
2.25.1

[PATCH v2 1/9] dt-bindings: Arm: Add Firmware Framework for Armv8-A (FF-A) binding

[Sudeep Holla]

From: Will Deacon <will@kernel.org>

Add devicetree bindings for a FF-A-compliant hypervisor, its partitions
and their memory regions. The naming is ludicrous but also not by fault.

Cc: Rob Herring <robh+dt@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
(sudeep.holla: Dropped PSA from name and elsewhere as it seem to have
 disappeared mysteriously just before the final release, renamed
 arm,ffa.yaml as arm,ffa-hyp.yaml)
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 .../devicetree/bindings/arm/arm,ffa-hyp.yaml  | 102 ++++++++++++++++++
 .../reserved-memory/arm,ffa-memory.yaml       |  71 ++++++++++++
 2 files changed, 173 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/arm,ffa-hyp.yaml
 create mode 100644 Documentation/devicetree/bindings/reserved-memory/arm,ffa-memory.yaml

Hi Rob,

Sorry I haven't waited for the discussion on the previous version to
conclude and posting new version to re-trigger (more) discussion on the
series.

Regards,
Sudeep

diff --git a/Documentation/devicetree/bindings/arm/arm,ffa-hyp.yaml b/Documentation/devicetree/bindings/arm/arm,ffa-hyp.yaml
new file mode 100644
index 000000000000..054e36d317ff
--- /dev/null
+++ b/Documentation/devicetree/bindings/arm/arm,ffa-hyp.yaml
@@ -0,0 +1,102 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/arm/arm,ffa-hyp.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Arm Firmware Framework for Arm v8-A
+
+maintainers:
+  - Will Deacon <will@kernel.org>
+
+description: |
+  Firmware frameworks implementing partition setup according to the FF-A
+  specification defined by ARM document number ARM DEN 0077A ("Arm Firmware
+  Framework for Arm v8-A") [0] must provide a "manifest and image" for each
+  partition to the "partition manager" so that the partition execution contexts
+  can be initialised.
+
+  In the case of a virtual FFA instance, the manifest and image details can be
+  passed to the hypervisor (e.g. Linux KVM) using this binding.
+
+  [0] https://developer.arm.com/docs/den0077/latest
+
+properties:
+  $nodename:
+    const: ffa_hyp
+
+  compatible:
+    oneOf:
+      - const: arm,ffa-1.0-hypervisor
+
+  memory-region:
+    $ref: '/schemas/types.yaml#/definitions/phandle'
+    description: |
+      A phandle to the reserved memory region [1] to be used by the hypervisor.
+      The reserved memory region must be compatible with
+      "arm,ffa-1.0-hypervisor-memory-region".
+
+      [1] Documentation/devicetree/bindings/reserved-memory/reserved-memory.txt
+
+patternProperties:
+  "^ffa_partition[0-9]+$":
+    type: object
+    description: One or more child nodes, each describing an FFA partition.
+    properties:
+      $nodename:
+        const: ffa_partition
+
+      compatible:
+        oneOf:
+          - const: arm,ffa-1.0-partition
+
+      uuid:
+        $ref: '/schemas/types.yaml#definitions/string'
+        description: |
+          The 128-bit UUID [2] of the service implemented by this partition.
+
+          [2] https://tools.ietf.org/html/rfc4122
+
+      nr-exec-ctxs:
+        $ref: '/schemas/types.yaml#/definitions/uint32'
+        description: |
+          The number of virtual CPUs to instantiate for this partition.
+
+      exec-state:
+        description: The execution state in which to execute the partition.
+        oneOf:
+          - const: "AArch64"
+          - const: "AArch32"
+
+      entry-point:
+        $ref: '/schemas/types.yaml#/definitions/uint32-matrix'
+        description: |
+          The entry address of the partition specified as an Intermediate
+          Physical Address (IPA) encoded according to the '#address-cells'
+          property.
+
+      memory-region:
+        $ref: '/schemas/types.yaml#/definitions/phandle-array'
+        description: |
+          A list of phandles to FFA reserved memory regions [3] for this
+          partition.
+
+          [3] Documentation/devicetree/bindings/reserved-memory/arm,ffa-memory.yaml
+
+additionalProperties: false
+
+examples:
+  - |
+    ffa_hyp {
+        compatible = "arm,ffa-1.0-hypervisor";
+        memory-region = <&ffa_hyp_reserved>;
+
+        ffa_partition0 {
+            compatible = "arm,ffa-1.0-partition";
+            uuid = "12345678-9abc-def0-1234-56789abcdef0";
+            nr-exec-ctxs = <2>;
+            exec-state = "AArch64";
+            entry-point = <0x80000>;
+            memory-region = <&ffa_reserved0 &ffa_reserved1>;
+        };
+    };
diff --git a/Documentation/devicetree/bindings/reserved-memory/arm,ffa-memory.yaml b/Documentation/devicetree/bindings/reserved-memory/arm,ffa-memory.yaml
new file mode 100644
index 000000000000..5335e07abcfc
--- /dev/null
+++ b/Documentation/devicetree/bindings/reserved-memory/arm,ffa-memory.yaml
@@ -0,0 +1,71 @@
+# SPDX-License-Identifier: GPL-2.0
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/reserved-memory/arm,ffa-memory.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Memory Region for Arm Firmware Framework for Arm v8-A
+
+maintainers:
+  - Will Deacon <will@kernel.org>
+
+description: |
+  This binding allows a FF-A implementation to describe the normal memory
+  regions of a partition [1] to a hypervisor according to [2].
+
+  The physical address range reserved for the partition can be specified as a
+  static allocation using the 'reg' property or as a dynamic allocation using
+  the 'size' property. If both properties are omitted, then the hypervisor can
+  allocate physical memory for the partition however it sees fit.
+
+  [1] Documentation/devicetree/bindings/arm/arm,ffa.yaml
+  [2] Documentation/devicetree/bindings/reserved-memory/reserved-memory.txt
+
+properties:
+  $nodename:
+    pattern: "^ffa_mem(@[0-9a-f]+)?$"
+
+  compatible:
+    oneOf:
+      - const: arm,ffa-1.0-partition-memory-region
+
+  ipa-range:
+    $ref: '/schemas/types.yaml#/definitions/uint32-matrix'
+    description: |
+      The Intermediate Physical Address (IPA) range (encoded in the same way as
+      a 'reg' property) at which to map the physical memory. If the IPA range is
+      larger than the physical memory region then the region is mapped starting
+      at the base of the IPA range.
+
+  read-only:
+    type: boolean
+    description: |
+      (static allocation only) The memory region has been pre-populated
+      by the firmware framework and must be mapped without write permission
+      at stage 2.
+
+  non-executable:
+    type: boolean
+    description: |
+      The memory region must be mapped without execute permission at stage 2.
+
+
+required:
+  - compatible
+
+# The "reserved-memory" binding defines additional properties.
+additionalProperties: true
+
+examples:
+  - |
+    reserved-memory {
+        #address-cells = <2>;
+        #size-cells = <2>;
+
+        ffa_reserved0: ffa_mem@100000000 {
+            compatible = "arm,ffa-1.0-partition-memory-region";
+            reg = <0x1 0x0 0x0 0x04000000>;          // 64M @ 1GB
+            ipa-range = <0x0 0x0 0x0 0x04000000>;    // 64M @ 0x0
+            read-only;
+        };
+    };
-- 
2.25.1

[PATCH v2 2/9] dt-bindings: Arm: Extend FF-A binding to support in-kernel usage of partitions

[Sudeep Holla]

Since the FF-A v1.0 specification doesn't list the UUID of all the
partitions in the discovery API, we need to specify the UUID of the
partitions that need to be accessed by drivers within the kernel.

This extends the binding to provide the list of partitions that kernel
drivers may need to access and are not part of the partitions managed
by the hypervisor.

Cc: Rob Herring <robh+dt@kernel.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 .../devicetree/bindings/arm/arm,ffa.yaml      | 58 +++++++++++++++++++
 1 file changed, 58 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/arm,ffa.yaml

Hi,

I struggled to keep both these(arm,ffa and arm,ffa-hyp) in the single yaml
file. I couldn't understand the errors and couldn't find an example that
has something similar. I will continue to explore, just posting as separate
for now to continue discussions.

Regards,
Sudeep


diff --git a/Documentation/devicetree/bindings/arm/arm,ffa.yaml b/Documentation/devicetree/bindings/arm/arm,ffa.yaml
new file mode 100644
index 000000000000..a014a5801c34
--- /dev/null
+++ b/Documentation/devicetree/bindings/arm/arm,ffa.yaml
@@ -0,0 +1,58 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/arm/arm,ffa.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Arm Firmware Framework for Arm v8-A (in-kernel users)
+
+maintainers:
+  - Sudeep Holla <sudeep.holla@arm.com>
+
+description: |
+  Firmware frameworks implementing partition according to the FF-A
+  specification defined by ARM document number ARM DEN 0077A ("Arm Firmware
+  Framework for Arm v8-A") [0], providing services to be used by other
+  partitions.
+
+  [0] https://developer.arm.com/docs/den0077/latest
+
+properties:
+  $nodename:
+    const: ffa
+
+  compatible:
+    oneOf:
+      - const: arm,ffa-1.0
+
+patternProperties:
+  "^ffa_partition[0-9]+$":
+    type: object
+    description: One or more child nodes, each describing an FFA partition.
+    properties:
+      $nodename:
+        const: ffa_partition
+
+      compatible:
+        oneOf:
+          - const: arm,ffa-1.0-partition
+
+      uuid:
+        $ref: '/schemas/types.yaml#definitions/string'
+        description: |
+          The 128-bit UUID [2] of the service implemented by this partition.
+
+          [2] https://tools.ietf.org/html/rfc4122
+
+additionalProperties: false
+
+examples:
+  - |
+    ffa {
+      compatible = "arm,ffa-1.0";
+
+      ffa_partition0 {
+        compatible = "arm,ffa-1.0-partition";
+        uuid = "12345678-9abc-def0-1234-56789abcdef0";
+      };
+    };
-- 
2.25.1

[PATCH v2 3/9] arm64: smccc: Add support for SMCCCv1.2 input/output registers

[Sudeep Holla]

SMCCC v1.2 allows x8-x17 to be used as parameter registers and x4—x17
to be used as result registers in SMC64/HVC64. Arm Firmware Framework
for Armv8-A specification makes use of x0-x7 as parameter and result
registers.

Current SMCCC interface in the kernel just use x0-x7 as parameter and
x0-x3 as result registers. Let us add new interface to support x0-x7
as parameter and result registers. This can be extended to include
x8-x17 when there are users for the same.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 arch/arm64/kernel/asm-offsets.c |  4 +++
 arch/arm64/kernel/smccc-call.S  | 22 +++++++++++++++
 include/linux/arm-smccc.h       | 50 +++++++++++++++++++++++++++++++++
 3 files changed, 76 insertions(+)

diff --git a/arch/arm64/kernel/asm-offsets.c b/arch/arm64/kernel/asm-offsets.c
index 7d32fc959b1a..32bcc25337ce 100644
--- a/arch/arm64/kernel/asm-offsets.c
+++ b/arch/arm64/kernel/asm-offsets.c
@@ -122,6 +122,10 @@ int main(void)
   DEFINE(ARM_SMCCC_RES_X2_OFFS,		offsetof(struct arm_smccc_res, a2));
   DEFINE(ARM_SMCCC_QUIRK_ID_OFFS,	offsetof(struct arm_smccc_quirk, id));
   DEFINE(ARM_SMCCC_QUIRK_STATE_OFFS,	offsetof(struct arm_smccc_quirk, state));
+  DEFINE(ARM_SMCCC_V1_2_RES_X0_OFFS,	offsetof(struct arm_smccc_v1_2_res, a0));
+  DEFINE(ARM_SMCCC_V1_2_RES_X2_OFFS,	offsetof(struct arm_smccc_v1_2_res, a2));
+  DEFINE(ARM_SMCCC_V1_2_RES_X4_OFFS,	offsetof(struct arm_smccc_v1_2_res, a4));
+  DEFINE(ARM_SMCCC_V1_2_RES_X6_OFFS,	offsetof(struct arm_smccc_v1_2_res, a6));
   BLANK();
   DEFINE(HIBERN_PBE_ORIG,	offsetof(struct pbe, orig_address));
   DEFINE(HIBERN_PBE_ADDR,	offsetof(struct pbe, address));
diff --git a/arch/arm64/kernel/smccc-call.S b/arch/arm64/kernel/smccc-call.S
index d62447964ed9..0ea15c1742f3 100644
--- a/arch/arm64/kernel/smccc-call.S
+++ b/arch/arm64/kernel/smccc-call.S
@@ -43,3 +43,25 @@ SYM_FUNC_START(__arm_smccc_hvc)
 	SMCCC	hvc
 SYM_FUNC_END(__arm_smccc_hvc)
 EXPORT_SYMBOL(__arm_smccc_hvc)
+
+	.macro SMCCC_v1_2 instr
+	.cfi_startproc
+	\instr #0
+	ldr x8, [sp]
+	stp x0, x1, [x8, #ARM_SMCCC_V1_2_RES_X0_OFFS]
+	stp x2, x3, [x8, #ARM_SMCCC_V1_2_RES_X2_OFFS]
+	stp x4, x5, [x8, #ARM_SMCCC_V1_2_RES_X4_OFFS]
+	stp x6, x7, [x8, #ARM_SMCCC_V1_2_RES_X6_OFFS]
+	ret
+	.cfi_endproc
+.endm
+
+SYM_FUNC_START(arm_smccc_v1_2_hvc)
+	SMCCC_v1_2 hvc
+SYM_FUNC_END(arm_smccc_v1_2_hvc)
+EXPORT_SYMBOL(arm_smccc_v1_2_hvc)
+
+SYM_FUNC_START(arm_smccc_v1_2_smc)
+	SMCCC_v1_2 smc
+SYM_FUNC_END(arm_smccc_v1_2_smc)
+EXPORT_SYMBOL(arm_smccc_v1_2_smc)
diff --git a/include/linux/arm-smccc.h b/include/linux/arm-smccc.h
index f860645f6512..aed27214c88f 100644
--- a/include/linux/arm-smccc.h
+++ b/include/linux/arm-smccc.h
@@ -155,6 +155,56 @@ struct arm_smccc_res {
 	unsigned long a3;
 };
 
+#ifdef CONFIG_ARM64
+/* TODO Need to implement for ARM too */
+/**
+ * struct arm_smccc_v1_2_res - Result from SMC/HVC call
+ * @a0-a3 result values from registers 0 to 3
+ */
+struct arm_smccc_v1_2_res {
+	unsigned long a0;
+	unsigned long a1;
+	unsigned long a2;
+	unsigned long a3;
+	unsigned long a4;
+	unsigned long a5;
+	unsigned long a6;
+	unsigned long a7;
+};
+
+/**
+ * arm_smccc_v1_2_hvc() - make HVC calls
+ * @a0-a7: arguments passed in registers 0 to 7
+ * @res: result values from registers 0 to 7
+ *
+ * This function is used to make SMC calls following SMC Calling Convention
+ * v1.2 or above. The content of the supplied param are copied to registers
+ * 0 to 7 prior to the SMC instruction. The return values are updated with
+ * the content from register 0 to 7 on return from the SMC instruction.
+ */
+asmlinkage
+void arm_smccc_v1_2_hvc(unsigned long a0, unsigned long a1, unsigned long a2,
+			unsigned long a3, unsigned long a4, unsigned long a5,
+			unsigned long a6, unsigned long a7,
+			struct arm_smccc_v1_2_res  *res);
+
+/**
+ * arm_smccc_v1_2_smc() - make SMC calls
+ * @a0-a7: arguments passed in registers 0 to 7
+ * @res: result values from registers 0 to 7
+ *
+ * This function is used to make SMC calls following SMC Calling Convention
+ * v1.2 or above. The content of the supplied param are copied to registers
+ * 0 to 7 prior to the SMC instruction. The return values are updated with
+ * the content from register 0 to 7 on return from the SMC instruction.
+ */
+asmlinkage
+void arm_smccc_v1_2_smc(unsigned long a0, unsigned long a1, unsigned long a2,
+			unsigned long a3, unsigned long a4, unsigned long a5,
+			unsigned long a6, unsigned long a7,
+			struct arm_smccc_v1_2_res  *res);
+#endif
+
 /**
  * struct arm_smccc_quirk - Contains quirk information
  * @id: quirk identification
-- 
2.25.1

[PATCH v2 4/9] firmware: arm_ffa: Add initial FFA bus support for device enumeration

[Sudeep Holla]

The Arm FF for Armv8-A specification has concept of endpoints or
partitions. In the Normal world, a partition could be a VM when
the Virtualization extension is enabled or the kernel itself.

In order to handle multiple partitions, we can create a FFA device for
each partition on a dedicated FFA bus. Similarly, different drivers
requiring FFA transport can be registered on the same bus. We can match
the device and drivers using UUID. This is mostly for the in-kernel
users with FFA drivers.

However, to support usage of FFA transport from user-space, there is also
a provision to create character device interface for the same.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/Kconfig          |   1 +
 drivers/firmware/Makefile         |   1 +
 drivers/firmware/arm_ffa/Kconfig  |  16 +++
 drivers/firmware/arm_ffa/Makefile |   4 +
 drivers/firmware/arm_ffa/bus.c    | 205 ++++++++++++++++++++++++++++++
 include/linux/arm_ffa.h           |  81 ++++++++++++
 6 files changed, 308 insertions(+)
 create mode 100644 drivers/firmware/arm_ffa/Kconfig
 create mode 100644 drivers/firmware/arm_ffa/Makefile
 create mode 100644 drivers/firmware/arm_ffa/bus.c
 create mode 100644 include/linux/arm_ffa.h

diff --git a/drivers/firmware/Kconfig b/drivers/firmware/Kconfig
index 3315e3c21586..be9056226a07 100644
--- a/drivers/firmware/Kconfig
+++ b/drivers/firmware/Kconfig
@@ -295,6 +295,7 @@ config TURRIS_MOX_RWTM
 	  other manufacturing data and also utilize the Entropy Bit Generator
 	  for hardware random number generation.
 
+source "drivers/firmware/arm_ffa/Kconfig"
 source "drivers/firmware/broadcom/Kconfig"
 source "drivers/firmware/google/Kconfig"
 source "drivers/firmware/efi/Kconfig"
diff --git a/drivers/firmware/Makefile b/drivers/firmware/Makefile
index 5e013b6a3692..546ac8e7f6d0 100644
--- a/drivers/firmware/Makefile
+++ b/drivers/firmware/Makefile
@@ -22,6 +22,7 @@ obj-$(CONFIG_TI_SCI_PROTOCOL)	+= ti_sci.o
 obj-$(CONFIG_TRUSTED_FOUNDATIONS) += trusted_foundations.o
 obj-$(CONFIG_TURRIS_MOX_RWTM)	+= turris-mox-rwtm.o
 
+obj-y				+= arm_ffa/
 obj-y				+= arm_scmi/
 obj-y				+= broadcom/
 obj-y				+= meson/
diff --git a/drivers/firmware/arm_ffa/Kconfig b/drivers/firmware/arm_ffa/Kconfig
new file mode 100644
index 000000000000..261a3660650a
--- /dev/null
+++ b/drivers/firmware/arm_ffa/Kconfig
@@ -0,0 +1,16 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config ARM_FFA_TRANSPORT
+	tristate "Arm Firmware Framework for Armv8-A"
+	depends on OF
+	depends on ARM64
+	default n
+	help
+	  This Firmware Framework(FF) for Arm A-profile processors describes
+	  interfaces that standardize communication between the various
+	  software images which includes communication between images in
+	  the Secure world and Normal world. It also leverages the
+	  virtualization extension to isolate software images provided
+	  by an ecosystem of vendors from each other.
+
+	  This driver provides interface for all the client drivers making
+	  use of the features offered by ARM FF-A.
diff --git a/drivers/firmware/arm_ffa/Makefile b/drivers/firmware/arm_ffa/Makefile
new file mode 100644
index 000000000000..bfe4323a8784
--- /dev/null
+++ b/drivers/firmware/arm_ffa/Makefile
@@ -0,0 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0-only
+ffa-bus-y = bus.o
+ffa-module-objs := $(ffa-bus-y)
+obj-$(CONFIG_ARM_FFA_TRANSPORT) = ffa-module.o
diff --git a/drivers/firmware/arm_ffa/bus.c b/drivers/firmware/arm_ffa/bus.c
new file mode 100644
index 000000000000..5d7102419e6c
--- /dev/null
+++ b/drivers/firmware/arm_ffa/bus.c
@@ -0,0 +1,205 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020 ARM Ltd.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/arm_ffa.h>
+#include <linux/device.h>
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+
+#define DEVICE_NAME "arm_ffa"
+#define FFA_MAX_CDEVS	32
+
+static DEFINE_IDA(ffa_dev_id);
+static dev_t ffa_devt;
+
+static int ffa_device_match(struct device *dev, struct device_driver *drv)
+{
+	const struct ffa_device_id *id_table;
+	struct ffa_device *ffa_dev;
+
+	id_table = to_ffa_driver(drv)->id_table;
+	ffa_dev = to_ffa_dev(dev);
+
+	while (!uuid_is_null(&id_table->uuid)) {
+		if (uuid_equal(&ffa_dev->uuid, &id_table->uuid))
+			return 1;
+		id_table++;
+	}
+
+	return 0;
+}
+
+static int ffa_device_probe(struct device *dev)
+{
+	struct ffa_driver *ffa_drv = to_ffa_driver(dev->driver);
+	struct ffa_device *ffa_dev = to_ffa_dev(dev);
+
+	if (!ffa_device_match(dev, dev->driver))
+		return -ENODEV;
+
+	return ffa_drv->probe(ffa_dev);
+}
+
+static int ffa_device_uevent(struct device *dev, struct kobj_uevent_env *env)
+{
+	uuid_t *dev_id = &to_ffa_dev(dev)->uuid;
+
+	return add_uevent_var(env, "MODALIAS=arm_ffa:%pUb", dev_id);
+}
+
+struct bus_type ffa_bus_type = {
+	.name		= "arm_ffa",
+	.match		= ffa_device_match,
+	.probe		= ffa_device_probe,
+	.uevent		= ffa_device_uevent,
+};
+EXPORT_SYMBOL_GPL(ffa_bus_type);
+
+int ffa_driver_register(struct ffa_driver *driver, struct module *owner,
+			const char *mod_name)
+{
+	int ret;
+
+	driver->driver.bus = &ffa_bus_type;
+	driver->driver.name = driver->name;
+	driver->driver.owner = owner;
+	driver->driver.mod_name = mod_name;
+
+	ret = driver_register(&driver->driver);
+	if (!ret)
+		pr_debug("registered new ffa driver %s\n", driver->name);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(ffa_driver_register);
+
+void ffa_driver_unregister(struct ffa_driver *driver)
+{
+	driver_unregister(&driver->driver);
+}
+EXPORT_SYMBOL_GPL(ffa_driver_unregister);
+
+static void ffa_release_device(struct device *dev)
+{
+	struct ffa_device *ffa_dev = to_ffa_dev(dev);
+
+	kfree(ffa_dev);
+}
+
+static int __ffa_devices_unregister(struct device *dev, void *data)
+{
+	ffa_release_device(dev);
+
+	return 0;
+}
+
+static void ffa_devices_unregister(void)
+{
+	bus_for_each_dev(&ffa_bus_type, NULL, NULL,
+			 __ffa_devices_unregister);
+}
+
+static char *
+ffa_devnode(struct device *dev, umode_t *mode, kuid_t *uid, kgid_t *gid)
+{
+	return kasprintf(GFP_KERNEL, DEVICE_NAME "/%s", dev_name(dev));
+}
+
+static struct device_type ffa_dev_type = {
+	.devnode = ffa_devnode,
+};
+
+int ffa_device_register(struct ffa_device *ffa_dev)
+{
+	int ret;
+	struct cdev *cdev;
+	struct device *dev;
+
+	if (!ffa_dev)
+		return -EINVAL;
+
+	dev = &ffa_dev->dev;
+	cdev = &ffa_dev->cdev;
+
+	dev->bus = &ffa_bus_type;
+	dev->type = &ffa_dev_type;
+	dev->release = ffa_release_device;
+
+	device_initialize(dev);
+
+	if (cdev->ops) {
+		ret = ida_simple_get(&ffa_dev_id, 0, FFA_MAX_CDEVS, GFP_KERNEL);
+		if (ret < 0) {
+			put_device(dev);
+			return ret;
+		}
+
+		dev->devt = MKDEV(MAJOR(ffa_devt), ret);
+
+		cdev->owner = cdev->ops->owner;
+	}
+
+	ret = cdev_device_add(cdev, dev);
+	if (ret) {
+		dev_err(dev, "unable to cdev_device_add() %s, major %d, minor %d, err=%d\n",
+			dev_name(dev), MAJOR(dev->devt), MINOR(dev->devt),
+			ret);
+		put_device(dev);
+		return ret;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(ffa_device_register);
+
+void ffa_device_unregister(struct ffa_device *ffa_dev)
+{
+	if (!ffa_dev)
+		return;
+
+	cdev_device_del(&ffa_dev->cdev, &ffa_dev->dev);
+
+	put_device(&ffa_dev->dev);
+}
+EXPORT_SYMBOL_GPL(ffa_device_unregister);
+
+static int __init arm_ffa_bus_init(void)
+{
+	int ret;
+
+	ret = alloc_chrdev_region(&ffa_devt, 0, FFA_MAX_CDEVS, DEVICE_NAME);
+	if (ret) {
+		pr_err("failed to allocate char dev region\n");
+		return ret;
+	}
+
+	ret = bus_register(&ffa_bus_type);
+	if (ret) {
+		pr_err("ffa bus register failed (%d)\n", ret);
+		unregister_chrdev_region(ffa_devt, FFA_MAX_CDEVS);
+	}
+
+	return ret;
+}
+subsys_initcall(arm_ffa_bus_init);
+
+static void __exit arm_ffa_bus_exit(void)
+{
+	ffa_devices_unregister();
+	bus_unregister(&ffa_bus_type);
+	unregister_chrdev_region(ffa_devt, FFA_MAX_CDEVS);
+}
+
+module_exit(arm_ffa_bus_exit);
+
+MODULE_ALIAS("arm-ffa-bus");
+MODULE_AUTHOR("Sudeep Holla <sudeep.holla@arm.com>");
+MODULE_DESCRIPTION("Arm FF-A bus driver");
+MODULE_LICENSE("GPL v2");
diff --git a/include/linux/arm_ffa.h b/include/linux/arm_ffa.h
new file mode 100644
index 000000000000..2fe16176149f
--- /dev/null
+++ b/include/linux/arm_ffa.h
@@ -0,0 +1,81 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2020 ARM Ltd.
+ */
+
+#ifndef _LINUX_ARM_FFA_H
+#define _LINUX_ARM_FFA_H
+
+#include <linux/cdev.h>
+#include <linux/device.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/uuid.h>
+
+struct ffa_device {
+	u32 vm_id;
+	uuid_t uuid;
+	struct device dev;
+	struct cdev cdev;
+};
+
+#define to_ffa_dev(d) container_of(d, struct ffa_device, dev)
+
+struct ffa_device_id {
+	uuid_t uuid;
+};
+
+struct ffa_driver {
+	const char *name;
+	int (*probe)(struct ffa_device *sdev);
+	void (*remove)(struct ffa_device *sdev);
+	const struct ffa_device_id *id_table;
+
+	struct device_driver driver;
+};
+
+#define to_ffa_driver(d) container_of(d, struct ffa_driver, driver)
+
+#if IS_REACHABLE(CONFIG_ARM_FFA_TRANSPORT)
+int ffa_device_register(struct ffa_device *ffa_dev);
+void ffa_device_unregister(struct ffa_device *ffa_dev);
+int ffa_driver_register(struct ffa_driver *driver, struct module *owner,
+			const char *mod_name);
+void ffa_driver_unregister(struct ffa_driver *driver);
+
+#else
+static inline int ffa_device_register(struct ffa_device *ffa_dev)
+{
+	return -EINVAL;
+}
+
+static inline void ffa_device_unregister(struct ffa_device *dev) {}
+
+static inline int
+ffa_driver_register(struct ffa_driver *driver, struct module *owner,
+		    const char *mod_name)
+{
+	return -EINVAL;
+}
+
+static inline void ffa_driver_unregister(struct ffa_driver *driver) {}
+
+#endif /* CONFIG_ARM_FFA_TRANSPORT */
+
+#define ffa_register(driver) \
+	ffa_driver_register(driver, THIS_MODULE, KBUILD_MODNAME)
+#define ffa_unregister(driver) \
+	ffa_driver_unregister(driver)
+
+/**
+ * module_ffa_driver() - Helper macro for registering a psa_ffa driver
+ * @__ffa_driver: ffa_driver structure
+ *
+ * Helper macro for psa_ffa drivers to set up proper module init / exit
+ * functions.  Replaces module_init() and module_exit() and keeps people from
+ * printing pointless things to the kernel log when their driver is loaded.
+ */
+#define module_ffa_driver(__ffa_driver)	\
+	module_driver(__ffa_driver, ffa_register, ffa_unregister)
+
+#endif /* _LINUX_ARM_FFA_H */
-- 
2.25.1

[PATCH v2 5/9] firmware: arm_ffa: Add initial Arm FFA driver support

[Sudeep Holla]

This just add a basic driver that sets up the transport(e.g. SMCCC),
checks the FFA version implemented, get the partition ID for self and
sets up the Tx/Rx buffers for communication.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/arm_ffa/Makefile |   3 +-
 drivers/firmware/arm_ffa/bus.c    |  12 +-
 drivers/firmware/arm_ffa/common.h |  26 ++++
 drivers/firmware/arm_ffa/driver.c | 219 ++++++++++++++++++++++++++++++
 include/linux/arm_ffa.h           |  77 +++++++++++
 5 files changed, 326 insertions(+), 11 deletions(-)
 create mode 100644 drivers/firmware/arm_ffa/common.h
 create mode 100644 drivers/firmware/arm_ffa/driver.c

diff --git a/drivers/firmware/arm_ffa/Makefile b/drivers/firmware/arm_ffa/Makefile
index bfe4323a8784..82d0d35c5324 100644
--- a/drivers/firmware/arm_ffa/Makefile
+++ b/drivers/firmware/arm_ffa/Makefile
@@ -1,4 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
 ffa-bus-y = bus.o
-ffa-module-objs := $(ffa-bus-y)
+ffa-driver-y = driver.o
+ffa-module-objs := $(ffa-bus-y) $(ffa-driver-y)
 obj-$(CONFIG_ARM_FFA_TRANSPORT) = ffa-module.o
diff --git a/drivers/firmware/arm_ffa/bus.c b/drivers/firmware/arm_ffa/bus.c
index 5d7102419e6c..2dbcf356bd8b 100644
--- a/drivers/firmware/arm_ffa/bus.c
+++ b/drivers/firmware/arm_ffa/bus.c
@@ -170,7 +170,7 @@ void ffa_device_unregister(struct ffa_device *ffa_dev)
 }
 EXPORT_SYMBOL_GPL(ffa_device_unregister);
 
-static int __init arm_ffa_bus_init(void)
+int __init arm_ffa_bus_init(void)
 {
 	int ret;
 
@@ -188,18 +188,10 @@ static int __init arm_ffa_bus_init(void)
 
 	return ret;
 }
-subsys_initcall(arm_ffa_bus_init);
 
-static void __exit arm_ffa_bus_exit(void)
+void __exit arm_ffa_bus_exit(void)
 {
 	ffa_devices_unregister();
 	bus_unregister(&ffa_bus_type);
 	unregister_chrdev_region(ffa_devt, FFA_MAX_CDEVS);
 }
-
-module_exit(arm_ffa_bus_exit);
-
-MODULE_ALIAS("arm-ffa-bus");
-MODULE_AUTHOR("Sudeep Holla <sudeep.holla@arm.com>");
-MODULE_DESCRIPTION("Arm FF-A bus driver");
-MODULE_LICENSE("GPL v2");
diff --git a/drivers/firmware/arm_ffa/common.h b/drivers/firmware/arm_ffa/common.h
new file mode 100644
index 000000000000..fc6948efe9f3
--- /dev/null
+++ b/drivers/firmware/arm_ffa/common.h
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2020 ARM Ltd.
+ */
+
+#ifndef _FFA_COMMON_H
+#define _FFA_COMMON_H
+
+#include <linux/arm-smccc.h>
+#include <linux/err.h>
+
+typedef struct arm_smccc_v1_2_res ffa_res_t;
+
+typedef ffa_res_t
+(ffa_fn)(unsigned long, unsigned long, unsigned long, unsigned long,
+	 unsigned long, unsigned long, unsigned long, unsigned long);
+
+int __init arm_ffa_bus_init(void);
+void __exit arm_ffa_bus_exit(void);
+
+static inline int __init ffa_transport_init(ffa_fn **invoke_ffa_fn)
+{
+	return -EOPNOTSUPP;
+}
+
+#endif /* _FFA_COMMON_H */
diff --git a/drivers/firmware/arm_ffa/driver.c b/drivers/firmware/arm_ffa/driver.c
new file mode 100644
index 000000000000..1c4a5e5095b5
--- /dev/null
+++ b/drivers/firmware/arm_ffa/driver.c
@@ -0,0 +1,219 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Arm Firmware Framework for ARMv8-A(FFA) interface driver
+ *
+ * The Arm FFA specification[1] describes a software architecture to
+ * leverages the virtualization extension to isolate software images
+ * provided by an ecosystem of vendors from each other and describes
+ * interfaces that standardize communication between the various software
+ * images including communication between images in the Secure world and
+ * Normal world. Any Hypervisor could use the FFA interfaces to enable
+ * communication between VMs it manages.
+ *
+ * The Hypervisor a.k.a Partition managers in FFA terminology can assign
+ * system resources(Memory regions, Devices, CPU cycles) to the partitions
+ * and manage isolation amongst them.
+ *
+ * [1] https://developer.arm.com/docs/den0077/latest
+ *
+ * Copyright (C) 2020 Arm Ltd.
+ */
+
+#define DRIVER_NAME "ARM FF-A"
+#define pr_fmt(fmt) DRIVER_NAME ": " fmt
+
+#include <linux/arm_ffa.h>
+#include <linux/bitfield.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#include "common.h"
+
+#define FFA_DRIVER_VERSION	FFA_VERSION_1_0
+
+#define SENDER_ID_MASK		GENMASK(31, 16)
+#define RECEIVER_ID_MASK	GENMASK(15, 0)
+#define SENDER_ID(x)		(u16)(FIELD_GET(SENDER_ID_MASK, (x)))
+#define RECEIVER_ID(x)		(u16)(FIELD_GET(RECEIVER_ID_MASK, (x)))
+#define PACK_TARGET_INFO(s, r)		\
+	(FIELD_PREP(SENDER_ID_MASK, (s)) | FIELD_PREP(RECEIVER_ID_MASK, (r)))
+
+/**
+ * FF-A specification mentions explicitly about '4K pages'. This should
+ * not be confused with the kernel PAGE_SIZE, which is the translation
+ * granule kernel is configured and may be one among 4K, 16K and 64K.
+ */
+#define FFA_PAGE_SIZE		SZ_4K
+/* Keeping RX TX buffer size as 64K for now */
+#define RXTX_BUFFER_SIZE	SZ_64K
+
+static ffa_fn *invoke_ffa_fn;
+
+static const int ffa_linux_errmap[] = {
+	/* better than switch case as long as return value is continuous */
+	0,		/* FFA_RET_SUCCESS */
+	-EOPNOTSUPP,	/* FFA_RET_NOT_SUPPORTED */
+	-EINVAL,	/* FFA_RET_INVALID_PARAMETERS */
+	-ENOMEM,	/* FFA_RET_NO_MEMORY */
+	-EBUSY,		/* FFA_RET_BUSY */
+	-EINTR,		/* FFA_RET_INTERRUPTED */
+	-EACCES,	/* FFA_RET_DENIED */
+	-EAGAIN,	/* FFA_RET_RETRY */
+	-ECANCELED,	/* FFA_RET_ABORTED */
+};
+
+static inline int ffa_to_linux_errno(int errno)
+{
+	if (errno < FFA_RET_SUCCESS && errno >= -ARRAY_SIZE(ffa_linux_errmap))
+		return ffa_linux_errmap[-errno];
+	return -EINVAL;
+}
+
+struct ffa_drv_info {
+	u32 version;
+	u16 vm_id;
+	struct mutex rx_lock; /* lock to protect Rx buffer */
+	struct mutex tx_lock; /* lock to protect Tx buffer */
+	void *rx_buffer;
+	void *tx_buffer;
+};
+
+static struct ffa_drv_info *drv_info;
+
+static int ffa_version_check(u32 *version)
+{
+	ffa_res_t ver;
+
+	ver = invoke_ffa_fn(FFA_VERSION, FFA_DRIVER_VERSION, 0, 0, 0, 0, 0, 0);
+
+	if (ver.a0 == FFA_RET_NOT_SUPPORTED) {
+		pr_info("FFA_VERSION returned not supported\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (ver.a0 < FFA_MIN_VERSION || ver.a0 > FFA_DRIVER_VERSION) {
+		pr_err("Incompatible version %d.%d found\n",
+		       MAJOR_VERSION(ver.a0), MINOR_VERSION(ver.a0));
+		return -EINVAL;
+	}
+
+	*version = ver.a0;
+	pr_info("Version %d.%d found\n", MAJOR_VERSION(ver.a0),
+		MINOR_VERSION(ver.a0));
+	return 0;
+}
+
+static int ffa_rxtx_map(phys_addr_t tx_buf, phys_addr_t rx_buf, u32 pg_cnt)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_RXTX_MAP, tx_buf, rx_buf, pg_cnt, 0, 0, 0, 0);
+
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	return 0;
+}
+
+static int ffa_rxtx_unmap(u16 vm_id)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_RXTX_UNMAP, vm_id, 0, 0, 0, 0, 0, 0);
+
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	return 0;
+}
+
+#define VM_ID_MASK	GENMASK(15, 0)
+static int ffa_id_get(u16 *vm_id)
+{
+	ffa_res_t id;
+
+	id = invoke_ffa_fn(FFA_ID_GET, 0, 0, 0, 0, 0, 0, 0);
+
+	if (id.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)id.a2);
+
+	*vm_id = FIELD_GET(VM_ID_MASK, (id.a2));
+
+	return 0;
+}
+
+static int __init ffa_init(void)
+{
+	int ret;
+
+	ret = arm_ffa_bus_init();
+	if (ret)
+		return ret;
+
+	ret = ffa_transport_init(&invoke_ffa_fn);
+	if (ret)
+		return ret;
+
+	drv_info = kzalloc(sizeof(*drv_info), GFP_KERNEL);
+	if (!drv_info)
+		return -ENOMEM;
+
+	ret = ffa_version_check(&drv_info->version);
+	if (ret)
+		goto free_drv_info;
+
+	if (ffa_id_get(&drv_info->vm_id)) {
+		pr_err("failed to obtain VM id for self\n");
+		ret = -ENODEV;
+		goto free_drv_info;
+	}
+
+	drv_info->rx_buffer = alloc_pages_exact(RXTX_BUFFER_SIZE, GFP_KERNEL);
+	if (!drv_info->rx_buffer) {
+		ret = -ENOMEM;
+		goto free_pages;
+	}
+
+	drv_info->tx_buffer = alloc_pages_exact(RXTX_BUFFER_SIZE, GFP_KERNEL);
+	if (!drv_info->tx_buffer) {
+		ret = -ENOMEM;
+		goto free_pages;
+	}
+
+	ret = ffa_rxtx_map(virt_to_phys(drv_info->tx_buffer),
+			   virt_to_phys(drv_info->rx_buffer),
+			   RXTX_BUFFER_SIZE / FFA_PAGE_SIZE);
+	if (ret) {
+		pr_err("failed to register FFA RxTx buffers\n");
+		goto free_pages;
+	}
+
+	mutex_init(&drv_info->rx_lock);
+	mutex_init(&drv_info->tx_lock);
+
+	return 0;
+free_pages:
+	if (drv_info->tx_buffer)
+		free_pages_exact(drv_info->tx_buffer, RXTX_BUFFER_SIZE);
+	free_pages_exact(drv_info->rx_buffer, RXTX_BUFFER_SIZE);
+free_drv_info:
+	kfree(drv_info);
+	return ret;
+}
+module_init(ffa_init);
+
+static void __exit ffa_exit(void)
+{
+	ffa_rxtx_unmap(drv_info->vm_id);
+	free_pages_exact(drv_info->tx_buffer, RXTX_BUFFER_SIZE);
+	free_pages_exact(drv_info->rx_buffer, RXTX_BUFFER_SIZE);
+	kfree(drv_info);
+	arm_ffa_bus_exit();
+}
+module_exit(ffa_exit);
+
+MODULE_ALIAS("arm-ffa");
+MODULE_AUTHOR("Sudeep Holla <sudeep.holla@arm.com>");
+MODULE_DESCRIPTION("Arm FF-A interface driver");
+MODULE_LICENSE("GPL v2");
diff --git a/include/linux/arm_ffa.h b/include/linux/arm_ffa.h
index 2fe16176149f..6185d0d12f15 100644
--- a/include/linux/arm_ffa.h
+++ b/include/linux/arm_ffa.h
@@ -12,6 +12,83 @@
 #include <linux/types.h>
 #include <linux/uuid.h>
 
+#define FFA_SMC(calling_convention, func_num)				\
+	ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL, (calling_convention),	\
+			   ARM_SMCCC_OWNER_STANDARD, (func_num))
+
+#define FFA_SMC_32(func_num)	FFA_SMC(ARM_SMCCC_SMC_32, (func_num))
+#define FFA_SMC_64(func_num)	FFA_SMC(ARM_SMCCC_SMC_64, (func_num))
+
+#define FFA_ERROR			FFA_SMC_32(0x60)
+#define FFA_SUCCESS			FFA_SMC_32(0x61)
+#define FFA_INTERRUPT			FFA_SMC_32(0x62)
+#define FFA_VERSION			FFA_SMC_32(0x63)
+#define FFA_FEATURES			FFA_SMC_32(0x64)
+#define FFA_RX_RELEASE			FFA_SMC_32(0x65)
+#define FFA_RXTX_MAP			FFA_SMC_32(0x66)
+#define FFA_RXTX_UNMAP			FFA_SMC_32(0x67)
+#define FFA_PARTITION_INFO_GET		FFA_SMC_32(0x68)
+#define FFA_ID_GET			FFA_SMC_32(0x69)
+#define FFA_MSG_POLL			FFA_SMC_32(0x6A)
+#define FFA_MSG_WAIT			FFA_SMC_32(0x6B)
+#define FFA_YIELD			FFA_SMC_32(0x6C)
+#define FFA_RUN				FFA_SMC_32(0x6D)
+#define FFA_MSG_SEND			FFA_SMC_32(0x6E)
+#define FFA_MSG_SEND_DIRECT_REQ		FFA_SMC_32(0x6F)
+#define FFA_FN64_MSG_SEND_DIRECT_REQ	FFA_SMC_64(0x6F)
+#define FFA_MSG_SEND_DIRECT_RESP	FFA_SMC_32(0x70)
+#define FFA_FN64_MSG_SEND_DIRECT_RESP	FFA_SMC_64(0x70)
+#define FFA_MEM_DONATE			FFA_SMC_32(0x71)
+#define FFA_FN64_MEM_DONATE		FFA_SMC_32(0x71)
+#define FFA_MEM_LEND			FFA_SMC_32(0x72)
+#define FFA_FN64_MEM_LEND		FFA_SMC_32(0x72)
+#define FFA_MEM_SHARE			FFA_SMC_32(0x73)
+#define FFA_FN64_MEM_SHARE		FFA_SMC_64(0x73)
+#define FFA_MEM_RETRIEVE_REQ		FFA_SMC_32(0x74)
+#define FFA_FN64_MEM_RETRIEVE_REQ	FFA_SMC_64(0x74)
+#define FFA_MEM_RETRIEVE_RESP		FFA_SMC_32(0x75)
+#define FFA_MEM_RELINQUISH		FFA_SMC_32(0x76)
+#define FFA_MEM_RECLAIM			FFA_SMC_32(0x77)
+#define FFA_MEM_OP_PAUSE		FFA_SMC_32(0x78)
+#define FFA_MEM_OP_RESUME		FFA_SMC_32(0x79)
+#define FFA_MEM_FRAG_RX			FFA_SMC_32(0x7A)
+#define FFA_MEM_FRAG_TX			FFA_SMC_32(0x7B)
+#define FFA_NORMAL_WORLD_RESUME		FFA_SMC_32(0x7C)
+
+/*
+ * For some calls it is necessary to use SMC64 to pass or return 64-bit values.
+ * For such calls FFA_FN_NATIVE(name) will choose the appropriate
+ * (native-width) function ID.
+ */
+#ifdef CONFIG_64BIT
+#define FFA_FN_NATIVE(name)	FFA_FN64_##name
+#else
+#define FFA_FN_NATIVE(name)	FFA_##name
+#endif
+
+/* FFA error codes. */
+#define FFA_RET_SUCCESS            (0)
+#define FFA_RET_NOT_SUPPORTED      (-1)
+#define FFA_RET_INVALID_PARAMETERS (-2)
+#define FFA_RET_NO_MEMORY          (-3)
+#define FFA_RET_BUSY               (-4)
+#define FFA_RET_INTERRUPTED        (-5)
+#define FFA_RET_DENIED             (-6)
+#define FFA_RET_RETRY              (-7)
+#define FFA_RET_ABORTED            (-8)
+
+#define MAJOR_VERSION_MASK	GENMASK(30, 16)
+#define MINOR_VERSION_MASK	GENMASK(15, 0)
+#define MAJOR_VERSION(x)	(u16)(FIELD_GET(MAJOR_VERSION_MASK, (x)))
+#define MINOR_VERSION(x)	(u16)(FIELD_GET(MINOR_VERSION_MASK, (x)))
+#define PACK_VERSION_INFO(major, minor)			\
+	(FIELD_PREP(MAJOR_VERSION_MASK, (major)) |	\
+	 FIELD_PREP(MINOR_VERSION_MASK, (minor)))
+#define FFA_VERSION_1_0		PACK_VERSION_INFO(1, 0)
+#define FFA_MIN_VERSION		FFA_VERSION_1_0
+
+/* FFA Device/Bus related */
+
 struct ffa_device {
 	u32 vm_id;
 	uuid_t uuid;
-- 
2.25.1

[PATCH v2 6/9] firmware: arm_ffa: Add support for SMCCC as transport to FFA driver

[Sudeep Holla]

There are requests to keep the transport separate in order to allow
other possible transports like virtio. So let us keep the SMCCC transport
specific routines abstracted.

It is kept simple for now. Once we add another transport, we can develop
better abstraction.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/arm_ffa/Kconfig  |  5 +++
 drivers/firmware/arm_ffa/Makefile |  3 +-
 drivers/firmware/arm_ffa/common.h |  4 +++
 drivers/firmware/arm_ffa/smccc.c  | 54 +++++++++++++++++++++++++++++++
 4 files changed, 65 insertions(+), 1 deletion(-)
 create mode 100644 drivers/firmware/arm_ffa/smccc.c

diff --git a/drivers/firmware/arm_ffa/Kconfig b/drivers/firmware/arm_ffa/Kconfig
index 261a3660650a..5e3ae5cf82e8 100644
--- a/drivers/firmware/arm_ffa/Kconfig
+++ b/drivers/firmware/arm_ffa/Kconfig
@@ -14,3 +14,8 @@ config ARM_FFA_TRANSPORT
 
 	  This driver provides interface for all the client drivers making
 	  use of the features offered by ARM FF-A.
+
+config ARM_FFA_SMCCC
+	bool
+	default ARM_FFA_TRANSPORT
+	depends on ARM64 && HAVE_ARM_SMCCC_DISCOVERY
diff --git a/drivers/firmware/arm_ffa/Makefile b/drivers/firmware/arm_ffa/Makefile
index 82d0d35c5324..9d9f37523200 100644
--- a/drivers/firmware/arm_ffa/Makefile
+++ b/drivers/firmware/arm_ffa/Makefile
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
 ffa-bus-y = bus.o
 ffa-driver-y = driver.o
-ffa-module-objs := $(ffa-bus-y) $(ffa-driver-y)
+ffa-transport-$(CONFIG_ARM_FFA_SMCCC) += smccc.o
+ffa-module-objs := $(ffa-bus-y) $(ffa-driver-y) $(ffa-transport-y)
 obj-$(CONFIG_ARM_FFA_TRANSPORT) = ffa-module.o
diff --git a/drivers/firmware/arm_ffa/common.h b/drivers/firmware/arm_ffa/common.h
index fc6948efe9f3..d019348bf67d 100644
--- a/drivers/firmware/arm_ffa/common.h
+++ b/drivers/firmware/arm_ffa/common.h
@@ -18,9 +18,13 @@ typedef ffa_res_t
 int __init arm_ffa_bus_init(void);
 void __exit arm_ffa_bus_exit(void);
 
+#ifdef CONFIG_ARM_FFA_SMCCC
+int __init ffa_transport_init(ffa_fn **invoke_ffa_fn);
+#else
 static inline int __init ffa_transport_init(ffa_fn **invoke_ffa_fn)
 {
 	return -EOPNOTSUPP;
 }
+#endif
 
 #endif /* _FFA_COMMON_H */
diff --git a/drivers/firmware/arm_ffa/smccc.c b/drivers/firmware/arm_ffa/smccc.c
new file mode 100644
index 000000000000..b93d281d2399
--- /dev/null
+++ b/drivers/firmware/arm_ffa/smccc.c
@@ -0,0 +1,54 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2020 ARM Ltd.
+ */
+
+#include <linux/printk.h>
+
+#include "common.h"
+
+static struct arm_smccc_v1_2_res
+__arm_ffa_fn_smc(unsigned long function_id, unsigned long arg0,
+		 unsigned long arg1, unsigned long arg2, unsigned long arg3,
+		 unsigned long arg4, unsigned long arg5, unsigned long arg6)
+{
+	struct arm_smccc_v1_2_res res;
+
+	arm_smccc_v1_2_smc(function_id, arg0, arg1, arg2, arg3, arg4, arg5,
+			   arg6, &res);
+
+	return res;
+}
+
+static struct arm_smccc_v1_2_res
+__arm_ffa_fn_hvc(unsigned long function_id, unsigned long arg0,
+		 unsigned long arg1, unsigned long arg2, unsigned long arg3,
+		 unsigned long arg4, unsigned long arg5, unsigned long arg6)
+{
+	struct arm_smccc_v1_2_res res;
+
+	arm_smccc_v1_2_hvc(function_id, arg0, arg1, arg2, arg3, arg4, arg5,
+			   arg6, &res);
+	return res;
+}
+
+int __init ffa_transport_init(ffa_fn **invoke_ffa_fn)
+{
+	enum arm_smccc_conduit conduit;
+
+	if (arm_smccc_get_version() < ARM_SMCCC_VERSION_1_2)
+		return -EOPNOTSUPP;
+
+	conduit = arm_smccc_1_1_get_conduit();
+	if (conduit == SMCCC_CONDUIT_NONE) {
+		pr_err("%s: invalid SMCCC conduit\n", __func__);
+		return -EOPNOTSUPP;
+	}
+
+	if (conduit == SMCCC_CONDUIT_SMC)
+		*invoke_ffa_fn = __arm_ffa_fn_smc;
+	else
+		*invoke_ffa_fn = __arm_ffa_fn_hvc;
+
+	return 0;
+}
-- 
2.25.1

[PATCH v2 7/9] firmware: arm_ffa: Setup in-kernel users of FFA partitions

[Sudeep Holla]

Parse the FFA nodes from the device-tree and register all the partitions
whose services will be used in the kernel.

In order to also enable in-kernel users of FFA interface, let us add
simple set of operations for such devices.

The in-kernel users are registered without the character device interface.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/arm_ffa/bus.c    |   2 +
 drivers/firmware/arm_ffa/driver.c | 281 ++++++++++++++++++++++++++++++
 include/linux/arm_ffa.h           |  15 ++
 include/uapi/linux/arm_ffa.h      |  56 ++++++
 4 files changed, 354 insertions(+)
 create mode 100644 include/uapi/linux/arm_ffa.h

diff --git a/drivers/firmware/arm_ffa/bus.c b/drivers/firmware/arm_ffa/bus.c
index 2dbcf356bd8b..de4f8b54271a 100644
--- a/drivers/firmware/arm_ffa/bus.c
+++ b/drivers/firmware/arm_ffa/bus.c
@@ -90,6 +90,7 @@ static void ffa_release_device(struct device *dev)
 {
 	struct ffa_device *ffa_dev = to_ffa_dev(dev);
 
+	mutex_destroy(&ffa_dev->mutex);
 	kfree(ffa_dev);
 }
 
@@ -127,6 +128,7 @@ int ffa_device_register(struct ffa_device *ffa_dev)
 
 	dev = &ffa_dev->dev;
 	cdev = &ffa_dev->cdev;
+	mutex_init(&ffa_dev->mutex);
 
 	dev->bus = &ffa_bus_type;
 	dev->type = &ffa_dev_type;
diff --git a/drivers/firmware/arm_ffa/driver.c b/drivers/firmware/arm_ffa/driver.c
index 1c4a5e5095b5..2e5fa56fffb7 100644
--- a/drivers/firmware/arm_ffa/driver.c
+++ b/drivers/firmware/arm_ffa/driver.c
@@ -24,9 +24,16 @@
 
 #include <linux/arm_ffa.h>
 #include <linux/bitfield.h>
+#include <linux/device.h>
+#include <linux/fs.h>
 #include <linux/io.h>
+#include <linux/kernel.h>
 #include <linux/module.h>
+#include <linux/of.h>
 #include <linux/slab.h>
+#include <linux/uaccess.h>
+#include <linux/uuid.h>
+#include <uapi/linux/arm_ffa.h>
 
 #include "common.h"
 
@@ -48,6 +55,13 @@
 /* Keeping RX TX buffer size as 64K for now */
 #define RXTX_BUFFER_SIZE	SZ_64K
 
+#define list_to_ffa_dev(n)	container_of(n, struct ffa_device, node)
+
+/* List of all FFA devices active in system */
+static LIST_HEAD(ffa_devs_list);
+/* Protection for the entire list */
+static DEFINE_MUTEX(ffa_devs_list_mutex);
+
 static ffa_fn *invoke_ffa_fn;
 
 static const int ffa_linux_errmap[] = {
@@ -104,6 +118,20 @@ static int ffa_version_check(u32 *version)
 	return 0;
 }
 
+static int ffa_rx_release(void)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_RX_RELEASE, 0, 0, 0, 0, 0, 0, 0);
+
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	/* check for ret.a0 == FFA_RX_RELEASE ? */
+
+	return 0;
+}
+
 static int ffa_rxtx_map(phys_addr_t tx_buf, phys_addr_t rx_buf, u32 pg_cnt)
 {
 	ffa_res_t ret;
@@ -128,6 +156,50 @@ static int ffa_rxtx_unmap(u16 vm_id)
 	return 0;
 }
 
+static int ffa_partition_info_get(u32 uuid0, u32 uuid1, u32 uuid2, u32 uuid3,
+				  struct ffa_partition_info **buffer)
+{
+	int count;
+	ffa_res_t partition_info;
+
+	mutex_lock(&drv_info->rx_lock);
+	partition_info = invoke_ffa_fn(FFA_PARTITION_INFO_GET, uuid0, uuid1,
+				       uuid2, uuid3, 0, 0, 0);
+
+	if (partition_info.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)partition_info.a2);
+
+	count = partition_info.a2;
+
+	if (buffer)
+		memcpy(*buffer, drv_info->rx_buffer, sizeof(*buffer) * count);
+
+	ffa_rx_release();
+
+	mutex_unlock(&drv_info->rx_lock);
+
+	return count;
+}
+
+static int ffa_partition_probe(const char *uuid_str,
+			       struct ffa_partition_info *buffer)
+{
+	int count;
+	uuid_t uuid;
+	u32 uuid0_4[4] = { 0 };
+
+	if (uuid_parse(uuid_str, &uuid)) {
+		pr_err("invalid uuid (%s)\n", uuid_str);
+		return -ENODEV;
+	}
+
+	export_uuid((u8 *)uuid0_4, &uuid);
+	count = ffa_partition_info_get(uuid0_4[0], uuid0_4[1], uuid0_4[2],
+				       uuid0_4[3], &buffer);
+
+	return count;
+}
+
 #define VM_ID_MASK	GENMASK(15, 0)
 static int ffa_id_get(u16 *vm_id)
 {
@@ -143,6 +215,205 @@ static int ffa_id_get(u16 *vm_id)
 	return 0;
 }
 
+static int ffa_msg_send_direct_req(u16 src_id, u16 dst_id,
+				   struct ffa_send_direct_data *data)
+{
+	u32 src_dst_ids = PACK_TARGET_INFO(src_id, dst_id);
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_FN_NATIVE(MSG_SEND_DIRECT_REQ), src_dst_ids, 0,
+			    data->data0, data->data1, data->data2,
+			    data->data3, data->data4);
+
+	while (ret.a0 == FFA_INTERRUPT)
+		ret = invoke_ffa_fn(FFA_RUN, ret.a1, 0, 0, 0, 0, 0, 0);
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	if (ret.a0 == FFA_FN_NATIVE(MSG_SEND_DIRECT_RESP)) {
+		data->data0 = ret.a3;
+		data->data1 = ret.a4;
+		data->data2 = ret.a5;
+		data->data3 = ret.a6;
+		data->data4 = ret.a7;
+	}
+
+	return 0;
+}
+
+static void ffa_device_get(struct ffa_device *ffa_dev)
+{
+	mutex_lock(&ffa_dev->mutex);
+	ffa_dev->num_users++;
+	mutex_unlock(&ffa_dev->mutex);
+}
+
+static void ffa_device_put(struct ffa_device *ffa_dev)
+{
+	mutex_lock(&ffa_dev->mutex);
+	ffa_dev->num_users--;
+	mutex_unlock(&ffa_dev->mutex);
+}
+
+static const struct ffa_dev_ops ffa_ops;
+
+const struct ffa_dev_ops *ffa_dev_ops_get(struct ffa_device *dev)
+{
+	struct list_head *p;
+	const struct ffa_dev_ops *ops = NULL;
+
+	if (uuid_is_null(&dev->uuid))
+		return NULL;
+
+	mutex_lock(&ffa_devs_list_mutex);
+
+	list_for_each(p, &ffa_devs_list) {
+		struct ffa_device *tmp_dev;
+
+		tmp_dev = list_to_ffa_dev(p);
+
+		if (uuid_equal(&tmp_dev->uuid, &dev->uuid)) {
+			ops = &ffa_ops;
+			break;
+		}
+	}
+
+	mutex_unlock(&ffa_devs_list_mutex);
+
+	return ops;
+}
+
+static int ffa_dev_open(struct ffa_device *ffa_dev)
+{
+	ffa_device_get(ffa_dev);
+
+	return 0;
+}
+
+static int ffa_dev_close(struct ffa_device *ffa_dev)
+{
+	ffa_device_put(ffa_dev);
+
+	return 0;
+}
+
+static long
+ffa_dev_ioctl(struct ffa_device *ffa_dev, unsigned int ioctl, void *arg)
+{
+	long r = -EINVAL;
+
+	switch (ioctl) {
+	case FFA_GET_API_VERSION:
+		return drv_info->version;
+	case FFA_GET_PARTITION_ID:
+		return ffa_dev->vm_id;
+	case FFA_GET_PARTITION_INFO: {
+		struct ffa_part_info *pinfo = arg;
+
+		if (ffa_partition_probe(pinfo->uuid_str, &pinfo->info) != 1)
+			r = -E2BIG;
+		break;
+	}
+	case FFA_SEND_RECEIVE_SYNC: {
+		struct ffa_send_recv_sync *kdata = arg;
+
+		r = ffa_msg_send_direct_req(ffa_dev->vm_id, kdata->endpoint_id,
+					    &kdata->data);
+		break;
+	}
+	default:
+		r = -EINVAL;
+	}
+
+	return r;
+}
+
+static const struct ffa_dev_ops ffa_ops = {
+	.open = ffa_dev_open,
+	.close = ffa_dev_close,
+	.ioctl = ffa_dev_ioctl,
+};
+
+static int ffa_device_alloc_register(const char *name, u16 vm_id, uuid_t *uuid,
+				     const struct file_operations *fops)
+{
+	int ret;
+	struct ffa_device *ffa_dev;
+
+	ffa_dev = kzalloc(sizeof(*ffa_dev), GFP_KERNEL);
+	if (!ffa_dev)
+		return -ENOMEM;
+
+	ffa_dev->vm_id = vm_id;
+	if (uuid)
+		uuid_copy(&ffa_dev->uuid, uuid);
+
+	dev_set_name(&ffa_dev->dev, name);
+	dev_set_drvdata(&ffa_dev->dev, drv_info);
+	cdev_init(&ffa_dev->cdev, fops);
+
+	ret = ffa_device_register(ffa_dev);
+	if (ret)
+		return ret;
+
+	mutex_lock(&ffa_devs_list_mutex);
+	list_add_tail(&ffa_dev->node, &ffa_devs_list);
+	mutex_unlock(&ffa_devs_list_mutex);
+
+	return 0;
+}
+
+int ffa_setup_partitions(const char *compatible,
+			 const struct file_operations *fops)
+{
+	int ret;
+	struct ffa_partition_info pbuf;
+	struct device_node *child, *parent;
+	const char *p_uuid, *pfx = "Ignoring FFA partition";
+	uuid_t uuid = UUID_INIT(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+
+	parent = of_find_compatible_node(NULL, NULL, compatible);
+	if (!parent)
+		return 0;
+
+	for_each_child_of_node(parent, child) {
+		if (!of_device_is_compatible(child, "arm,ffa-1.0-partition")) {
+			of_node_put(child);
+			continue;
+		}
+
+		if (of_property_read_string(child, "uuid", &p_uuid)) {
+			pr_err("%s: failed to parse \"uuid\" property\n", pfx);
+			of_node_put(child);
+			continue;
+		}
+
+		of_node_put(child);
+
+		if (uuid_parse(p_uuid, &uuid)) {
+			pr_err("%s: invalid \"uuid\" property (%s)\n",
+			       pfx, p_uuid);
+			continue;
+		}
+
+		ret = ffa_partition_probe(p_uuid, &pbuf);
+		if (ret != 1) {
+			pr_err("%s: %s partition info probe failed\n",
+			       pfx, p_uuid);
+			return -EINVAL;
+		}
+
+		ret = ffa_device_alloc_register(p_uuid, pbuf.id, &uuid, fops);
+		if (ret) {
+			pr_err("%s: failed to register %s\n", pfx, p_uuid);
+			continue;
+		}
+	}
+
+	of_node_put(parent);
+	return 0;
+}
+
 static int __init ffa_init(void)
 {
 	int ret;
@@ -192,6 +463,9 @@ static int __init ffa_init(void)
 	mutex_init(&drv_info->rx_lock);
 	mutex_init(&drv_info->tx_lock);
 
+	/* Set up all the partitions which have in-kernel drivers */
+	ffa_setup_partitions("arm,ffa-1.0", NULL);
+
 	return 0;
 free_pages:
 	if (drv_info->tx_buffer)
@@ -205,6 +479,13 @@ module_init(ffa_init);
 
 static void __exit ffa_exit(void)
 {
+	struct list_head *p;
+
+	mutex_lock(&ffa_devs_list_mutex);
+	list_for_each(p, &ffa_devs_list)
+		ffa_device_unregister(list_to_ffa_dev(p));
+	mutex_unlock(&ffa_devs_list_mutex);
+
 	ffa_rxtx_unmap(drv_info->vm_id);
 	free_pages_exact(drv_info->tx_buffer, RXTX_BUFFER_SIZE);
 	free_pages_exact(drv_info->rx_buffer, RXTX_BUFFER_SIZE);
diff --git a/include/linux/arm_ffa.h b/include/linux/arm_ffa.h
index 6185d0d12f15..719ef02fe42d 100644
--- a/include/linux/arm_ffa.h
+++ b/include/linux/arm_ffa.h
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/types.h>
 #include <linux/uuid.h>
+#include <uapi/linux/arm_ffa.h>
 
 #define FFA_SMC(calling_convention, func_num)				\
 	ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL, (calling_convention),	\
@@ -94,6 +95,9 @@ struct ffa_device {
 	uuid_t uuid;
 	struct device dev;
 	struct cdev cdev;
+	size_t num_users;
+	struct mutex mutex;	/* protects num_users */
+	struct list_head node;
 };
 
 #define to_ffa_dev(d) container_of(d, struct ffa_device, dev)
@@ -113,12 +117,19 @@ struct ffa_driver {
 
 #define to_ffa_driver(d) container_of(d, struct ffa_driver, driver)
 
+struct ffa_dev_ops {
+	int (*open)(struct ffa_device *dev);
+	int (*close)(struct ffa_device *dev);
+	long (*ioctl)(struct ffa_device *dev, unsigned int ioctl, void *arg);
+};
+
 #if IS_REACHABLE(CONFIG_ARM_FFA_TRANSPORT)
 int ffa_device_register(struct ffa_device *ffa_dev);
 void ffa_device_unregister(struct ffa_device *ffa_dev);
 int ffa_driver_register(struct ffa_driver *driver, struct module *owner,
 			const char *mod_name);
 void ffa_driver_unregister(struct ffa_driver *driver);
+const struct ffa_dev_ops *ffa_dev_ops_get(struct ffa_device *dev);
 
 #else
 static inline int ffa_device_register(struct ffa_device *ffa_dev)
@@ -137,6 +148,10 @@ ffa_driver_register(struct ffa_driver *driver, struct module *owner,
 
 static inline void ffa_driver_unregister(struct ffa_driver *driver) {}
 
+const struct ffa_dev_ops *ffa_dev_ops_get(struct ffa_device *dev)
+{
+	return NULL;
+}
 #endif /* CONFIG_ARM_FFA_TRANSPORT */
 
 #define ffa_register(driver) \
diff --git a/include/uapi/linux/arm_ffa.h b/include/uapi/linux/arm_ffa.h
new file mode 100644
index 000000000000..88ddddb4742f
--- /dev/null
+++ b/include/uapi/linux/arm_ffa.h
@@ -0,0 +1,56 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/*
+ * Copyright (C) 2020 Arm Ltd.
+ */
+
+#ifndef _UAPI_LINUX_ARM_FFA_H
+#define _UAPI_LINUX_ARM_FFA_H
+
+#include <linux/types.h>
+#include <linux/ioctl.h>
+
+#define FFA_BASE	0xFF
+
+struct ffa_partition_info {
+	__u16 id;
+	__u16 exec_ctxt;
+/* partition supports receipt of direct requests */
+#define FFA_PARTITION_DIRECT_RECV	BIT(0)
+/* partition can send direct requests. */
+#define FFA_PARTITION_DIRECT_SEND	BIT(1)
+/* partition can send and receive indirect messages. */
+#define FFA_PARTITION_INDIRECT_MSG	BIT(2)
+	__u32 properties;
+};
+
+struct ffa_part_info {
+	char uuid_str[36];
+	struct ffa_partition_info info;
+};
+
+struct ffa_send_direct_data {
+	unsigned long data0;
+	unsigned long data1;
+	unsigned long data2;
+	unsigned long data3;
+	unsigned long data4;
+};
+
+struct ffa_send_recv_sync {
+	__u16 endpoint_id;
+	struct ffa_send_direct_data data;
+};
+
+struct ffa_send_recv_async {
+	__u16 endpoint_id;
+	int length;
+	char buffer[];
+};
+
+#define FFA_GET_API_VERSION	_IO(FFA_BASE, 0x00)
+#define FFA_GET_PARTITION_ID	_IO(FFA_BASE, 0x01)
+#define FFA_GET_PARTITION_INFO	_IOWR(FFA_BASE, 0x02, struct ffa_part_info)
+#define FFA_SEND_RECEIVE_SYNC	_IOWR(FFA_BASE, 0x03, struct ffa_send_recv_sync)
+#define FFA_SEND_RECEIVE_ASYNC	_IOW(FFA_BASE, 0x04, struct ffa_send_recv_async)
+
+#endif /*_UAPI_LINUX_ARM_FFA_H*/
-- 
2.25.1

[PATCH v2 8/9] firmware: arm_ffa: Setup and register all the KVM managed partitions

[Sudeep Holla]

Parse the FFA nodes from the device-tree and register all the partitions
managed by the KVM hypervisor.

All the partitions including the host(self) are registered as the
character device with a set of file operations. Most of the interface
will concentrated in the ioctl.

For now, we have a tiny set of initial ioctls implemented.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/arm_ffa/Makefile         |   2 +-
 drivers/firmware/arm_ffa/common.h         |   5 +
 drivers/firmware/arm_ffa/driver.c         |  27 +++++
 drivers/firmware/arm_ffa/hyp_partitions.c | 132 ++++++++++++++++++++++
 4 files changed, 165 insertions(+), 1 deletion(-)
 create mode 100644 drivers/firmware/arm_ffa/hyp_partitions.c

diff --git a/drivers/firmware/arm_ffa/Makefile b/drivers/firmware/arm_ffa/Makefile
index 9d9f37523200..eac280c27cac 100644
--- a/drivers/firmware/arm_ffa/Makefile
+++ b/drivers/firmware/arm_ffa/Makefile
@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
 ffa-bus-y = bus.o
-ffa-driver-y = driver.o
+ffa-driver-y = driver.o hyp_partitions.o
 ffa-transport-$(CONFIG_ARM_FFA_SMCCC) += smccc.o
 ffa-module-objs := $(ffa-bus-y) $(ffa-driver-y) $(ffa-transport-y)
 obj-$(CONFIG_ARM_FFA_TRANSPORT) = ffa-module.o
diff --git a/drivers/firmware/arm_ffa/common.h b/drivers/firmware/arm_ffa/common.h
index d019348bf67d..12bb960deb09 100644
--- a/drivers/firmware/arm_ffa/common.h
+++ b/drivers/firmware/arm_ffa/common.h
@@ -8,6 +8,7 @@
 
 #include <linux/arm-smccc.h>
 #include <linux/err.h>
+#include <linux/fs.h>
 
 typedef struct arm_smccc_v1_2_res ffa_res_t;
 
@@ -15,8 +16,12 @@ typedef ffa_res_t
 (ffa_fn)(unsigned long, unsigned long, unsigned long, unsigned long,
 	 unsigned long, unsigned long, unsigned long, unsigned long);
 
+int __init ffa_hyp_partitions_init(void);
 int __init arm_ffa_bus_init(void);
 void __exit arm_ffa_bus_exit(void);
+struct ffa_device *ffa_device_get_from_minor(int minor);
+int ffa_setup_partitions(const char *compatible,
+			 const struct file_operations *fops);
 
 #ifdef CONFIG_ARM_FFA_SMCCC
 int __init ffa_transport_init(ffa_fn **invoke_ffa_fn);
diff --git a/drivers/firmware/arm_ffa/driver.c b/drivers/firmware/arm_ffa/driver.c
index 2e5fa56fffb7..0aa3e4d02896 100644
--- a/drivers/firmware/arm_ffa/driver.c
+++ b/drivers/firmware/arm_ffa/driver.c
@@ -241,6 +241,31 @@ static int ffa_msg_send_direct_req(u16 src_id, u16 dst_id,
 	return 0;
 }
 
+struct ffa_device *ffa_device_get_from_minor(int minor)
+{
+	struct list_head *p;
+	struct ffa_device *ffa_dev = NULL;
+
+	mutex_lock(&ffa_devs_list_mutex);
+
+	list_for_each(p, &ffa_devs_list) {
+		struct device *dev;
+		struct ffa_device *tmp_dev;
+
+		tmp_dev = list_to_ffa_dev(p);
+		dev = &tmp_dev->dev;
+
+		if (minor == MINOR(dev->devt)) {
+			ffa_dev = tmp_dev;
+			break;
+		}
+	}
+
+	mutex_unlock(&ffa_devs_list_mutex);
+
+	return ffa_dev;
+}
+
 static void ffa_device_get(struct ffa_device *ffa_dev)
 {
 	mutex_lock(&ffa_dev->mutex);
@@ -466,6 +491,8 @@ static int __init ffa_init(void)
 	/* Set up all the partitions which have in-kernel drivers */
 	ffa_setup_partitions("arm,ffa-1.0", NULL);
 
+	ffa_hyp_partitions_init();
+
 	return 0;
 free_pages:
 	if (drv_info->tx_buffer)
diff --git a/drivers/firmware/arm_ffa/hyp_partitions.c b/drivers/firmware/arm_ffa/hyp_partitions.c
new file mode 100644
index 000000000000..985f66d4322b
--- /dev/null
+++ b/drivers/firmware/arm_ffa/hyp_partitions.c
@@ -0,0 +1,132 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Arm Firmware Framework for ARMv8-A(FFA) userspace interface
+ *
+ * Copyright (C) 2020 Arm Ltd.
+ */
+
+#define pr_fmt(fmt) "ARM FF-A USER : " fmt
+
+#include <linux/arm_ffa.h>
+#include <linux/device.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/scatterlist.h>
+#include <linux/slab.h>
+#include <linux/uaccess.h>
+#include <linux/uuid.h>
+#include <uapi/linux/arm_ffa.h>
+
+#include "common.h"
+
+struct ffa_dev_data {
+	struct ffa_device *dev;
+	const struct ffa_dev_ops *ops;
+};
+
+static int ffa_open(struct inode *inode, struct file *filp)
+{
+	struct ffa_dev_data *ffa_data;
+
+	ffa_data = kzalloc(sizeof(*ffa_data), GFP_KERNEL);
+	if (!ffa_data)
+		return -ENOMEM;
+
+	ffa_data->dev = ffa_device_get_from_minor(iminor(inode));
+	if (!ffa_data->dev) {
+		kfree(ffa_data);
+		return -EINVAL;
+	}
+
+	ffa_data->ops = ffa_dev_ops_get(ffa_data->dev);
+	if (!ffa_data->ops) {
+		kfree(ffa_data);
+		return -EINVAL;
+	}
+
+	ffa_data->ops->open(ffa_data->dev);
+
+	filp->private_data = ffa_data;
+
+	return 0;
+}
+
+static int ffa_release(struct inode *inode, struct file *filp)
+{
+	struct ffa_dev_data *ffa_data = filp->private_data;
+
+	ffa_data->ops->close(ffa_data->dev);
+
+	filp->private_data = NULL;
+	kfree(ffa_data);
+
+	return 0;
+}
+
+static long ffa_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
+{
+	long r = -EINVAL;
+	void __user *argp = (void __user *)arg;
+	struct ffa_dev_data *ffa_data = filp->private_data;
+	struct ffa_device *ffa_dev = ffa_data->dev;
+
+	switch (ioctl) {
+	case FFA_GET_API_VERSION:
+	case FFA_GET_PARTITION_ID:
+		if (arg)
+			return r;
+		return ffa_data->ops->ioctl(ffa_dev, ioctl, NULL);
+	case FFA_GET_PARTITION_INFO: {
+		struct ffa_part_info __user *pinfo = argp;
+		struct ffa_part_info info;
+		unsigned int count;
+
+		r = -EFAULT;
+		if (copy_from_user(&info, pinfo, sizeof(info)))
+			break;
+		count = ffa_data->ops->ioctl(ffa_dev, ioctl, &info);
+		if (count > 1) {
+			r = -E2BIG;
+			break;
+		}
+		r = -EFAULT;
+		if (copy_to_user(pinfo, &info, sizeof(info)))
+			break;
+		r = 0;
+		break;
+	}
+	case FFA_SEND_RECEIVE_SYNC: {
+		struct ffa_send_recv_sync __user *udata = argp;
+		struct ffa_send_recv_sync kdata;
+
+		r = -EFAULT;
+		if (copy_from_user(&kdata, udata, sizeof(kdata)))
+			break;
+		r = ffa_data->ops->ioctl(ffa_dev, ioctl, &kdata);
+		if (r)
+			break;
+		if (copy_to_user(udata, &kdata, sizeof(kdata)))
+			break;
+		break;
+	}
+	default:
+		r = -EINVAL;
+	}
+	return r;
+}
+
+static const struct file_operations ffa_fops = {
+	.owner		= THIS_MODULE,
+	.open		= ffa_open,
+	.release        = ffa_release,
+	.unlocked_ioctl = ffa_ioctl,
+	.llseek		= noop_llseek,
+};
+
+int __init ffa_hyp_partitions_init(void)
+{
+	/* Set up all the partitions that KVM hypervisor maintains */
+	ffa_setup_partitions("arm,ffa-1.0-hypervisor", &ffa_fops);
+
+	return 0;
+}
-- 
2.25.1

[PATCH v2 9/9] firmware: arm_ffa: Add support for MEM_* interfaces

[Sudeep Holla]

Most of the MEM_* APIs share the same parameters, so they can be
generalised. Currently only MEM_SHARE is implemented and the user space
interface for that is not added yet.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
---
 drivers/firmware/arm_ffa/driver.c | 210 ++++++++++++++++++++++++++++++
 include/linux/arm_ffa.h           | 137 +++++++++++++++++++
 include/uapi/linux/arm_ffa.h      |  11 ++
 3 files changed, 358 insertions(+)

diff --git a/drivers/firmware/arm_ffa/driver.c b/drivers/firmware/arm_ffa/driver.c
index 0aa3e4d02896..6e838f2e15cb 100644
--- a/drivers/firmware/arm_ffa/driver.c
+++ b/drivers/firmware/arm_ffa/driver.c
@@ -29,7 +29,9 @@
 #include <linux/io.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
+#include <linux/mm.h>
 #include <linux/of.h>
+#include <linux/scatterlist.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
 #include <linux/uuid.h>
@@ -118,6 +120,18 @@ static int ffa_version_check(u32 *version)
 	return 0;
 }
 
+static int ffa_features(u32 func_id)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_FEATURES, func_id, 0, 0, 0, 0, 0, 0);
+
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	return ret.a2;
+}
+
 static int ffa_rx_release(void)
 {
 	ffa_res_t ret;
@@ -215,6 +229,25 @@ static int ffa_id_get(u16 *vm_id)
 	return 0;
 }
 
+static int ffa_msg_send(u16 src_id, u16 dst_id, void *buf, u32 len)
+{
+	ffa_res_t ret;
+	u32 src_dst_ids = PACK_TARGET_INFO(src_id, dst_id);
+
+	mutex_lock(&drv_info->tx_lock);
+
+	memcpy(drv_info->tx_buffer, buf, len);
+
+	ret = invoke_ffa_fn(FFA_MSG_SEND, src_dst_ids, 0, len, 0, 0, 0, 0);
+
+	mutex_unlock(&drv_info->tx_lock);
+
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	return 0;
+}
+
 static int ffa_msg_send_direct_req(u16 src_id, u16 dst_id,
 				   struct ffa_send_direct_data *data)
 {
@@ -241,6 +274,177 @@ static int ffa_msg_send_direct_req(u16 src_id, u16 dst_id,
 	return 0;
 }
 
+static int ffa_mem_first_frag(u32 func_id, phys_addr_t buf, u32 buf_sz,
+			      u32 frag_len, u32 len, u64 *handle)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(func_id, len, frag_len, buf, buf_sz, 0, 0, 0);
+
+	while (ret.a0 == FFA_MEM_OP_PAUSE)
+		ret = invoke_ffa_fn(FFA_MEM_OP_RESUME, ret.a1, ret.a2,
+				    0, 0, 0, 0, 0);
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	if (ret.a0 != FFA_SUCCESS)
+		return -EOPNOTSUPP;
+
+	if (handle)
+		*handle = PACK_HANDLE(ret.a3, ret.a2);
+
+	return frag_len;
+}
+
+static int ffa_mem_next_frag(u64 handle, u32 frag_len)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_MEM_FRAG_TX, HANDLE_LOW(handle),
+			    HANDLE_HIGH(handle), frag_len, 0, 0, 0, 0);
+
+	while (ret.a0 == FFA_MEM_OP_PAUSE)
+		ret = invoke_ffa_fn(FFA_MEM_OP_RESUME, ret.a1, ret.a2,
+				    0, 0, 0, 0, 0);
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	if (ret.a0 != FFA_MEM_FRAG_RX)
+		return -EOPNOTSUPP;
+
+	return ret.a3;
+}
+
+static int
+ffa_transmit_fragment(u32 func_id, phys_addr_t buf, u32 buf_sz, u32 frag_len,
+		      u32 len, u64 *handle, bool first)
+{
+	if (!first)
+		return ffa_mem_next_frag(*handle, frag_len);
+
+	return ffa_mem_first_frag(func_id, buf, buf_sz, frag_len,
+				      len, handle);
+}
+
+static u32 ffa_get_num_pages_sg(struct scatterlist *sg)
+{
+	u32 num_pages = 0;
+
+	do {
+		num_pages += sg->length / FFA_PAGE_SIZE;
+	} while ((sg = sg_next(sg)));
+
+	return num_pages;
+}
+
+static int
+ffa_setup_and_transmit(u32 func_id, void *buffer, u32 max_fragsize,
+		       struct ffa_mem_ops_args *args)
+{
+	int rc = 0;
+	bool first = true;
+	phys_addr_t addr = 0;
+	struct ffa_composite_mem_region *composite;
+	struct ffa_mem_region_addr_range *constituents;
+	struct ffa_mem_region_attributes *ep_mem_access;
+	struct ffa_mem_region *mem_region = buffer;
+	u32 idx, frag_len, length, num_entries = sg_nents(args->sg);
+	u32 buf_sz = max_fragsize / FFA_PAGE_SIZE;
+
+	mem_region->tag = args->tag;
+	mem_region->flags = args->flags;
+	mem_region->sender_id = drv_info->vm_id;
+	mem_region->attributes = FFA_MEM_NORMAL | FFA_MEM_WRITE_BACK |
+				 FFA_MEM_INNER_SHAREABLE;
+	ep_mem_access = &mem_region->ep_mem_access[0];
+
+	for (idx = 0; idx < args->nattrs; idx++, ep_mem_access++) {
+		ep_mem_access->receiver = args->attrs[idx].receiver;
+		ep_mem_access->attrs = args->attrs[idx].attrs;
+		ep_mem_access->composite_off = COMPOSITE_OFFSET(args->nattrs);
+	}
+	mem_region->ep_count = args->nattrs;
+
+	composite = buffer + COMPOSITE_OFFSET(args->nattrs);
+	composite->total_pg_cnt = ffa_get_num_pages_sg(args->sg);
+	composite->addr_range_cnt = num_entries;
+
+	length = COMPOSITE_CONSTITUENTS_OFFSET(args->nattrs, num_entries);
+	frag_len = COMPOSITE_CONSTITUENTS_OFFSET(args->nattrs, 0);
+	if (frag_len > max_fragsize)
+		return -ENXIO;
+
+	if (!args->use_txbuf)
+		addr = virt_to_phys(buffer);
+
+	constituents = buffer + frag_len;
+	idx = 0;
+	do {
+		if (frag_len == max_fragsize) {
+			rc = ffa_transmit_fragment(func_id, addr, buf_sz,
+						   frag_len, length,
+						   args->g_handle, first);
+			if (rc < 0)
+				return -ENXIO;
+
+			first = false;
+			idx = 0;
+			frag_len = 0;
+			constituents = buffer;
+		}
+
+		if ((void *)constituents - buffer > max_fragsize) {
+			pr_err("Memory Region Fragment > Tx Buffer size\n");
+			return -EFAULT;
+		}
+
+		constituents->address = sg_phys(args->sg);
+		constituents->pg_cnt = args->sg->length / FFA_PAGE_SIZE;
+		constituents++;
+		frag_len += sizeof(struct ffa_mem_region_addr_range);
+	} while ((args->sg = sg_next(args->sg)));
+
+	return ffa_transmit_fragment(func_id, addr, buf_sz, frag_len,
+				     length, args->g_handle, first);
+}
+
+static int ffa_memory_ops(u32 func_id, struct ffa_mem_ops_args *args)
+{
+	int ret;
+	void *buffer;
+
+	if (!args->use_txbuf) {
+		buffer = alloc_pages_exact(RXTX_BUFFER_SIZE, GFP_KERNEL);
+		if (!buffer)
+			return -ENOMEM;
+	} else {
+		buffer = drv_info->tx_buffer;
+		mutex_lock(&drv_info->tx_lock);
+	}
+
+	ret = ffa_setup_and_transmit(func_id, buffer, RXTX_BUFFER_SIZE, args);
+
+	if (args->use_txbuf)
+		mutex_unlock(&drv_info->tx_lock);
+	else
+		free_pages_exact(buffer, RXTX_BUFFER_SIZE);
+
+	return ret < 0 ? ret : 0;
+}
+
+static int ffa_memory_reclaim(u64 g_handle, u32 flags)
+{
+	ffa_res_t ret;
+
+	ret = invoke_ffa_fn(FFA_MEM_RECLAIM, HANDLE_LOW(g_handle),
+			    HANDLE_HIGH(g_handle), flags, 0, 0, 0, 0);
+
+	if (ret.a0 == FFA_ERROR)
+		return ffa_to_linux_errno((int)ret.a2);
+
+	return 0;
+}
+
 struct ffa_device *ffa_device_get_from_minor(int minor)
 {
 	struct list_head *p;
@@ -346,6 +550,12 @@ ffa_dev_ioctl(struct ffa_device *ffa_dev, unsigned int ioctl, void *arg)
 					    &kdata->data);
 		break;
 	}
+	case FFA_SHARE_MEMORY: {
+		struct ffa_mem_ops_args *kdata = arg;
+
+		r = ffa_memory_ops(FFA_MEM_SHARE, kdata);
+		break;
+	}
 	default:
 		r = -EINVAL;
 	}
diff --git a/include/linux/arm_ffa.h b/include/linux/arm_ffa.h
index 719ef02fe42d..12765a7926fc 100644
--- a/include/linux/arm_ffa.h
+++ b/include/linux/arm_ffa.h
@@ -88,6 +88,143 @@
 #define FFA_VERSION_1_0		PACK_VERSION_INFO(1, 0)
 #define FFA_MIN_VERSION		FFA_VERSION_1_0
 
+struct ffa_mem_region_addr_range {
+	/* The base IPA of the constituent memory region, aligned to 4 kiB */
+	u64 address;
+	/* The number of 4 kiB pages in the constituent memory region. */
+	u32 pg_cnt;
+	u32 reserved;
+};
+
+struct ffa_composite_mem_region {
+	/*
+	 * The total number of 4 kiB pages included in this memory region. This
+	 * must be equal to the sum of page counts specified in each
+	 * `struct ffa_mem_region_addr_range`.
+	 */
+	u32 total_pg_cnt;
+	/* The number of constituents included in this memory region range */
+	u32 addr_range_cnt;
+	u64 reserved;
+	/** An array of `addr_range_cnt` memory region constituents. */
+	struct ffa_mem_region_addr_range constituents[];
+};
+
+struct ffa_mem_region_attributes {
+	/* The ID of the VM to which the memory is being given or shared. */
+	u16 receiver;
+	/*
+	 * The permissions with which the memory region should be mapped in the
+	 * receiver's page table.
+	 */
+#define FFA_MEM_EXEC	BIT(3)
+#define FFA_MEM_NO_EXEC	BIT(2)
+#define FFA_MEM_RW		BIT(1)
+#define FFA_MEM_RO		BIT(0)
+	u8 attrs;
+	/*
+	 * Flags used during FFA_MEM_RETRIEVE_REQ and FFA_MEM_RETRIEVE_RESP
+	 * for memory regions with multiple borrowers.
+	 */
+#define FFA_MEM_RETRIEVE_SELF_BORROWER	BIT(0)
+	u8 flag;
+	u32 composite_off;
+	/*
+	 * Offset in bytes from the start of the outer `ffa_memory_region` to
+	 * an `struct ffa_mem_region_addr_range`.
+	 */
+	u64 reserved;
+};
+
+struct ffa_mem_region {
+	/* The ID of the VM/owner which originally sent the memory region */
+	u16 sender_id;
+#define FFA_MEM_NORMAL		BIT(5)
+#define FFA_MEM_DEVICE		BIT(4)
+
+#define FFA_MEM_WRITE_BACK	(3 << 2)
+#define FFA_MEM_NON_CACHEABLE	(1 << 2)
+
+#define FFA_DEV_nGnRnE		(0 << 2)
+#define FFA_DEV_nGnRE		(1 << 2)
+#define FFA_DEV_nGRE		(2 << 2)
+#define FFA_DEV_GRE		(3 << 2)
+
+#define FFA_MEM_NON_SHAREABLE	(0)
+#define FFA_MEM_OUTER_SHAREABLE	(2)
+#define FFA_MEM_INNER_SHAREABLE	(3)
+	u8 attributes;
+	u8 reserved_0;
+/*
+ * Clear memory region contents after unmapping it from the sender and
+ * before mapping it for any receiver.
+ */
+#define FFA_MEM_CLEAR			BIT(0)
+/*
+ * Whether the hypervisor may time slice the memory sharing or retrieval
+ * operation.
+ */
+#define FFA_TIME_SLICE_ENABLE		BIT(1)
+
+/*
+ * Whether the hypervisor should clear the memory region before the receiver
+ * relinquishes it or is aborted.
+ */
+#define FFA_MEM_CLEAR_BEFORE_RELINQUISH	BIT(0)
+/*
+ * Whether the hypervisor should clear the memory region after the receiver
+ * relinquishes it or is aborted.
+ */
+#define FFA_MEM_CLEAR_AFTER_RELINQUISH	BIT(2)
+
+#define FFA_MEM_RETRIEVE_TYPE_IN_RESP	(0 << 3)
+#define FFA_MEM_RETRIEVE_TYPE_SHARE	(1 << 3)
+#define FFA_MEM_RETRIEVE_TYPE_LEND	(2 << 3)
+#define FFA_MEM_RETRIEVE_TYPE_DONATE	(3 << 3)
+
+#define FFA_MEM_RETRIEVE_ADDR_ALIGN_HINT	BIT(9)
+#define FFA_MEM_RETRIEVE_ADDR_ALIGN(x)		((x) << 5)
+	/* Flags to control behaviour of the transaction. */
+	u32 flags;
+#define HANDLE_LOW_MASK		GENMASK_ULL(31, 0)
+#define HANDLE_HIGH_MASK	GENMASK_ULL(63, 32)
+#define HANDLE_LOW(x)		(u32)(FIELD_GET(HANDLE_LOW_MASK, (x)))
+#define	HANDLE_HIGH(x)		(u32)(FIELD_GET(HANDLE_HIGH_MASK, (x)))
+
+#define PACK_HANDLE(l, h)		\
+	(FIELD_PREP(HANDLE_LOW_MASK, (l)) | FIELD_PREP(HANDLE_HIGH_MASK, (h)))
+	/*
+	 * A globally-unique ID assigned by the hypervisor for a region
+	 * of memory being sent between VMs.
+	 */
+	u64 handle;
+	/*
+	 * An implementation defined value associated with the receiver and the
+	 * memory region.
+	 */
+	u64 tag;
+	u32 reserved_1;
+	/*
+	 * The number of `ffa_mem_region_attributes` entries included in this
+	 * transaction.
+	 */
+	u32 ep_count;
+	/*
+	 * An array of endpoint memory access descriptors.
+	 * Each one specifies a memory region offset, an endpoint and the
+	 * attributes with which this memory region should be mapped in that
+	 * endpoint's page table.
+	 */
+	struct ffa_mem_region_attributes ep_mem_access[];
+};
+
+#define	COMPOSITE_OFFSET(x)	\
+	(offsetof(struct ffa_mem_region, ep_mem_access[x]))
+#define CONSTITUENTS_OFFSET(x)	\
+	(offsetof(struct ffa_composite_mem_region, constituents[x]))
+#define COMPOSITE_CONSTITUENTS_OFFSET(x, y)	\
+	(COMPOSITE_OFFSET(x) + CONSTITUENTS_OFFSET(y))
+
 /* FFA Device/Bus related */
 
 struct ffa_device {
diff --git a/include/uapi/linux/arm_ffa.h b/include/uapi/linux/arm_ffa.h
index 88ddddb4742f..be2cbf01e51c 100644
--- a/include/uapi/linux/arm_ffa.h
+++ b/include/uapi/linux/arm_ffa.h
@@ -47,10 +47,21 @@ struct ffa_send_recv_async {
 	char buffer[];
 };
 
+struct ffa_mem_ops_args {
+	bool use_txbuf;
+	u64 tag;
+	u32 flags;
+	struct ffa_mem_region_attributes *attrs;
+	u32 nattrs;
+	struct scatterlist *sg;
+	u64 *g_handle;
+};
+
 #define FFA_GET_API_VERSION	_IO(FFA_BASE, 0x00)
 #define FFA_GET_PARTITION_ID	_IO(FFA_BASE, 0x01)
 #define FFA_GET_PARTITION_INFO	_IOWR(FFA_BASE, 0x02, struct ffa_part_info)
 #define FFA_SEND_RECEIVE_SYNC	_IOWR(FFA_BASE, 0x03, struct ffa_send_recv_sync)
 #define FFA_SEND_RECEIVE_ASYNC	_IOW(FFA_BASE, 0x04, struct ffa_send_recv_async)
+#define FFA_SHARE_MEMORY	_IOW(FFA_BASE, 0x05, struct ffa_mem_ops_args)
 
 #endif /*_UAPI_LINUX_ARM_FFA_H*/
-- 
2.25.1

Re: [PATCH v2 9/9] firmware: arm_ffa: Add support for MEM_* interfaces

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 2120 bytes --]

Hi Sudeep,

I love your patch! Yet something to improve:

[auto build test ERROR on robh/for-next]
[also build test ERROR on arm/for-next keystone/next rockchip/for-next shawnguo/for-next soc/for-next linus/master v5.10-rc2 next-20201103]
[cannot apply to arm64/for-next/core linux-rpi/for-rpi-next at91/at91-next]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Sudeep-Holla/firmware-Add-initial-support-for-Arm-FF-A/20201104-014609
base:   https://git.kernel.org/pub/scm/linux/kernel/git/robh/linux.git for-next
config: x86_64-randconfig-a005-20201103 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0
reproduce (this is a W=1 build):
        # https://github.com/0day-ci/linux/commit/2e8f3c2680d2f574b4e4efcce5dfd75912527386
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Sudeep-Holla/firmware-Add-initial-support-for-Arm-FF-A/20201104-014609
        git checkout 2e8f3c2680d2f574b4e4efcce5dfd75912527386
        # save the attached .config to linux build tree
        make W=1 ARCH=x86_64 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   In file included from <command-line>:32:
>> ./usr/include/linux/arm_ffa.h:51:2: error: unknown type name 'bool'
      51 |  bool use_txbuf;
         |  ^~~~
>> ./usr/include/linux/arm_ffa.h:52:2: error: unknown type name 'u64'
      52 |  u64 tag;
         |  ^~~
>> ./usr/include/linux/arm_ffa.h:53:2: error: unknown type name 'u32'
      53 |  u32 flags;
         |  ^~~
   ./usr/include/linux/arm_ffa.h:55:2: error: unknown type name 'u32'
      55 |  u32 nattrs;
         |  ^~~
   ./usr/include/linux/arm_ffa.h:57:2: error: unknown type name 'u64'
      57 |  u64 *g_handle;
         |  ^~~

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 33681 bytes --]

Re: [PATCH v2 3/9] arm64: smccc: Add support for SMCCCv1.2 input/output registers

[Andrew Walbran]

[-- Attachment #1: Type: text/plain, Size: 5397 bytes --]

On Tue, 3 Nov 2020 at 17:44, Sudeep Holla <sudeep.holla@arm.com> wrote:
>
> SMCCC v1.2 allows x8-x17 to be used as parameter registers and x4—x17
> to be used as result registers in SMC64/HVC64. Arm Firmware Framework
> for Armv8-A specification makes use of x0-x7 as parameter and result
> registers.
>
> Current SMCCC interface in the kernel just use x0-x7 as parameter and
> x0-x3 as result registers. Let us add new interface to support x0-x7
> as parameter and result registers. This can be extended to include
> x8-x17 when there are users for the same.
>
> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
> ---
>  arch/arm64/kernel/asm-offsets.c |  4 +++
>  arch/arm64/kernel/smccc-call.S  | 22 +++++++++++++++
>  include/linux/arm-smccc.h       | 50 +++++++++++++++++++++++++++++++++
>  3 files changed, 76 insertions(+)
>
> diff --git a/arch/arm64/kernel/asm-offsets.c b/arch/arm64/kernel/asm-offsets.c
> index 7d32fc959b1a..32bcc25337ce 100644
> --- a/arch/arm64/kernel/asm-offsets.c
> +++ b/arch/arm64/kernel/asm-offsets.c
> @@ -122,6 +122,10 @@ int main(void)
>    DEFINE(ARM_SMCCC_RES_X2_OFFS,                offsetof(struct arm_smccc_res, a2));
>    DEFINE(ARM_SMCCC_QUIRK_ID_OFFS,      offsetof(struct arm_smccc_quirk, id));
>    DEFINE(ARM_SMCCC_QUIRK_STATE_OFFS,   offsetof(struct arm_smccc_quirk, state));
> +  DEFINE(ARM_SMCCC_V1_2_RES_X0_OFFS,   offsetof(struct arm_smccc_v1_2_res, a0));
> +  DEFINE(ARM_SMCCC_V1_2_RES_X2_OFFS,   offsetof(struct arm_smccc_v1_2_res, a2));
> +  DEFINE(ARM_SMCCC_V1_2_RES_X4_OFFS,   offsetof(struct arm_smccc_v1_2_res, a4));
> +  DEFINE(ARM_SMCCC_V1_2_RES_X6_OFFS,   offsetof(struct arm_smccc_v1_2_res, a6));
>    BLANK();
>    DEFINE(HIBERN_PBE_ORIG,      offsetof(struct pbe, orig_address));
>    DEFINE(HIBERN_PBE_ADDR,      offsetof(struct pbe, address));
> diff --git a/arch/arm64/kernel/smccc-call.S b/arch/arm64/kernel/smccc-call.S
> index d62447964ed9..0ea15c1742f3 100644
> --- a/arch/arm64/kernel/smccc-call.S
> +++ b/arch/arm64/kernel/smccc-call.S
> @@ -43,3 +43,25 @@ SYM_FUNC_START(__arm_smccc_hvc)
>         SMCCC   hvc
>  SYM_FUNC_END(__arm_smccc_hvc)
>  EXPORT_SYMBOL(__arm_smccc_hvc)
> +
> +       .macro SMCCC_v1_2 instr
> +       .cfi_startproc
> +       \instr #0
> +       ldr x8, [sp]
> +       stp x0, x1, [x8, #ARM_SMCCC_V1_2_RES_X0_OFFS]
> +       stp x2, x3, [x8, #ARM_SMCCC_V1_2_RES_X2_OFFS]
> +       stp x4, x5, [x8, #ARM_SMCCC_V1_2_RES_X4_OFFS]
> +       stp x6, x7, [x8, #ARM_SMCCC_V1_2_RES_X6_OFFS]
> +       ret
> +       .cfi_endproc
> +.endm
> +
> +SYM_FUNC_START(arm_smccc_v1_2_hvc)
> +       SMCCC_v1_2 hvc
> +SYM_FUNC_END(arm_smccc_v1_2_hvc)
> +EXPORT_SYMBOL(arm_smccc_v1_2_hvc)
> +
> +SYM_FUNC_START(arm_smccc_v1_2_smc)
> +       SMCCC_v1_2 smc
> +SYM_FUNC_END(arm_smccc_v1_2_smc)
> +EXPORT_SYMBOL(arm_smccc_v1_2_smc)
> diff --git a/include/linux/arm-smccc.h b/include/linux/arm-smccc.h
> index f860645f6512..aed27214c88f 100644
> --- a/include/linux/arm-smccc.h
> +++ b/include/linux/arm-smccc.h
> @@ -155,6 +155,56 @@ struct arm_smccc_res {
>         unsigned long a3;
>  };
>
> +#ifdef CONFIG_ARM64
> +/* TODO Need to implement for ARM too */
> +/**
> + * struct arm_smccc_v1_2_res - Result from SMC/HVC call
> + * @a0-a3 result values from registers 0 to 3

Should this be a0-a7?

> + */
> +struct arm_smccc_v1_2_res {
> +       unsigned long a0;
> +       unsigned long a1;
> +       unsigned long a2;
> +       unsigned long a3;
> +       unsigned long a4;
> +       unsigned long a5;
> +       unsigned long a6;
> +       unsigned long a7;
> +};
> +
> +/**
> + * arm_smccc_v1_2_hvc() - make HVC calls
> + * @a0-a7: arguments passed in registers 0 to 7
> + * @res: result values from registers 0 to 7
> + *
> + * This function is used to make SMC calls following SMC Calling Convention
> + * v1.2 or above. The content of the supplied param are copied to registers
> + * 0 to 7 prior to the SMC instruction. The return values are updated with
> + * the content from register 0 to 7 on return from the SMC instruction.

s/SMC/HVC/ in several places above.

> + */
> +asmlinkage
> +void arm_smccc_v1_2_hvc(unsigned long a0, unsigned long a1, unsigned long a2,
> +                       unsigned long a3, unsigned long a4, unsigned long a5,
> +                       unsigned long a6, unsigned long a7,
> +                       struct arm_smccc_v1_2_res  *res);
> +
> +/**
> + * arm_smccc_v1_2_smc() - make SMC calls
> + * @a0-a7: arguments passed in registers 0 to 7
> + * @res: result values from registers 0 to 7
> + *
> + * This function is used to make SMC calls following SMC Calling Convention
> + * v1.2 or above. The content of the supplied param are copied to registers
> + * 0 to 7 prior to the SMC instruction. The return values are updated with
> + * the content from register 0 to 7 on return from the SMC instruction.
> + */
> +asmlinkage
> +void arm_smccc_v1_2_smc(unsigned long a0, unsigned long a1, unsigned long a2,
> +                       unsigned long a3, unsigned long a4, unsigned long a5,
> +                       unsigned long a6, unsigned long a7,
> +                       struct arm_smccc_v1_2_res  *res);
> +#endif
> +
>  /**
>   * struct arm_smccc_quirk - Contains quirk information
>   * @id: quirk identification
> --
> 2.25.1
>

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 3998 bytes --]

Re: [PATCH v2 0/9] firmware: Add initial support for Arm FF-A

[Jens Wiklander]

Hi Sudeep,

On Tue, Nov 03, 2020 at 05:43:41PM +0000, Sudeep Holla wrote:
> Hi all,
> 
> Let me start stating this is just initial implementation to check on
> the idea of providing more in-kernel and userspace support. Lot of things
> are still work in progress, I am posting just to get the early feedback
> before building lot of things on this idea. Consider this more as RFC
> though not tagged explicity(just to avoid it being ignored :))
> 
> Arm Firmware Framework for Armv8-A specification[1] describes a software
> architecture that provides mechanism to utilise the virtualization
> extension to isolate software images and describes interfaces that
> standardize communication between the various software images. This
> includes communication between images in the Secure and Normal world.
> 
> The main idea here is to create FFA device to establish any communication
> with a partition(secure or normal world VM).
> 
> If it is a partition managed by hypervisor, then we will register chardev
> associated with each of those partition FFA device.
> 
> /dev/arm_ffa:
> 
> e3a48fa5-dc54-4a8b-898b-bdc4dfeeb7b8
> 49f65057-d002-4ae2-b4ee-d31c7940a13d
> 
> For in-kernel usage(mostly communication with secure partitions), only
> in-kernel APIs are accessible(no userspace). There may be a need to
> provide userspace access instead of in-kernel, it is not yet support
> in this series as we need way to identify those and I am not sure if
> that belong to DT.

With unfiltered VM to VM commnication from user space there's no easy
way for two VMs to exchange privileged information that excludes user
space. Perhaps access to the FFA device is considered privileged and
enough for all purposes.

If I've understood it correctly is VM to SP communication only allowed
via kernel mode in the VM. The communication with OP-TEE depends on this
with the recent commit c5b4312bea5d ("tee: optee: Add support for session
login client UUID generation").

Cheers,
Jens

Re: [PATCH v2 7/9] firmware: arm_ffa: Setup in-kernel users of FFA partitions

[Jens Wiklander]

Hi Sudeep,

On Tue, Nov 03, 2020 at 05:43:48PM +0000, Sudeep Holla wrote:
> Parse the FFA nodes from the device-tree and register all the partitions
> whose services will be used in the kernel.
> 
> In order to also enable in-kernel users of FFA interface, let us add
> simple set of operations for such devices.
> 
> The in-kernel users are registered without the character device interface.
> 
> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
> ---
>  drivers/firmware/arm_ffa/bus.c    |   2 +
>  drivers/firmware/arm_ffa/driver.c | 281 ++++++++++++++++++++++++++++++
>  include/linux/arm_ffa.h           |  15 ++
>  include/uapi/linux/arm_ffa.h      |  56 ++++++
>  4 files changed, 354 insertions(+)
>  create mode 100644 include/uapi/linux/arm_ffa.h
> 
> diff --git a/drivers/firmware/arm_ffa/bus.c b/drivers/firmware/arm_ffa/bus.c
> index 2dbcf356bd8b..de4f8b54271a 100644
> --- a/drivers/firmware/arm_ffa/bus.c
> +++ b/drivers/firmware/arm_ffa/bus.c
> @@ -90,6 +90,7 @@ static void ffa_release_device(struct device *dev)
>  {
>  	struct ffa_device *ffa_dev = to_ffa_dev(dev);
>  
> +	mutex_destroy(&ffa_dev->mutex);
>  	kfree(ffa_dev);
>  }
>  
> @@ -127,6 +128,7 @@ int ffa_device_register(struct ffa_device *ffa_dev)
>  
>  	dev = &ffa_dev->dev;
>  	cdev = &ffa_dev->cdev;
> +	mutex_init(&ffa_dev->mutex);
>  
>  	dev->bus = &ffa_bus_type;
>  	dev->type = &ffa_dev_type;
> diff --git a/drivers/firmware/arm_ffa/driver.c b/drivers/firmware/arm_ffa/driver.c
> index 1c4a5e5095b5..2e5fa56fffb7 100644
> --- a/drivers/firmware/arm_ffa/driver.c
> +++ b/drivers/firmware/arm_ffa/driver.c
> @@ -24,9 +24,16 @@
>  
>  #include <linux/arm_ffa.h>
>  #include <linux/bitfield.h>
> +#include <linux/device.h>
> +#include <linux/fs.h>
>  #include <linux/io.h>
> +#include <linux/kernel.h>
>  #include <linux/module.h>
> +#include <linux/of.h>
>  #include <linux/slab.h>
> +#include <linux/uaccess.h>
> +#include <linux/uuid.h>
> +#include <uapi/linux/arm_ffa.h>
>  
>  #include "common.h"
>  
> @@ -48,6 +55,13 @@
>  /* Keeping RX TX buffer size as 64K for now */
>  #define RXTX_BUFFER_SIZE	SZ_64K
>  
> +#define list_to_ffa_dev(n)	container_of(n, struct ffa_device, node)
> +
> +/* List of all FFA devices active in system */
> +static LIST_HEAD(ffa_devs_list);
> +/* Protection for the entire list */
> +static DEFINE_MUTEX(ffa_devs_list_mutex);
> +
>  static ffa_fn *invoke_ffa_fn;
>  
>  static const int ffa_linux_errmap[] = {
> @@ -104,6 +118,20 @@ static int ffa_version_check(u32 *version)
>  	return 0;
>  }
>  
> +static int ffa_rx_release(void)
> +{
> +	ffa_res_t ret;
> +
> +	ret = invoke_ffa_fn(FFA_RX_RELEASE, 0, 0, 0, 0, 0, 0, 0);
> +
> +	if (ret.a0 == FFA_ERROR)
> +		return ffa_to_linux_errno((int)ret.a2);
> +
> +	/* check for ret.a0 == FFA_RX_RELEASE ? */
> +
> +	return 0;
> +}
> +
>  static int ffa_rxtx_map(phys_addr_t tx_buf, phys_addr_t rx_buf, u32 pg_cnt)
>  {
>  	ffa_res_t ret;
> @@ -128,6 +156,50 @@ static int ffa_rxtx_unmap(u16 vm_id)
>  	return 0;
>  }
>  
> +static int ffa_partition_info_get(u32 uuid0, u32 uuid1, u32 uuid2, u32 uuid3,
> +				  struct ffa_partition_info **buffer)
> +{
> +	int count;
> +	ffa_res_t partition_info;
> +
> +	mutex_lock(&drv_info->rx_lock);
> +	partition_info = invoke_ffa_fn(FFA_PARTITION_INFO_GET, uuid0, uuid1,
> +				       uuid2, uuid3, 0, 0, 0);
> +
> +	if (partition_info.a0 == FFA_ERROR)
> +		return ffa_to_linux_errno((int)partition_info.a2);
> +
> +	count = partition_info.a2;
> +
> +	if (buffer)
> +		memcpy(*buffer, drv_info->rx_buffer, sizeof(*buffer) * count);
> +
> +	ffa_rx_release();
> +
> +	mutex_unlock(&drv_info->rx_lock);
> +
> +	return count;
> +}
> +
> +static int ffa_partition_probe(const char *uuid_str,
> +			       struct ffa_partition_info *buffer)
> +{
> +	int count;
> +	uuid_t uuid;
> +	u32 uuid0_4[4] = { 0 };
> +
> +	if (uuid_parse(uuid_str, &uuid)) {
> +		pr_err("invalid uuid (%s)\n", uuid_str);
> +		return -ENODEV;
> +	}
> +
> +	export_uuid((u8 *)uuid0_4, &uuid);
> +	count = ffa_partition_info_get(uuid0_4[0], uuid0_4[1], uuid0_4[2],
> +				       uuid0_4[3], &buffer);
> +
> +	return count;
> +}
> +
>  #define VM_ID_MASK	GENMASK(15, 0)
>  static int ffa_id_get(u16 *vm_id)
>  {
> @@ -143,6 +215,205 @@ static int ffa_id_get(u16 *vm_id)
>  	return 0;
>  }
>  
> +static int ffa_msg_send_direct_req(u16 src_id, u16 dst_id,
> +				   struct ffa_send_direct_data *data)
> +{
> +	u32 src_dst_ids = PACK_TARGET_INFO(src_id, dst_id);
> +	ffa_res_t ret;
> +
> +	ret = invoke_ffa_fn(FFA_FN_NATIVE(MSG_SEND_DIRECT_REQ), src_dst_ids, 0,
> +			    data->data0, data->data1, data->data2,
> +			    data->data3, data->data4);
> +
> +	while (ret.a0 == FFA_INTERRUPT)
> +		ret = invoke_ffa_fn(FFA_RUN, ret.a1, 0, 0, 0, 0, 0, 0);
> +	if (ret.a0 == FFA_ERROR)
> +		return ffa_to_linux_errno((int)ret.a2);
> +
> +	if (ret.a0 == FFA_FN_NATIVE(MSG_SEND_DIRECT_RESP)) {
> +		data->data0 = ret.a3;
> +		data->data1 = ret.a4;
> +		data->data2 = ret.a5;
> +		data->data3 = ret.a6;
> +		data->data4 = ret.a7;
> +	}
> +
> +	return 0;
> +}
> +
> +static void ffa_device_get(struct ffa_device *ffa_dev)
> +{
> +	mutex_lock(&ffa_dev->mutex);
> +	ffa_dev->num_users++;
> +	mutex_unlock(&ffa_dev->mutex);
> +}
> +
> +static void ffa_device_put(struct ffa_device *ffa_dev)
> +{
> +	mutex_lock(&ffa_dev->mutex);
> +	ffa_dev->num_users--;
> +	mutex_unlock(&ffa_dev->mutex);
> +}
> +
> +static const struct ffa_dev_ops ffa_ops;
> +
> +const struct ffa_dev_ops *ffa_dev_ops_get(struct ffa_device *dev)
> +{
> +	struct list_head *p;
> +	const struct ffa_dev_ops *ops = NULL;
> +
> +	if (uuid_is_null(&dev->uuid))
> +		return NULL;
> +
> +	mutex_lock(&ffa_devs_list_mutex);
> +
> +	list_for_each(p, &ffa_devs_list) {
> +		struct ffa_device *tmp_dev;
> +
> +		tmp_dev = list_to_ffa_dev(p);
> +
> +		if (uuid_equal(&tmp_dev->uuid, &dev->uuid)) {
> +			ops = &ffa_ops;
> +			break;
> +		}
> +	}
> +
> +	mutex_unlock(&ffa_devs_list_mutex);
> +
> +	return ops;
> +}
> +
> +static int ffa_dev_open(struct ffa_device *ffa_dev)
> +{
> +	ffa_device_get(ffa_dev);
> +
> +	return 0;
> +}
> +
> +static int ffa_dev_close(struct ffa_device *ffa_dev)
> +{
> +	ffa_device_put(ffa_dev);
> +
> +	return 0;
> +}
> +
> +static long
> +ffa_dev_ioctl(struct ffa_device *ffa_dev, unsigned int ioctl, void *arg)
> +{
> +	long r = -EINVAL;
> +
> +	switch (ioctl) {
> +	case FFA_GET_API_VERSION:
> +		return drv_info->version;
> +	case FFA_GET_PARTITION_ID:
> +		return ffa_dev->vm_id;
> +	case FFA_GET_PARTITION_INFO: {
> +		struct ffa_part_info *pinfo = arg;
> +
> +		if (ffa_partition_probe(pinfo->uuid_str, &pinfo->info) != 1)
> +			r = -E2BIG;
> +		break;
> +	}
> +	case FFA_SEND_RECEIVE_SYNC: {
> +		struct ffa_send_recv_sync *kdata = arg;
> +
> +		r = ffa_msg_send_direct_req(ffa_dev->vm_id, kdata->endpoint_id,
> +					    &kdata->data);
> +		break;
> +	}
> +	default:
> +		r = -EINVAL;
> +	}
> +
> +	return r;
> +}
> +
> +static const struct ffa_dev_ops ffa_ops = {
> +	.open = ffa_dev_open,
> +	.close = ffa_dev_close,
> +	.ioctl = ffa_dev_ioctl,
> +};
> +
> +static int ffa_device_alloc_register(const char *name, u16 vm_id, uuid_t *uuid,
> +				     const struct file_operations *fops)
> +{
> +	int ret;
> +	struct ffa_device *ffa_dev;
> +
> +	ffa_dev = kzalloc(sizeof(*ffa_dev), GFP_KERNEL);
> +	if (!ffa_dev)
> +		return -ENOMEM;
> +
> +	ffa_dev->vm_id = vm_id;
> +	if (uuid)
> +		uuid_copy(&ffa_dev->uuid, uuid);
> +
> +	dev_set_name(&ffa_dev->dev, name);
> +	dev_set_drvdata(&ffa_dev->dev, drv_info);
> +	cdev_init(&ffa_dev->cdev, fops);
> +
> +	ret = ffa_device_register(ffa_dev);
> +	if (ret)
> +		return ret;
> +
> +	mutex_lock(&ffa_devs_list_mutex);
> +	list_add_tail(&ffa_dev->node, &ffa_devs_list);
> +	mutex_unlock(&ffa_devs_list_mutex);
> +
> +	return 0;
> +}
> +
> +int ffa_setup_partitions(const char *compatible,
> +			 const struct file_operations *fops)
> +{
> +	int ret;
> +	struct ffa_partition_info pbuf;
> +	struct device_node *child, *parent;
> +	const char *p_uuid, *pfx = "Ignoring FFA partition";
> +	uuid_t uuid = UUID_INIT(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
> +
> +	parent = of_find_compatible_node(NULL, NULL, compatible);
> +	if (!parent)
> +		return 0;
> +
> +	for_each_child_of_node(parent, child) {
> +		if (!of_device_is_compatible(child, "arm,ffa-1.0-partition")) {
> +			of_node_put(child);
> +			continue;
> +		}
> +
> +		if (of_property_read_string(child, "uuid", &p_uuid)) {
> +			pr_err("%s: failed to parse \"uuid\" property\n", pfx);
> +			of_node_put(child);
> +			continue;
> +		}
> +
> +		of_node_put(child);
> +
> +		if (uuid_parse(p_uuid, &uuid)) {
> +			pr_err("%s: invalid \"uuid\" property (%s)\n",
> +			       pfx, p_uuid);
> +			continue;
> +		}
> +
> +		ret = ffa_partition_probe(p_uuid, &pbuf);
> +		if (ret != 1) {
> +			pr_err("%s: %s partition info probe failed\n",
> +			       pfx, p_uuid);
> +			return -EINVAL;
> +		}
> +
> +		ret = ffa_device_alloc_register(p_uuid, pbuf.id, &uuid, fops);
> +		if (ret) {
> +			pr_err("%s: failed to register %s\n", pfx, p_uuid);
> +			continue;
> +		}
> +	}
> +
> +	of_node_put(parent);
> +	return 0;
> +}
> +
>  static int __init ffa_init(void)
>  {
>  	int ret;
> @@ -192,6 +463,9 @@ static int __init ffa_init(void)
>  	mutex_init(&drv_info->rx_lock);
>  	mutex_init(&drv_info->tx_lock);
>  
> +	/* Set up all the partitions which have in-kernel drivers */
> +	ffa_setup_partitions("arm,ffa-1.0", NULL);
> +
>  	return 0;
>  free_pages:
>  	if (drv_info->tx_buffer)
> @@ -205,6 +479,13 @@ module_init(ffa_init);
>  
>  static void __exit ffa_exit(void)
>  {
> +	struct list_head *p;
> +
> +	mutex_lock(&ffa_devs_list_mutex);
> +	list_for_each(p, &ffa_devs_list)
> +		ffa_device_unregister(list_to_ffa_dev(p));
> +	mutex_unlock(&ffa_devs_list_mutex);
> +
>  	ffa_rxtx_unmap(drv_info->vm_id);
>  	free_pages_exact(drv_info->tx_buffer, RXTX_BUFFER_SIZE);
>  	free_pages_exact(drv_info->rx_buffer, RXTX_BUFFER_SIZE);
> diff --git a/include/linux/arm_ffa.h b/include/linux/arm_ffa.h
> index 6185d0d12f15..719ef02fe42d 100644
> --- a/include/linux/arm_ffa.h
> +++ b/include/linux/arm_ffa.h
> @@ -11,6 +11,7 @@
>  #include <linux/module.h>
>  #include <linux/types.h>
>  #include <linux/uuid.h>
> +#include <uapi/linux/arm_ffa.h>
>  
>  #define FFA_SMC(calling_convention, func_num)				\
>  	ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL, (calling_convention),	\
> @@ -94,6 +95,9 @@ struct ffa_device {
>  	uuid_t uuid;
>  	struct device dev;
>  	struct cdev cdev;
> +	size_t num_users;
> +	struct mutex mutex;	/* protects num_users */
> +	struct list_head node;
>  };
>  
>  #define to_ffa_dev(d) container_of(d, struct ffa_device, dev)
> @@ -113,12 +117,19 @@ struct ffa_driver {
>  
>  #define to_ffa_driver(d) container_of(d, struct ffa_driver, driver)
>  
> +struct ffa_dev_ops {
> +	int (*open)(struct ffa_device *dev);
> +	int (*close)(struct ffa_device *dev);
> +	long (*ioctl)(struct ffa_device *dev, unsigned int ioctl, void *arg);
> +};

I assume that all interaction with a SP is done via ffa_ops->ioctl().
For example the ffa_msg_send_direct_req() function is then called via:
struct ffa_send_recv_sync arg = { .endpoint_id = xxx, .data = yyy };
rc = ffa_ops->ioctl(ffa_dev, FFA_SEND_RECEIVE_SYNC, &arg);

That isn't too hard to use, but is a bit inconvenient and less safe
compared to a plain:
rc = ffa_ops->send_recieve_sync(ffa_dev, xxx, yyy);

I don't see any big win in then next patch with ffa_ioctl() either. That
function must still do some actions specific for each ioctl id. So I'm a
bit curious about the design choice.

Cheers,
Jens

> +
>  #if IS_REACHABLE(CONFIG_ARM_FFA_TRANSPORT)
>  int ffa_device_register(struct ffa_device *ffa_dev);
>  void ffa_device_unregister(struct ffa_device *ffa_dev);
>  int ffa_driver_register(struct ffa_driver *driver, struct module *owner,
>  			const char *mod_name);
>  void ffa_driver_unregister(struct ffa_driver *driver);
> +const struct ffa_dev_ops *ffa_dev_ops_get(struct ffa_device *dev);
>  
>  #else
>  static inline int ffa_device_register(struct ffa_device *ffa_dev)
> @@ -137,6 +148,10 @@ ffa_driver_register(struct ffa_driver *driver, struct module *owner,
>  
>  static inline void ffa_driver_unregister(struct ffa_driver *driver) {}
>  
> +const struct ffa_dev_ops *ffa_dev_ops_get(struct ffa_device *dev)
> +{
> +	return NULL;
> +}
>  #endif /* CONFIG_ARM_FFA_TRANSPORT */
>  
>  #define ffa_register(driver) \
> diff --git a/include/uapi/linux/arm_ffa.h b/include/uapi/linux/arm_ffa.h
> new file mode 100644
> index 000000000000..88ddddb4742f
> --- /dev/null
> +++ b/include/uapi/linux/arm_ffa.h
> @@ -0,0 +1,56 @@
> +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
> +/*
> + * Copyright (C) 2020 Arm Ltd.
> + */
> +
> +#ifndef _UAPI_LINUX_ARM_FFA_H
> +#define _UAPI_LINUX_ARM_FFA_H
> +
> +#include <linux/types.h>
> +#include <linux/ioctl.h>
> +
> +#define FFA_BASE	0xFF
> +
> +struct ffa_partition_info {
> +	__u16 id;
> +	__u16 exec_ctxt;
> +/* partition supports receipt of direct requests */
> +#define FFA_PARTITION_DIRECT_RECV	BIT(0)
> +/* partition can send direct requests. */
> +#define FFA_PARTITION_DIRECT_SEND	BIT(1)
> +/* partition can send and receive indirect messages. */
> +#define FFA_PARTITION_INDIRECT_MSG	BIT(2)
> +	__u32 properties;
> +};
> +
> +struct ffa_part_info {
> +	char uuid_str[36];
> +	struct ffa_partition_info info;
> +};
> +
> +struct ffa_send_direct_data {
> +	unsigned long data0;
> +	unsigned long data1;
> +	unsigned long data2;
> +	unsigned long data3;
> +	unsigned long data4;
> +};
> +
> +struct ffa_send_recv_sync {
> +	__u16 endpoint_id;
> +	struct ffa_send_direct_data data;
> +};
> +
> +struct ffa_send_recv_async {
> +	__u16 endpoint_id;
> +	int length;
> +	char buffer[];
> +};
> +
> +#define FFA_GET_API_VERSION	_IO(FFA_BASE, 0x00)
> +#define FFA_GET_PARTITION_ID	_IO(FFA_BASE, 0x01)
> +#define FFA_GET_PARTITION_INFO	_IOWR(FFA_BASE, 0x02, struct ffa_part_info)
> +#define FFA_SEND_RECEIVE_SYNC	_IOWR(FFA_BASE, 0x03, struct ffa_send_recv_sync)
> +#define FFA_SEND_RECEIVE_ASYNC	_IOW(FFA_BASE, 0x04, struct ffa_send_recv_async)
> +
> +#endif /*_UAPI_LINUX_ARM_FFA_H*/
> -- 
> 2.25.1
> 

Re: [PATCH v2 7/9] firmware: arm_ffa: Setup in-kernel users of FFA partitions

[Sudeep Holla]

On Sat, Nov 28, 2020 at 02:36:55PM +0100, Jens Wiklander wrote:
> Hi Sudeep,
>

[...]

> > @@ -113,12 +117,19 @@ struct ffa_driver {
> >
> >  #define to_ffa_driver(d) container_of(d, struct ffa_driver, driver)
> >
> > +struct ffa_dev_ops {
> > +	int (*open)(struct ffa_device *dev);
> > +	int (*close)(struct ffa_device *dev);
> > +	long (*ioctl)(struct ffa_device *dev, unsigned int ioctl, void *arg);
> > +};
>
> I assume that all interaction with a SP is done via ffa_ops->ioctl().

Yes that was the idea.

> For example the ffa_msg_send_direct_req() function is then called via:
> struct ffa_send_recv_sync arg = { .endpoint_id = xxx, .data = yyy };
> rc = ffa_ops->ioctl(ffa_dev, FFA_SEND_RECEIVE_SYNC, &arg);
>

Correct.

> That isn't too hard to use, but is a bit inconvenient and less safe
> compared to a plain:
> rc = ffa_ops->send_recieve_sync(ffa_dev, xxx, yyy);
>

Agreed.

> I don't see any big win in then next patch with ffa_ioctl() either. That
> function must still do some actions specific for each ioctl id. So I'm a
> bit curious about the design choice.
>

Initial idea was to keep both in-kernel and user-space interface inline.
Also the assumption was that expect few old/legacy usecases, the userspace
is the way forward. While that is still ideal, but things have changed
since the main user of userspace interface(pKVM) is no longer using
FFA. I will change the interface as you mention above. I was also more
inclined towards that after dropping userspace. Good timing though, I was
about to post revised version dropping userspace. I will modify the
interface something on lines of your suggestion.

--
Regards,
Sudeep

Re: [PATCH v2 0/9] firmware: Add initial support for Arm FF-A

[Sudeep Holla]

On Sat, Nov 28, 2020 at 01:25:02PM +0100, Jens Wiklander wrote:
> Hi Sudeep,
> 
> On Tue, Nov 03, 2020 at 05:43:41PM +0000, Sudeep Holla wrote:
> > Hi all,
> > 
> > Let me start stating this is just initial implementation to check on
> > the idea of providing more in-kernel and userspace support. Lot of things
> > are still work in progress, I am posting just to get the early feedback
> > before building lot of things on this idea. Consider this more as RFC
> > though not tagged explicity(just to avoid it being ignored :))
> > 
> > Arm Firmware Framework for Armv8-A specification[1] describes a software
> > architecture that provides mechanism to utilise the virtualization
> > extension to isolate software images and describes interfaces that
> > standardize communication between the various software images. This
> > includes communication between images in the Secure and Normal world.
> > 
> > The main idea here is to create FFA device to establish any communication
> > with a partition(secure or normal world VM).
> > 
> > If it is a partition managed by hypervisor, then we will register chardev
> > associated with each of those partition FFA device.
> > 
> > /dev/arm_ffa:
> > 
> > e3a48fa5-dc54-4a8b-898b-bdc4dfeeb7b8
> > 49f65057-d002-4ae2-b4ee-d31c7940a13d
> > 
> > For in-kernel usage(mostly communication with secure partitions), only
> > in-kernel APIs are accessible(no userspace). There may be a need to
> > provide userspace access instead of in-kernel, it is not yet support
> > in this series as we need way to identify those and I am not sure if
> > that belong to DT.
> 
> With unfiltered VM to VM commnication from user space there's no easy
> way for two VMs to exchange privileged information that excludes user
> space.

Though this usercase is dropped now, it was targeted for VMM and may be
it was not an issue there.

> Perhaps access to the FFA device is considered privileged and
> enough for all purposes.
>

I don't know TBH.

> If I've understood it correctly is VM to SP communication only allowed
> via kernel mode in the VM.

Correct.

> The communication with OP-TEE depends on this with the recent commit
> c5b4312bea5d ("tee: optee: Add support for session login client UUID
> generation").
>

OK, thanks for the info.

-- 
Regards,
Sudeep