[PATCH v4 00/23] Linux SBI MPXY and RPMI drivers

[PATCH v4 00/23] Linux SBI MPXY and RPMI drivers

[Anup Patel]

The SBI v3.0 (MPXY extension) [1] and RPMI v1.0 [2] specifications
are frozen and in public review at the RISC-V International.

Currently, most of the RPMI and MPXY drivers are in OpenSBI whereas
Linux only has SBI MPXY mailbox controller driver, RPMI clock driver
and RPMI system MSI driver This series also includes ACPI support
for SBI MPXY mailbox controller and RPMI system MSI drivers.

These patches can be found in the riscv_sbi_mpxy_mailbox_v4 branch
at: https://github.com/avpatel/linux.git

To test these patches, boot Linux on "virt,rpmi=on,aia=aplic-imsic"
machine with OpenSBI and QEMU from the dev-upstream branch at:
https://github.com/ventanamicro/opensbi.git
https://github.com/ventanamicro/qemu.git

[1] https://github.com/riscv-non-isa/riscv-sbi-doc/releases
[2] https://github.com/riscv-non-isa/riscv-rpmi/releases

Changes since v3:
 - Rebased the series on Linux-6.15-rc7
 - Updated PATCH2 DT bindings as-per Rob's suggestion
 - Improved request_threaded_irq() usage in PATCH7
 - Updated PATCH10 clk-rpmi driver as-per commments from Andy
 - Updated PATCH13 irq-riscv-rpmi-sysmsi driver as-per comments
   from Andy and Tglx
 - Addressed ACPI related comments in PATCH14, PATCH15, PATCH18,
   PATCH20 and PATCH21

Changes since v2:
 - Dropped the "RFC" tag from series since the SBI v3.0 and
   RPMI v1.0 specifications are now frozen
 - Rebased the series on Linux-6.15-rc5
 - Split PATCH8 of v2 into two patches adding separate DT
   bindings for "riscv,rpmi-mpxy-clock" and "riscv,rpmi-clock"
 - Split PATCH10 of v2 into two patches adding separate DT
   bindings for "riscv,rpmi-mpxy-system-msi" and
   "riscv,rpmi-system-msi"
 - Addressed comments from TGLX on PATCH11 of v2 adding irqchip
   driver for RPMI system MSI
 - Addressed ACPI related comments in PATCH15 and PATCH16 of v2
 - New PATCH17 and PATCH18 in this series

Changes since v1:
 - Addressed DT bindings related comments in PATCH2, PATCH3, and
   PATCH7 of v1 series
 - Addressed comments in PATCH6 and PATCH8 of v1 series
 - New PATCH6 in v2 series to allow fwnode based mailbox channel
   request
 - New PATCH10 and PATCH11 to add RPMI system MSI based interrupt
   controller driver
 - New PATCH12 to PATCH16 which adds ACPI support in SBI MPXY
   mailbox driver and RPMI system MSI driver
 - New PATCH17 to enable required kconfig option to allow graceful
   shutdown on QEMU virt machine

Anup Patel (14):
  riscv: Add new error codes defined by SBI v3.0
  dt-bindings: mailbox: Add bindings for RPMI shared memory transport
  dt-bindings: mailbox: Add bindings for RISC-V SBI MPXY extension
  RISC-V: Add defines for the SBI message proxy extension
  mailbox: Add common header for RPMI messages sent via mailbox
  mailbox: Allow controller specific mapping using fwnode
  mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver
  dt-bindings: clock: Add RPMI clock service message proxy bindings
  dt-bindings: clock: Add RPMI clock service controller bindings
  dt-bindings: Add RPMI system MSI message proxy bindings
  dt-bindings: Add RPMI system MSI interrupt controller bindings
  irqchip: Add driver for the RPMI system MSI service group
  RISC-V: Enable GPIO keyboard and event device in RV64 defconfig
  MAINTAINERS: Add entry for RISC-V RPMI and MPXY drivers

Rahul Pathak (1):
  clk: Add clock driver for the RISC-V RPMI clock service group

Sunil V L (8):
  ACPI: property: Refactor acpi_fwnode_get_reference_args()
  ACPI: property: Add support for cells property
  ACPI: scan: Update honor list for RPMI System MSI
  ACPI: RISC-V: Create interrupt controller list in sorted order
  ACPI: RISC-V: Add support to update gsi range
  ACPI: RISC-V: Add RPMI System MSI to GSI mapping
  mailbox/riscv-sbi-mpxy: Add ACPI support
  irqchip/riscv-rpmi-sysmsi: Add ACPI support

 .../bindings/clock/riscv,rpmi-clock.yaml      |   61 +
 .../bindings/clock/riscv,rpmi-mpxy-clock.yaml |   64 ++
 .../riscv,rpmi-mpxy-system-msi.yaml           |   67 ++
 .../riscv,rpmi-system-msi.yaml                |   74 ++
 .../mailbox/riscv,rpmi-shmem-mbox.yaml        |  124 ++
 .../bindings/mailbox/riscv,sbi-mpxy-mbox.yaml |   51 +
 MAINTAINERS                                   |   15 +
 arch/riscv/configs/defconfig                  |    2 +
 arch/riscv/include/asm/irq.h                  |    6 +
 arch/riscv/include/asm/sbi.h                  |   72 ++
 drivers/acpi/property.c                       |  123 +-
 drivers/acpi/riscv/irq.c                      |   80 +-
 drivers/acpi/scan.c                           |    2 +
 drivers/base/property.c                       |    2 +-
 drivers/clk/Kconfig                           |    8 +
 drivers/clk/Makefile                          |    1 +
 drivers/clk/clk-rpmi.c                        |  589 ++++++++++
 drivers/irqchip/Kconfig                       |    7 +
 drivers/irqchip/Makefile                      |    1 +
 drivers/irqchip/irq-riscv-rpmi-sysmsi.c       |  324 ++++++
 drivers/mailbox/Kconfig                       |   11 +
 drivers/mailbox/Makefile                      |    2 +
 drivers/mailbox/mailbox.c                     |   47 +-
 drivers/mailbox/riscv-sbi-mpxy-mbox.c         | 1003 +++++++++++++++++
 include/linux/mailbox/riscv-rpmi-message.h    |  233 ++++
 include/linux/mailbox_controller.h            |    3 +
 include/linux/wordpart.h                      |    8 +
 27 files changed, 2905 insertions(+), 75 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
 create mode 100644 Documentation/devicetree/bindings/clock/riscv,rpmi-mpxy-clock.yaml
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
 create mode 100644 Documentation/devicetree/bindings/mailbox/riscv,rpmi-shmem-mbox.yaml
 create mode 100644 Documentation/devicetree/bindings/mailbox/riscv,sbi-mpxy-mbox.yaml
 create mode 100644 drivers/clk/clk-rpmi.c
 create mode 100644 drivers/irqchip/irq-riscv-rpmi-sysmsi.c
 create mode 100644 drivers/mailbox/riscv-sbi-mpxy-mbox.c
 create mode 100644 include/linux/mailbox/riscv-rpmi-message.h

-- 
2.43.0

[PATCH v4 01/23] riscv: Add new error codes defined by SBI v3.0

[Anup Patel]

The SBI v3.0 defines new error codes so add these new error codes
to the asm/sbi.h for use by newer SBI extensions.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 arch/riscv/include/asm/sbi.h | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/arch/riscv/include/asm/sbi.h b/arch/riscv/include/asm/sbi.h
index 3d250824178b..4dd6aafb8468 100644
--- a/arch/riscv/include/asm/sbi.h
+++ b/arch/riscv/include/asm/sbi.h
@@ -419,6 +419,11 @@ enum sbi_ext_nacl_feature {
 #define SBI_ERR_ALREADY_STARTED -7
 #define SBI_ERR_ALREADY_STOPPED -8
 #define SBI_ERR_NO_SHMEM	-9
+#define SBI_ERR_INVALID_STATE	-10
+#define SBI_ERR_BAD_RANGE	-11
+#define SBI_ERR_TIMEOUT		-12
+#define SBI_ERR_IO		-13
+#define SBI_ERR_DENIED_LOCKED	-14
 
 extern unsigned long sbi_spec_version;
 struct sbiret {
@@ -503,11 +508,18 @@ static inline int sbi_err_map_linux_errno(int err)
 	case SBI_SUCCESS:
 		return 0;
 	case SBI_ERR_DENIED:
+	case SBI_ERR_DENIED_LOCKED:
 		return -EPERM;
 	case SBI_ERR_INVALID_PARAM:
+	case SBI_ERR_INVALID_STATE:
+	case SBI_ERR_BAD_RANGE:
 		return -EINVAL;
 	case SBI_ERR_INVALID_ADDRESS:
 		return -EFAULT;
+	case SBI_ERR_TIMEOUT:
+		return -ETIMEDOUT;
+	case SBI_ERR_IO:
+		return -EIO;
 	case SBI_ERR_NOT_SUPPORTED:
 	case SBI_ERR_FAILURE:
 	default:
-- 
2.43.0

[PATCH v4 02/23] dt-bindings: mailbox: Add bindings for RPMI shared memory transport

[Anup Patel]

Add device tree bindings for the common RISC-V Platform Management
Interface (RPMI) shared memory transport as a mailbox controller.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 .../mailbox/riscv,rpmi-shmem-mbox.yaml        | 124 ++++++++++++++++++
 1 file changed, 124 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/mailbox/riscv,rpmi-shmem-mbox.yaml

diff --git a/Documentation/devicetree/bindings/mailbox/riscv,rpmi-shmem-mbox.yaml b/Documentation/devicetree/bindings/mailbox/riscv,rpmi-shmem-mbox.yaml
new file mode 100644
index 000000000000..a5bcf1fba464
--- /dev/null
+++ b/Documentation/devicetree/bindings/mailbox/riscv,rpmi-shmem-mbox.yaml
@@ -0,0 +1,124 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/mailbox/riscv,rpmi-shmem-mbox.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: RISC-V Platform Management Interface (RPMI) shared memory mailbox
+
+maintainers:
+  - Anup Patel <anup@brainfault.org>
+
+description: |
+  The RISC-V Platform Management Interface (RPMI) [1] defines a common shared
+  memory based RPMI transport. This RPMI shared memory transport integrates as
+  mailbox controller in the SBI implementation or supervisor software whereas
+  each RPMI service group is mailbox client in the SBI implementation and
+  supervisor software.
+
+  ===========================================
+  References
+  ===========================================
+
+  [1] RISC-V Platform Management Interface (RPMI)
+      https://github.com/riscv-non-isa/riscv-rpmi/releases
+
+properties:
+  compatible:
+    const: riscv,rpmi-shmem-mbox
+
+  reg:
+    minItems: 2
+    items:
+      - description: A2P request queue base address
+      - description: P2A acknowledgment queue base address
+      - description: P2A request queue base address
+      - description: A2P acknowledgment queue base address
+      - description: A2P doorbell address
+
+  reg-names:
+    minItems: 2
+    items:
+      - const: a2p-req
+      - const: p2a-ack
+      - enum: [ p2a-req, a2p-doorbell ]
+      - const: a2p-ack
+      - const: a2p-doorbell
+
+  interrupts:
+    maxItems: 1
+    description:
+      The RPMI shared memory transport supports P2A doorbell as a wired
+      interrupt and this property specifies the interrupt source.
+
+  msi-parent:
+    description:
+      The RPMI shared memory transport supports P2A doorbell as a system MSI
+      and this property specifies the target MSI controller.
+
+  riscv,slot-size:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    minimum: 64
+    description:
+      Power-of-2 RPMI slot size of the RPMI shared memory transport.
+
+  riscv,a2p-doorbell-value:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    default: 0x1
+    description:
+      Value written to the 32-bit A2P doorbell register.
+
+  riscv,p2a-doorbell-sysmsi-index:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    description:
+      The RPMI shared memory transport supports P2A doorbell as a system MSI
+      and this property specifies system MSI index to be used for configuring
+      the P2A doorbell MSI.
+
+  "#mbox-cells":
+    const: 1
+    description:
+      The first cell specifies RPMI service group ID.
+
+required:
+  - compatible
+  - reg
+  - reg-names
+  - riscv,slot-size
+  - "#mbox-cells"
+
+anyOf:
+  - required:
+      - interrupts
+  - required:
+      - msi-parent
+
+additionalProperties: false
+
+examples:
+  - |
+    // Example 1 (RPMI shared memory with only 2 queues):
+    mailbox@10080000 {
+        compatible = "riscv,rpmi-shmem-mbox";
+        reg = <0x10080000 0x10000>,
+              <0x10090000 0x10000>;
+        reg-names = "a2p-req", "p2a-ack";
+        msi-parent = <&imsic_mlevel>;
+        riscv,slot-size = <64>;
+        #mbox-cells = <1>;
+    };
+  - |
+    // Example 2 (RPMI shared memory with only 4 queues):
+    mailbox@10001000 {
+        compatible = "riscv,rpmi-shmem-mbox";
+        reg = <0x10001000 0x800>,
+              <0x10001800 0x800>,
+              <0x10002000 0x800>,
+              <0x10002800 0x800>,
+              <0x10003000 0x4>;
+        reg-names = "a2p-req", "p2a-ack", "p2a-req", "a2p-ack", "a2p-doorbell";
+        msi-parent = <&imsic_mlevel>;
+        riscv,slot-size = <64>;
+        riscv,a2p-doorbell-value = <0x00008000>;
+        #mbox-cells = <1>;
+    };
-- 
2.43.0

[PATCH v4 03/23] dt-bindings: mailbox: Add bindings for RISC-V SBI MPXY extension

[Anup Patel]

Add device tree bindings for the RISC-V SBI Message Proxy (MPXY)
extension as a mailbox controller.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 .../bindings/mailbox/riscv,sbi-mpxy-mbox.yaml | 51 +++++++++++++++++++
 1 file changed, 51 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/mailbox/riscv,sbi-mpxy-mbox.yaml

diff --git a/Documentation/devicetree/bindings/mailbox/riscv,sbi-mpxy-mbox.yaml b/Documentation/devicetree/bindings/mailbox/riscv,sbi-mpxy-mbox.yaml
new file mode 100644
index 000000000000..ad4c135e0510
--- /dev/null
+++ b/Documentation/devicetree/bindings/mailbox/riscv,sbi-mpxy-mbox.yaml
@@ -0,0 +1,51 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/mailbox/riscv,sbi-mpxy-mbox.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: RISC-V SBI Message Proxy (MPXY) extension based mailbox
+
+maintainers:
+  - Anup Patel <anup@brainfault.org>
+
+description: |
+  The RISC-V SBI Message Proxy (MPXY) extension [1] allows supervisor
+  software to send messages through the SBI implementation (M-mode
+  firmware or HS-mode hypervisor). The underlying message protocol
+  and message format used by the supervisor software could be some
+  other standard protocol compatible with the SBI MPXY extension
+  (such as RISC-V Platform Management Interface (RPMI) [2]).
+
+  ===========================================
+  References
+  ===========================================
+
+  [1] RISC-V Supervisor Binary Interface (SBI)
+      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
+
+  [2] RISC-V Platform Management Interface (RPMI)
+      https://github.com/riscv-non-isa/riscv-rpmi/releases
+
+properties:
+  compatible:
+    const: riscv,sbi-mpxy-mbox
+
+  "#mbox-cells":
+    const: 2
+    description:
+      The first cell specifies channel_id of the SBI MPXY channel,
+      the second cell specifies MSG_PROT_ID of the SBI MPXY channel
+
+required:
+  - compatible
+  - "#mbox-cells"
+
+additionalProperties: false
+
+examples:
+  - |
+    mailbox {
+          compatible = "riscv,sbi-mpxy-mbox";
+          #mbox-cells = <2>;
+    };
-- 
2.43.0

[PATCH v4 04/23] RISC-V: Add defines for the SBI message proxy extension

[Anup Patel]

Add defines for the new SBI message proxy extension which is part
of the SBI v3.0 specification.

Co-developed-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 arch/riscv/include/asm/sbi.h | 60 ++++++++++++++++++++++++++++++++++++
 1 file changed, 60 insertions(+)

diff --git a/arch/riscv/include/asm/sbi.h b/arch/riscv/include/asm/sbi.h
index 4dd6aafb8468..d295c26a7c26 100644
--- a/arch/riscv/include/asm/sbi.h
+++ b/arch/riscv/include/asm/sbi.h
@@ -35,6 +35,7 @@ enum sbi_ext_id {
 	SBI_EXT_DBCN = 0x4442434E,
 	SBI_EXT_STA = 0x535441,
 	SBI_EXT_NACL = 0x4E41434C,
+	SBI_EXT_MPXY = 0x4D505859,
 
 	/* Experimentals extensions must lie within this range */
 	SBI_EXT_EXPERIMENTAL_START = 0x08000000,
@@ -402,6 +403,65 @@ enum sbi_ext_nacl_feature {
 #define SBI_NACL_SHMEM_SRET_X(__i)		((__riscv_xlen / 8) * (__i))
 #define SBI_NACL_SHMEM_SRET_X_LAST		31
 
+enum sbi_ext_mpxy_fid {
+	SBI_EXT_MPXY_GET_SHMEM_SIZE,
+	SBI_EXT_MPXY_SET_SHMEM,
+	SBI_EXT_MPXY_GET_CHANNEL_IDS,
+	SBI_EXT_MPXY_READ_ATTRS,
+	SBI_EXT_MPXY_WRITE_ATTRS,
+	SBI_EXT_MPXY_SEND_MSG_WITH_RESP,
+	SBI_EXT_MPXY_SEND_MSG_WITHOUT_RESP,
+	SBI_EXT_MPXY_GET_NOTIFICATION_EVENTS,
+};
+
+enum sbi_mpxy_attribute_id {
+	/* Standard channel attributes managed by MPXY framework */
+	SBI_MPXY_ATTR_MSG_PROT_ID		= 0x00000000,
+	SBI_MPXY_ATTR_MSG_PROT_VER		= 0x00000001,
+	SBI_MPXY_ATTR_MSG_MAX_LEN		= 0x00000002,
+	SBI_MPXY_ATTR_MSG_SEND_TIMEOUT		= 0x00000003,
+	SBI_MPXY_ATTR_MSG_COMPLETION_TIMEOUT	= 0x00000004,
+	SBI_MPXY_ATTR_CHANNEL_CAPABILITY	= 0x00000005,
+	SBI_MPXY_ATTR_SSE_EVENT_ID		= 0x00000006,
+	SBI_MPXY_ATTR_MSI_CONTROL		= 0x00000007,
+	SBI_MPXY_ATTR_MSI_ADDR_LO		= 0x00000008,
+	SBI_MPXY_ATTR_MSI_ADDR_HI		= 0x00000009,
+	SBI_MPXY_ATTR_MSI_DATA			= 0x0000000A,
+	SBI_MPXY_ATTR_EVENTS_STATE_CONTROL	= 0x0000000B,
+	SBI_MPXY_ATTR_STD_ATTR_MAX_IDX,
+	/*
+	 * Message protocol specific attributes, managed by
+	 * the message protocol specification.
+	 */
+	SBI_MPXY_ATTR_MSGPROTO_ATTR_START	= 0x80000000,
+	SBI_MPXY_ATTR_MSGPROTO_ATTR_END		= 0xffffffff
+};
+
+/* Possible values of MSG_PROT_ID attribute */
+enum sbi_mpxy_msgproto_id {
+	SBI_MPXY_MSGPROTO_RPMI_ID = 0x0,
+};
+
+/** RPMI message protocol specific MPXY attributes */
+enum sbi_mpxy_rpmi_attribute_id {
+	SBI_MPXY_RPMI_ATTR_SERVICEGROUP_ID = SBI_MPXY_ATTR_MSGPROTO_ATTR_START,
+	SBI_MPXY_RPMI_ATTR_SERVICEGROUP_VERSION,
+	SBI_MPXY_RPMI_ATTR_MAX_ID,
+};
+
+/* Encoding of MSG_PROT_VER attribute */
+#define SBI_MPXY_MSG_PROT_VER_MAJOR(__ver)	(((__ver) >> 16) & 0xffff)
+#define SBI_MPXY_MSG_PROT_VER_MINOR(__ver)	((__ver) & 0xffff)
+#define SBI_MPXY_MSG_PROT_MKVER(__maj, __min)	(((__maj) << 16) | (__min))
+
+/* Capabilities available through CHANNEL_CAPABILITY attribute */
+#define SBI_MPXY_CHAN_CAP_MSI			BIT(0)
+#define SBI_MPXY_CHAN_CAP_SSE			BIT(1)
+#define SBI_MPXY_CHAN_CAP_EVENTS_STATE		BIT(2)
+#define SBI_MPXY_CHAN_CAP_SEND_WITH_RESP	BIT(3)
+#define SBI_MPXY_CHAN_CAP_SEND_WITHOUT_RESP	BIT(4)
+#define SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS	BIT(5)
+
 /* SBI spec version fields */
 #define SBI_SPEC_VERSION_DEFAULT	0x1
 #define SBI_SPEC_VERSION_MAJOR_SHIFT	24
-- 
2.43.0

[PATCH v4 05/23] mailbox: Add common header for RPMI messages sent via mailbox

[Anup Patel]

The RPMI based mailbox controller drivers and mailbox clients need to
share defines related to RPMI messages over mailbox interface so add
a common header for this purpose.

Co-developed-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 include/linux/mailbox/riscv-rpmi-message.h | 204 +++++++++++++++++++++
 1 file changed, 204 insertions(+)
 create mode 100644 include/linux/mailbox/riscv-rpmi-message.h

diff --git a/include/linux/mailbox/riscv-rpmi-message.h b/include/linux/mailbox/riscv-rpmi-message.h
new file mode 100644
index 000000000000..6201d4dd6047
--- /dev/null
+++ b/include/linux/mailbox/riscv-rpmi-message.h
@@ -0,0 +1,204 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Ventana Micro Systems Inc. */
+
+#ifndef _LINUX_RISCV_RPMI_MESSAGE_H_
+#define _LINUX_RISCV_RPMI_MESSAGE_H_
+
+#include <linux/mailbox_client.h>
+
+/* RPMI version encode/decode macros */
+#define RPMI_VER_MAJOR(__ver)		(((__ver) >> 16) & 0xffff)
+#define RPMI_VER_MINOR(__ver)		((__ver) & 0xffff)
+#define RPMI_MKVER(__maj, __min)	(((__maj) << 16) | (__min))
+
+/* RPMI message header */
+struct rpmi_message_header {
+	__le16 servicegroup_id;
+	u8 service_id;
+	u8 flags;
+	__le16 datalen;
+	__le16 token;
+};
+
+/* RPMI message */
+struct rpmi_message {
+	struct rpmi_message_header header;
+	u8 data[];
+};
+
+/* RPMI notification event */
+struct rpmi_notification_event {
+	__le16 event_datalen;
+	u8 event_id;
+	u8 reserved;
+	u8 event_data[];
+};
+
+/* RPMI error codes */
+enum rpmi_error_codes {
+	RPMI_SUCCESS			= 0,
+	RPMI_ERR_FAILED			= -1,
+	RPMI_ERR_NOTSUPP		= -2,
+	RPMI_ERR_INVALID_PARAM		= -3,
+	RPMI_ERR_DENIED			= -4,
+	RPMI_ERR_INVALID_ADDR		= -5,
+	RPMI_ERR_ALREADY		= -6,
+	RPMI_ERR_EXTENSION		= -7,
+	RPMI_ERR_HW_FAULT		= -8,
+	RPMI_ERR_BUSY			= -9,
+	RPMI_ERR_INVALID_STATE		= -10,
+	RPMI_ERR_BAD_RANGE		= -11,
+	RPMI_ERR_TIMEOUT		= -12,
+	RPMI_ERR_IO			= -13,
+	RPMI_ERR_NO_DATA		= -14,
+	RPMI_ERR_RESERVED_START		= -15,
+	RPMI_ERR_RESERVED_END		= -127,
+	RPMI_ERR_VENDOR_START		= -128
+};
+
+static inline int rpmi_to_linux_error(int rpmi_error)
+{
+	switch (rpmi_error) {
+	case RPMI_SUCCESS:
+		return 0;
+	case RPMI_ERR_INVALID_PARAM:
+	case RPMI_ERR_BAD_RANGE:
+	case RPMI_ERR_INVALID_STATE:
+		return -EINVAL;
+	case RPMI_ERR_DENIED:
+		return -EPERM;
+	case RPMI_ERR_INVALID_ADDR:
+	case RPMI_ERR_HW_FAULT:
+		return -EFAULT;
+	case RPMI_ERR_ALREADY:
+		return -EALREADY;
+	case RPMI_ERR_BUSY:
+		return -EBUSY;
+	case RPMI_ERR_TIMEOUT:
+		return -ETIMEDOUT;
+	case RPMI_ERR_IO:
+		return -ECOMM;
+	case RPMI_ERR_FAILED:
+	case RPMI_ERR_NOTSUPP:
+	case RPMI_ERR_NO_DATA:
+	case RPMI_ERR_EXTENSION:
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+/* RPMI linux mailbox attribute IDs */
+enum rpmi_mbox_attribute_id {
+	RPMI_MBOX_ATTR_SPEC_VERSION = 0,
+	RPMI_MBOX_ATTR_MAX_MSG_DATA_SIZE,
+	RPMI_MBOX_ATTR_SERVICEGROUP_ID,
+	RPMI_MBOX_ATTR_SERVICEGROUP_VERSION,
+	RPMI_MBOX_ATTR_MAX_ID
+};
+
+/* RPMI linux mailbox message types */
+enum rpmi_mbox_message_type {
+	RPMI_MBOX_MSG_TYPE_GET_ATTRIBUTE = 0,
+	RPMI_MBOX_MSG_TYPE_SET_ATTRIBUTE,
+	RPMI_MBOX_MSG_TYPE_SEND_WITH_RESPONSE,
+	RPMI_MBOX_MSG_TYPE_SEND_WITHOUT_RESPONSE,
+	RPMI_MBOX_MSG_TYPE_NOTIFICATION_EVENT,
+	RPMI_MBOX_MSG_MAX_TYPE
+};
+
+/* RPMI linux mailbox message instance */
+struct rpmi_mbox_message {
+	enum rpmi_mbox_message_type type;
+	union {
+		struct {
+			enum rpmi_mbox_attribute_id id;
+			u32 value;
+		} attr;
+
+		struct {
+			u32 service_id;
+			void *request;
+			unsigned long request_len;
+			void *response;
+			unsigned long max_response_len;
+			unsigned long out_response_len;
+		} data;
+
+		struct {
+			u16 event_datalen;
+			u8 event_id;
+			u8 *event_data;
+		} notif;
+	};
+	int error;
+};
+
+/* RPMI linux mailbox message helper routines */
+static inline void rpmi_mbox_init_get_attribute(struct rpmi_mbox_message *msg,
+						enum rpmi_mbox_attribute_id id)
+{
+	msg->type = RPMI_MBOX_MSG_TYPE_GET_ATTRIBUTE;
+	msg->attr.id = id;
+	msg->attr.value = 0;
+	msg->error = 0;
+}
+
+static inline void rpmi_mbox_init_set_attribute(struct rpmi_mbox_message *msg,
+						enum rpmi_mbox_attribute_id id,
+						u32 value)
+{
+	msg->type = RPMI_MBOX_MSG_TYPE_SET_ATTRIBUTE;
+	msg->attr.id = id;
+	msg->attr.value = value;
+	msg->error = 0;
+}
+
+static inline void rpmi_mbox_init_send_with_response(struct rpmi_mbox_message *msg,
+						     u32 service_id,
+						     void *request,
+						     unsigned long request_len,
+						     void *response,
+						     unsigned long max_response_len)
+{
+	msg->type = RPMI_MBOX_MSG_TYPE_SEND_WITH_RESPONSE;
+	msg->data.service_id = service_id;
+	msg->data.request = request;
+	msg->data.request_len = request_len;
+	msg->data.response = response;
+	msg->data.max_response_len = max_response_len;
+	msg->data.out_response_len = 0;
+	msg->error = 0;
+}
+
+static inline void rpmi_mbox_init_send_without_response(struct rpmi_mbox_message *msg,
+							u32 service_id,
+							void *request,
+							unsigned long request_len)
+{
+	msg->type = RPMI_MBOX_MSG_TYPE_SEND_WITHOUT_RESPONSE;
+	msg->data.service_id = service_id;
+	msg->data.request = request;
+	msg->data.request_len = request_len;
+	msg->data.response = NULL;
+	msg->data.max_response_len = 0;
+	msg->data.out_response_len = 0;
+	msg->error = 0;
+}
+
+static inline int rpmi_mbox_send_message(struct mbox_chan *chan,
+					 struct rpmi_mbox_message *msg)
+{
+	int ret;
+
+	/* Send message for the underlying mailbox channel */
+	ret = mbox_send_message(chan, msg);
+	if (ret < 0)
+		return ret;
+
+	/* Explicitly signal txdone for mailbox channel */
+	ret = msg->error;
+	mbox_client_txdone(chan, ret);
+	return ret;
+}
+
+#endif /* _LINUX_RISCV_RPMI_MESSAGE_H_ */
-- 
2.43.0

[PATCH v4 06/23] mailbox: Allow controller specific mapping using fwnode

[Anup Patel]

Introduce optional fw_node() callback which allows a mailbox controller
driver to provide controller specific mapping using fwnode.

The Linux OF framework already implements fwnode operations for the
Linux DD framework so the fw_xlate() callback works fine with device
tree as well.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 drivers/mailbox/mailbox.c          | 47 +++++++++++++++++-------------
 include/linux/mailbox_controller.h |  3 ++
 2 files changed, 30 insertions(+), 20 deletions(-)

diff --git a/drivers/mailbox/mailbox.c b/drivers/mailbox/mailbox.c
index 0593b4d03685..0100163c0390 100644
--- a/drivers/mailbox/mailbox.c
+++ b/drivers/mailbox/mailbox.c
@@ -399,35 +399,49 @@ EXPORT_SYMBOL_GPL(mbox_bind_client);
  */
 struct mbox_chan *mbox_request_channel(struct mbox_client *cl, int index)
 {
+	struct fwnode_reference_args fwspec;
 	struct device *dev = cl->dev;
 	struct mbox_controller *mbox;
 	struct of_phandle_args spec;
 	struct mbox_chan *chan;
-	int ret;
+	int i, ret;
 
-	if (!dev || !dev->of_node) {
-		pr_debug("%s: No owner device node\n", __func__);
+	if (!dev || !dev->fwnode) {
+		pr_debug("%s: No owner %s\n", __func__, !dev ? "device" : "fwnode");
 		return ERR_PTR(-ENODEV);
 	}
 
-	ret = of_parse_phandle_with_args(dev->of_node, "mboxes", "#mbox-cells",
-					 index, &spec);
+	ret = fwnode_property_get_reference_args(dev->fwnode, "mboxes",
+						 "#mbox-cells", 0, index, &fwspec);
 	if (ret) {
 		dev_dbg(dev, "%s: can't parse \"mboxes\" property\n", __func__);
 		return ERR_PTR(ret);
 	}
 
+	memset(&spec, 0, sizeof(spec));
+	if (dev->of_node) {
+		spec.np = to_of_node(fwspec.fwnode);
+		spec.args_count = fwspec.nargs;
+		for (i = 0; i < spec.args_count; i++)
+			spec.args[i] = fwspec.args[i];
+	}
+
 	mutex_lock(&con_mutex);
 
 	chan = ERR_PTR(-EPROBE_DEFER);
-	list_for_each_entry(mbox, &mbox_cons, node)
-		if (mbox->dev->of_node == spec.np) {
+	list_for_each_entry(mbox, &mbox_cons, node) {
+		if (mbox->fw_xlate && mbox->dev->fwnode == fwspec.fwnode) {
+			chan = mbox->fw_xlate(mbox, &fwspec);
+			if (!IS_ERR(chan))
+				break;
+		} else if (mbox->of_xlate && mbox->dev->of_node == spec.np) {
 			chan = mbox->of_xlate(mbox, &spec);
 			if (!IS_ERR(chan))
 				break;
 		}
+	}
 
-	of_node_put(spec.np);
+	fwnode_handle_put(fwspec.fwnode);
 
 	if (IS_ERR(chan)) {
 		mutex_unlock(&con_mutex);
@@ -446,15 +460,8 @@ EXPORT_SYMBOL_GPL(mbox_request_channel);
 struct mbox_chan *mbox_request_channel_byname(struct mbox_client *cl,
 					      const char *name)
 {
-	struct device_node *np = cl->dev->of_node;
-	int index;
-
-	if (!np) {
-		dev_err(cl->dev, "%s() currently only supports DT\n", __func__);
-		return ERR_PTR(-EINVAL);
-	}
+	int index = device_property_match_string(cl->dev, "mbox-names", name);
 
-	index = of_property_match_string(np, "mbox-names", name);
 	if (index < 0) {
 		dev_err(cl->dev, "%s() could not locate channel named \"%s\"\n",
 			__func__, name);
@@ -492,8 +499,8 @@ void mbox_free_channel(struct mbox_chan *chan)
 EXPORT_SYMBOL_GPL(mbox_free_channel);
 
 static struct mbox_chan *
-of_mbox_index_xlate(struct mbox_controller *mbox,
-		    const struct of_phandle_args *sp)
+fw_mbox_index_xlate(struct mbox_controller *mbox,
+		    const struct fwnode_reference_args *sp)
 {
 	int ind = sp->args[0];
 
@@ -544,8 +551,8 @@ int mbox_controller_register(struct mbox_controller *mbox)
 		spin_lock_init(&chan->lock);
 	}
 
-	if (!mbox->of_xlate)
-		mbox->of_xlate = of_mbox_index_xlate;
+	if (!mbox->fw_xlate && !mbox->of_xlate)
+		mbox->fw_xlate = fw_mbox_index_xlate;
 
 	mutex_lock(&con_mutex);
 	list_add_tail(&mbox->node, &mbox_cons);
diff --git a/include/linux/mailbox_controller.h b/include/linux/mailbox_controller.h
index 5fb0b65f45a2..b91379922cb3 100644
--- a/include/linux/mailbox_controller.h
+++ b/include/linux/mailbox_controller.h
@@ -66,6 +66,7 @@ struct mbox_chan_ops {
  *			no interrupt rises. Ignored if 'txdone_irq' is set.
  * @txpoll_period:	If 'txdone_poll' is in effect, the API polls for
  *			last TX's status after these many millisecs
+ * @fw_xlate:		Controller driver specific mapping of channel via fwnode
  * @of_xlate:		Controller driver specific mapping of channel via DT
  * @poll_hrt:		API private. hrtimer used to poll for TXDONE on all
  *			channels.
@@ -79,6 +80,8 @@ struct mbox_controller {
 	bool txdone_irq;
 	bool txdone_poll;
 	unsigned txpoll_period;
+	struct mbox_chan *(*fw_xlate)(struct mbox_controller *mbox,
+				      const struct fwnode_reference_args *sp);
 	struct mbox_chan *(*of_xlate)(struct mbox_controller *mbox,
 				      const struct of_phandle_args *sp);
 	/* Internal to API */
-- 
2.43.0

[PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Anup Patel]

Add a mailbox controller driver for the new SBI message proxy extension
which is part of the SBI v3.0 specification.

Co-developed-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 drivers/mailbox/Kconfig               |  11 +
 drivers/mailbox/Makefile              |   2 +
 drivers/mailbox/riscv-sbi-mpxy-mbox.c | 979 ++++++++++++++++++++++++++
 3 files changed, 992 insertions(+)
 create mode 100644 drivers/mailbox/riscv-sbi-mpxy-mbox.c

diff --git a/drivers/mailbox/Kconfig b/drivers/mailbox/Kconfig
index ed52db272f4d..cc29a1a1974a 100644
--- a/drivers/mailbox/Kconfig
+++ b/drivers/mailbox/Kconfig
@@ -330,4 +330,15 @@ config THEAD_TH1520_MBOX
 	  kernel is running, and E902 core used for power management among other
 	  things.
 
+config RISCV_SBI_MPXY_MBOX
+	tristate "RISC-V SBI Message Proxy (MPXY) Mailbox"
+	depends on RISCV_SBI
+	default RISCV
+	help
+	  Mailbox driver implementation for RISC-V SBI Message Proxy (MPXY)
+	  extension. This mailbox driver is used to send messages to the
+	  remote processor through the SBI implementation (M-mode firmware
+	  or HS-mode hypervisor). Say Y here if you want to have this support.
+	  If unsure say N.
+
 endif
diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
index 9a1542b55539..833d72649790 100644
--- a/drivers/mailbox/Makefile
+++ b/drivers/mailbox/Makefile
@@ -70,3 +70,5 @@ obj-$(CONFIG_QCOM_CPUCP_MBOX)	+= qcom-cpucp-mbox.o
 obj-$(CONFIG_QCOM_IPCC)		+= qcom-ipcc.o
 
 obj-$(CONFIG_THEAD_TH1520_MBOX)	+= mailbox-th1520.o
+
+obj-$(CONFIG_RISCV_SBI_MPXY_MBOX)	+= riscv-sbi-mpxy-mbox.o
diff --git a/drivers/mailbox/riscv-sbi-mpxy-mbox.c b/drivers/mailbox/riscv-sbi-mpxy-mbox.c
new file mode 100644
index 000000000000..121ee5fd3d0d
--- /dev/null
+++ b/drivers/mailbox/riscv-sbi-mpxy-mbox.c
@@ -0,0 +1,979 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * RISC-V SBI Message Proxy (MPXY) mailbox controller driver
+ *
+ * Copyright (C) 2025 Ventana Micro Systems Inc.
+ */
+
+#include <asm/sbi.h>
+#include <linux/cpu.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/jump_label.h>
+#include <linux/kernel.h>
+#include <linux/mailbox_controller.h>
+#include <linux/mailbox/riscv-rpmi-message.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/msi.h>
+#include <linux/of_irq.h>
+#include <linux/percpu.h>
+#include <linux/platform_device.h>
+#include <linux/smp.h>
+
+/* ====== SBI MPXY extension data structures ====== */
+
+/* SBI MPXY MSI related channel attributes */
+struct sbi_mpxy_msi_info {
+	/* Lower 32-bits of the MSI target address */
+	u32 msi_addr_lo;
+	/* Upper 32-bits of the MSI target address */
+	u32 msi_addr_hi;
+	/* MSI data value */
+	u32 msi_data;
+};
+
+/*
+ * SBI MPXY standard channel attributes.
+ *
+ * NOTE: The sequence of attribute fields are as-per the
+ * defined sequence in the attribute table in spec (or
+ * as-per the enum sbi_mpxy_attribute_id).
+ */
+struct sbi_mpxy_channel_attrs {
+	/* Message protocol ID */
+	u32 msg_proto_id;
+	/* Message protocol version */
+	u32 msg_proto_version;
+	/* Message protocol maximum message length */
+	u32 msg_max_len;
+	/* Message protocol message send timeout in microseconds */
+	u32 msg_send_timeout;
+	/* Message protocol message completion timeout in microseconds */
+	u32 msg_completion_timeout;
+	/* Bit array for channel capabilities */
+	u32 capability;
+	/* SSE event ID */
+	u32 sse_event_id;
+	/* MSI enable/disable control knob */
+	u32 msi_control;
+	/* Channel MSI info */
+	struct sbi_mpxy_msi_info msi_info;
+	/* Events state control */
+	u32 events_state_ctrl;
+};
+
+/*
+ * RPMI specific SBI MPXY channel attributes.
+ *
+ * NOTE: The sequence of attribute fields are as-per the
+ * defined sequence in the attribute table in spec (or
+ * as-per the enum sbi_mpxy_rpmi_attribute_id).
+ */
+struct sbi_mpxy_rpmi_channel_attrs {
+	/* RPMI service group ID */
+	u32 servicegroup_id;
+	/* RPMI service group version */
+	u32 servicegroup_version;
+};
+
+/* SBI MPXY channel IDs data in shared memory */
+struct sbi_mpxy_channel_ids_data {
+	/* Remaining number of channel ids */
+	__le32 remaining;
+	/* Returned channel ids in current function call */
+	__le32 returned;
+	/* Returned channel id array */
+	__le32 channel_array[];
+};
+
+/* SBI MPXY notification data in shared memory */
+struct sbi_mpxy_notification_data {
+	/* Remaining number of notification events */
+	__le32 remaining;
+	/* Number of notification events returned */
+	__le32 returned;
+	/* Number of notification events lost */
+	__le32 lost;
+	/* Reserved for future use */
+	__le32 reserved;
+	/* Returned channel id array */
+	u8 events_data[];
+};
+
+/* ====== MPXY data structures & helper routines ====== */
+
+/* MPXY Per-CPU or local context */
+struct mpxy_local {
+	/* Shared memory base address */
+	void *shmem;
+	/* Shared memory physical address */
+	phys_addr_t shmem_phys_addr;
+	/* Flag representing whether shared memory is active or not */
+	bool shmem_active;
+};
+
+static DEFINE_PER_CPU(struct mpxy_local, mpxy_local);
+static unsigned long mpxy_shmem_size;
+static bool mpxy_shmem_init_done;
+
+static int mpxy_get_channel_count(u32 *channel_count)
+{
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	struct sbi_mpxy_channel_ids_data *sdata = mpxy->shmem;
+	u32 remaining, returned;
+	struct sbiret sret;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!channel_count)
+		return -EINVAL;
+
+	get_cpu();
+
+	/* Get the remaining and returned fields to calculate total */
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_CHANNEL_IDS,
+			 0, 0, 0, 0, 0, 0);
+	if (!sret.error) {
+		remaining = le32_to_cpu(sdata->remaining);
+		returned = le32_to_cpu(sdata->returned);
+		*channel_count = remaining + returned;
+	}
+
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static int mpxy_get_channel_ids(u32 channel_count, u32 *channel_ids)
+{
+	u32 remaining, returned, sidx, start_index = 0, cidx = 0;
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	struct sbi_mpxy_channel_ids_data *sdata = mpxy->shmem;
+	struct sbiret sret;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!channel_count || !channel_ids)
+		return -EINVAL;
+
+	get_cpu();
+
+	do {
+		sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_CHANNEL_IDS,
+				 start_index, 0, 0, 0, 0, 0);
+		if (sret.error)
+			goto done;
+
+		remaining = le32_to_cpu(sdata->remaining);
+		returned = le32_to_cpu(sdata->returned);
+
+		for (sidx = 0; sidx < returned && cidx < channel_count; sidx++) {
+			channel_ids[cidx] = le32_to_cpu(sdata->channel_array[sidx]);
+			cidx += 1;
+		}
+
+		start_index = cidx;
+
+	} while (remaining);
+
+done:
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static int mpxy_read_attrs(u32 channel_id, u32 base_attrid, u32 attr_count,
+			   u32 *attrs_buf)
+{
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	struct sbiret sret;
+	u32 i;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!attr_count || !attrs_buf)
+		return -EINVAL;
+
+	get_cpu();
+
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_READ_ATTRS,
+			 channel_id, base_attrid, attr_count, 0, 0, 0);
+	if (!sret.error) {
+		for (i = 0; i < attr_count; i++)
+			attrs_buf[i] = le32_to_cpu(((__le32 *)mpxy->shmem)[i]);
+	}
+
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static int mpxy_write_attrs(u32 channel_id, u32 base_attrid, u32 attr_count,
+			    u32 *attrs_buf)
+{
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	struct sbiret sret;
+	u32 i;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!attr_count || !attrs_buf)
+		return -EINVAL;
+
+	get_cpu();
+
+	for (i = 0; i < attr_count; i++)
+		((__le32 *)mpxy->shmem)[i] = cpu_to_le32(attrs_buf[i]);
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_WRITE_ATTRS,
+			 channel_id, base_attrid, attr_count, 0, 0, 0);
+
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static int mpxy_send_message_with_resp(u32 channel_id, u32 msg_id,
+				       void *tx, unsigned long tx_len,
+				       void *rx, unsigned long max_rx_len,
+				       unsigned long *rx_len)
+{
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	unsigned long rx_bytes;
+	struct sbiret sret;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!tx && tx_len)
+		return -EINVAL;
+
+	get_cpu();
+
+	/* Message protocols allowed to have no data in messages */
+	if (tx_len)
+		memcpy(mpxy->shmem, tx, tx_len);
+
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_SEND_MSG_WITH_RESP,
+			 channel_id, msg_id, tx_len, 0, 0, 0);
+	if (rx && !sret.error) {
+		rx_bytes = sret.value;
+		if (rx_bytes > max_rx_len) {
+			put_cpu();
+			return -ENOSPC;
+		}
+
+		memcpy(rx, mpxy->shmem, rx_bytes);
+		if (rx_len)
+			*rx_len = rx_bytes;
+	}
+
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static int mpxy_send_message_without_resp(u32 channel_id, u32 msg_id,
+					  void *tx, unsigned long tx_len)
+{
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	struct sbiret sret;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!tx && tx_len)
+		return -EINVAL;
+
+	get_cpu();
+
+	/* Message protocols allowed to have no data in messages */
+	if (tx_len)
+		memcpy(mpxy->shmem, tx, tx_len);
+
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_SEND_MSG_WITHOUT_RESP,
+			 channel_id, msg_id, tx_len, 0, 0, 0);
+
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static int mpxy_get_notifications(u32 channel_id,
+				  struct sbi_mpxy_notification_data *notif_data,
+				  unsigned long *events_data_len)
+{
+	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
+	struct sbiret sret;
+
+	if (!mpxy->shmem_active)
+		return -ENODEV;
+	if (!notif_data || !events_data_len)
+		return -EINVAL;
+
+	get_cpu();
+
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_NOTIFICATION_EVENTS,
+			 channel_id, 0, 0, 0, 0, 0);
+	if (!sret.error) {
+		memcpy(notif_data, mpxy->shmem, sret.value + 16);
+		*events_data_len = sret.value;
+	}
+
+	put_cpu();
+	return sbi_err_map_linux_errno(sret.error);
+}
+
+static unsigned long mpxy_get_shmem_size(void)
+{
+	struct sbiret sret;
+
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_SHMEM_SIZE,
+			 0, 0, 0, 0, 0, 0);
+	if (sret.error)
+		return sbi_err_map_linux_errno(sret.error);
+
+	return sret.value;
+}
+
+static int mpxy_setup_shmem(unsigned int cpu)
+{
+	struct page *shmem_page;
+	struct mpxy_local *mpxy;
+	struct sbiret sret;
+
+	mpxy = per_cpu_ptr(&mpxy_local, cpu);
+	if (mpxy->shmem_active)
+		return 0;
+
+	shmem_page = alloc_pages(GFP_KERNEL | __GFP_ZERO, get_order(mpxy_shmem_size));
+	if (!shmem_page)
+		return -ENOMEM;
+
+	/*
+	 * Linux setup of shmem is done in mpxy OVERWRITE mode.
+	 * flags[1:0] = 00b
+	 */
+	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_SET_SHMEM,
+			 page_to_phys(shmem_page), 0, 0, 0, 0, 0);
+	if (sret.error) {
+		free_pages((unsigned long)page_to_virt(shmem_page),
+			   get_order(mpxy_shmem_size));
+		return sbi_err_map_linux_errno(sret.error);
+	}
+
+	mpxy->shmem = page_to_virt(shmem_page);
+	mpxy->shmem_phys_addr = page_to_phys(shmem_page);
+	mpxy->shmem_active = true;
+
+	return 0;
+}
+
+/* ====== MPXY mailbox data structures ====== */
+
+/* MPXY mailbox channel */
+struct mpxy_mbox_channel {
+	struct mpxy_mbox *mbox;
+	u32 channel_id;
+	struct sbi_mpxy_channel_attrs attrs;
+	struct sbi_mpxy_rpmi_channel_attrs rpmi_attrs;
+	struct sbi_mpxy_notification_data *notif;
+	u32 max_xfer_len;
+	bool have_events_state;
+	u32 msi_index;
+	u32 msi_irq;
+	bool started;
+};
+
+/* MPXY mailbox */
+struct mpxy_mbox {
+	struct device *dev;
+	u32 channel_count;
+	struct mpxy_mbox_channel *channels;
+	u32 msi_count;
+	struct mpxy_mbox_channel **msi_index_to_channel;
+	struct mbox_controller controller;
+};
+
+/* ====== MPXY RPMI processing ====== */
+
+static int mpxy_mbox_send_rpmi_data(struct mpxy_mbox_channel *mchan,
+				    struct rpmi_mbox_message *msg)
+{
+	int rc = 0;
+
+	switch (msg->type) {
+	case RPMI_MBOX_MSG_TYPE_GET_ATTRIBUTE:
+		switch (msg->attr.id) {
+		case RPMI_MBOX_ATTR_SPEC_VERSION:
+			msg->attr.value = mchan->attrs.msg_proto_version;
+			break;
+		case RPMI_MBOX_ATTR_MAX_MSG_DATA_SIZE:
+			msg->attr.value = mchan->max_xfer_len;
+			break;
+		case RPMI_MBOX_ATTR_SERVICEGROUP_ID:
+			msg->attr.value = mchan->rpmi_attrs.servicegroup_id;
+			break;
+		case RPMI_MBOX_ATTR_SERVICEGROUP_VERSION:
+			msg->attr.value = mchan->rpmi_attrs.servicegroup_version;
+			break;
+		default:
+			rc = -EOPNOTSUPP;
+			break;
+		}
+		break;
+	case RPMI_MBOX_MSG_TYPE_SET_ATTRIBUTE:
+		/* None of the RPMI linux mailbox attributes are writeable */
+		rc = -EOPNOTSUPP;
+		break;
+	case RPMI_MBOX_MSG_TYPE_SEND_WITH_RESPONSE:
+		if ((!msg->data.request && msg->data.request_len) ||
+		    (msg->data.request &&
+		     msg->data.request_len > mchan->max_xfer_len) ||
+		    (!msg->data.response && msg->data.max_response_len)) {
+			rc = -EINVAL;
+			break;
+		}
+		if (!(mchan->attrs.capability & SBI_MPXY_CHAN_CAP_SEND_WITH_RESP)) {
+			rc = -EIO;
+			break;
+		}
+		rc = mpxy_send_message_with_resp(mchan->channel_id,
+						 msg->data.service_id,
+						 msg->data.request,
+						 msg->data.request_len,
+						 msg->data.response,
+						 msg->data.max_response_len,
+						 &msg->data.out_response_len);
+		break;
+	case RPMI_MBOX_MSG_TYPE_SEND_WITHOUT_RESPONSE:
+		if ((!msg->data.request && msg->data.request_len) ||
+		    (msg->data.request &&
+		     msg->data.request_len > mchan->max_xfer_len)) {
+			rc = -EINVAL;
+			break;
+		}
+		if (!(mchan->attrs.capability & SBI_MPXY_CHAN_CAP_SEND_WITHOUT_RESP)) {
+			rc = -EIO;
+			break;
+		}
+		rc = mpxy_send_message_without_resp(mchan->channel_id,
+						    msg->data.service_id,
+						    msg->data.request,
+						    msg->data.request_len);
+		break;
+	default:
+		rc = -EOPNOTSUPP;
+		break;
+	}
+
+	msg->error = rc;
+	return 0;
+}
+
+static void mpxy_mbox_peek_rpmi_data(struct mbox_chan *chan,
+				     struct mpxy_mbox_channel *mchan,
+				     struct sbi_mpxy_notification_data *notif,
+				     unsigned long events_data_len)
+{
+	struct rpmi_notification_event *event;
+	unsigned long pos = 0, event_size;
+	struct rpmi_mbox_message msg;
+
+	while ((pos < events_data_len) && !(pos & 0x3) &&
+	       ((events_data_len - pos) <= sizeof(*event))) {
+		event = (struct rpmi_notification_event *)(notif->events_data + pos);
+
+		msg.type = RPMI_MBOX_MSG_TYPE_NOTIFICATION_EVENT;
+		msg.notif.event_datalen = le16_to_cpu(event->event_datalen);
+		msg.notif.event_id = event->event_id;
+		msg.notif.event_data = event->event_data;
+		msg.error = 0;
+
+		event_size = sizeof(*event) + msg.notif.event_datalen;
+		if (event_size > (events_data_len - pos)) {
+			event_size = events_data_len - pos;
+			goto skip_event;
+		}
+		if (event_size & 0x3)
+			goto skip_event;
+
+		mbox_chan_received_data(chan, &msg);
+
+skip_event:
+		pos += event_size;
+	}
+}
+
+static int mpxy_mbox_read_rpmi_attrs(struct mpxy_mbox_channel *mchan)
+{
+	return mpxy_read_attrs(mchan->channel_id,
+			       SBI_MPXY_ATTR_MSGPROTO_ATTR_START,
+			       sizeof(mchan->rpmi_attrs) / sizeof(u32),
+			       (u32 *)&mchan->rpmi_attrs);
+}
+
+/* ====== MPXY mailbox callbacks ====== */
+
+static int mpxy_mbox_send_data(struct mbox_chan *chan, void *data)
+{
+	struct mpxy_mbox_channel *mchan = chan->con_priv;
+
+	if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID)
+		return mpxy_mbox_send_rpmi_data(mchan, data);
+
+	return -EOPNOTSUPP;
+}
+
+static bool mpxy_mbox_peek_data(struct mbox_chan *chan)
+{
+	struct mpxy_mbox_channel *mchan = chan->con_priv;
+	struct sbi_mpxy_notification_data *notif = mchan->notif;
+	bool have_notifications = false;
+	unsigned long data_len;
+	int rc;
+
+	if (!(mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS))
+		return false;
+
+	while (1) {
+		rc = mpxy_get_notifications(mchan->channel_id, notif, &data_len);
+		if (rc || !data_len)
+			break;
+
+		if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID)
+			mpxy_mbox_peek_rpmi_data(chan, mchan, notif, data_len);
+
+		have_notifications = true;
+	}
+
+	return have_notifications;
+}
+
+static irqreturn_t mpxy_mbox_irq_thread(int irq, void *dev_id)
+{
+	mpxy_mbox_peek_data(dev_id);
+	return IRQ_HANDLED;
+}
+
+static int mpxy_mbox_setup_msi(struct mbox_chan *chan,
+			       struct mpxy_mbox_channel *mchan)
+{
+	struct device *dev = mchan->mbox->dev;
+	int rc;
+
+	/* Do nothing if MSI not supported */
+	if (mchan->msi_irq == U32_MAX)
+		return 0;
+
+	/* Fail if MSI already enabled */
+	if (mchan->attrs.msi_control)
+		return -EALREADY;
+
+	/* Request channel MSI handler */
+	rc = request_threaded_irq(mchan->msi_irq, NULL, mpxy_mbox_irq_thread,
+				  0, dev_name(dev), chan);
+	if (rc) {
+		dev_err(dev, "failed to request MPXY channel 0x%x IRQ\n",
+			mchan->channel_id);
+		return rc;
+	}
+
+	/* Enable channel MSI control */
+	mchan->attrs.msi_control = 1;
+	rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_MSI_CONTROL,
+			      1, &mchan->attrs.msi_control);
+	if (rc) {
+		dev_err(dev, "enable MSI control failed for MPXY channel 0x%x\n",
+			mchan->channel_id);
+		mchan->attrs.msi_control = 0;
+		free_irq(mchan->msi_irq, chan);
+		return rc;
+	}
+
+	return 0;
+}
+
+static void mpxy_mbox_cleanup_msi(struct mbox_chan *chan,
+				  struct mpxy_mbox_channel *mchan)
+{
+	struct device *dev = mchan->mbox->dev;
+	int rc;
+
+	/* Do nothing if MSI not supported */
+	if (mchan->msi_irq == U32_MAX)
+		return;
+
+	/* Do nothing if MSI already disabled */
+	if (!mchan->attrs.msi_control)
+		return;
+
+	/* Disable channel MSI control */
+	mchan->attrs.msi_control = 0;
+	rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_MSI_CONTROL,
+			      1, &mchan->attrs.msi_control);
+	if (rc) {
+		dev_err(dev, "disable MSI control failed for MPXY channel 0x%x\n",
+			mchan->channel_id);
+	}
+
+	/* Free channel MSI handler */
+	free_irq(mchan->msi_irq, chan);
+}
+
+static int mpxy_mbox_setup_events(struct mpxy_mbox_channel *mchan)
+{
+	struct device *dev = mchan->mbox->dev;
+	int rc;
+
+	/* Do nothing if events state not supported */
+	if (!mchan->have_events_state)
+		return 0;
+
+	/* Fail if events state already enabled */
+	if (mchan->attrs.events_state_ctrl)
+		return -EALREADY;
+
+	/* Enable channel events state */
+	mchan->attrs.events_state_ctrl = 1;
+	rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_EVENTS_STATE_CONTROL,
+			      1, &mchan->attrs.events_state_ctrl);
+	if (rc) {
+		dev_err(dev, "enable events state failed for MPXY channel 0x%x\n",
+			mchan->channel_id);
+		mchan->attrs.events_state_ctrl = 0;
+		return rc;
+	}
+
+	return 0;
+}
+
+static void mpxy_mbox_cleanup_events(struct mpxy_mbox_channel *mchan)
+{
+	struct device *dev = mchan->mbox->dev;
+	int rc;
+
+	/* Do nothing if events state not supported */
+	if (!mchan->have_events_state)
+		return;
+
+	/* Do nothing if events state already disabled */
+	if (!mchan->attrs.events_state_ctrl)
+		return;
+
+	/* Disable channel events state */
+	mchan->attrs.events_state_ctrl = 0;
+	rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_EVENTS_STATE_CONTROL,
+			      1, &mchan->attrs.events_state_ctrl);
+	if (rc) {
+		dev_err(dev, "disable events state failed for MPXY channel 0x%x\n",
+			mchan->channel_id);
+	}
+}
+
+static int mpxy_mbox_startup(struct mbox_chan *chan)
+{
+	struct mpxy_mbox_channel *mchan = chan->con_priv;
+	int rc;
+
+	if (mchan->started)
+		return -EALREADY;
+
+	/* Setup channel MSI */
+	rc = mpxy_mbox_setup_msi(chan, mchan);
+	if (rc)
+		return rc;
+
+	/* Setup channel notification events */
+	rc = mpxy_mbox_setup_events(mchan);
+	if (rc) {
+		mpxy_mbox_cleanup_msi(chan, mchan);
+		return rc;
+	}
+
+	/* Mark the channel as started */
+	mchan->started = true;
+
+	return 0;
+}
+
+static void mpxy_mbox_shutdown(struct mbox_chan *chan)
+{
+	struct mpxy_mbox_channel *mchan = chan->con_priv;
+
+	if (!mchan->started)
+		return;
+
+	/* Mark the channel as stopped */
+	mchan->started = false;
+
+	/* Cleanup channel notification events */
+	mpxy_mbox_cleanup_events(mchan);
+
+	/* Cleanup channel MSI */
+	mpxy_mbox_cleanup_msi(chan, mchan);
+}
+
+static const struct mbox_chan_ops mpxy_mbox_ops = {
+	.send_data = mpxy_mbox_send_data,
+	.peek_data = mpxy_mbox_peek_data,
+	.startup = mpxy_mbox_startup,
+	.shutdown = mpxy_mbox_shutdown,
+};
+
+/* ====== MPXY platform driver ===== */
+
+static void mpxy_mbox_msi_write(struct msi_desc *desc, struct msi_msg *msg)
+{
+	struct device *dev = msi_desc_to_dev(desc);
+	struct mpxy_mbox *mbox = dev_get_drvdata(dev);
+	struct mpxy_mbox_channel *mchan;
+	struct sbi_mpxy_msi_info *minfo;
+	int rc;
+
+	mchan = mbox->msi_index_to_channel[desc->msi_index];
+	if (!mchan) {
+		dev_warn(dev, "MPXY channel not available for MSI index %d\n",
+			 desc->msi_index);
+		return;
+	}
+
+	minfo = &mchan->attrs.msi_info;
+	minfo->msi_addr_lo = msg->address_lo;
+	minfo->msi_addr_hi = msg->address_hi;
+	minfo->msi_data = msg->data;
+
+	rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_MSI_ADDR_LO,
+			      sizeof(*minfo) / sizeof(u32), (u32 *)minfo);
+	if (rc) {
+		dev_warn(dev, "failed to write MSI info for MPXY channel 0x%x\n",
+			 mchan->channel_id);
+	}
+}
+
+static struct mbox_chan *mpxy_mbox_fw_xlate(struct mbox_controller *ctlr,
+					    const struct fwnode_reference_args *pa)
+{
+	struct mpxy_mbox *mbox = container_of(ctlr, struct mpxy_mbox, controller);
+	struct mpxy_mbox_channel *mchan;
+	u32 i;
+
+	if (pa->nargs != 2)
+		return ERR_PTR(-EINVAL);
+
+	for (i = 0; i < mbox->channel_count; i++) {
+		mchan = &mbox->channels[i];
+		if (mchan->channel_id == pa->args[0] &&
+		    mchan->attrs.msg_proto_id == pa->args[1])
+			return &mbox->controller.chans[i];
+	}
+
+	return ERR_PTR(-ENOENT);
+}
+
+static int mpxy_mbox_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct mpxy_mbox_channel *mchan;
+	struct mpxy_mbox *mbox;
+	int i, msi_idx, rc;
+	u32 *channel_ids;
+
+	/*
+	 * Initialize MPXY shared memory only once. This also ensures
+	 * that SBI MPXY mailbox is probed only once.
+	 */
+	if (mpxy_shmem_init_done) {
+		dev_err(dev, "SBI MPXY mailbox already initialized\n");
+		return -EALREADY;
+	}
+
+	/* Probe for SBI MPXY extension */
+	if (sbi_spec_version < sbi_mk_version(1, 0) ||
+	    sbi_probe_extension(SBI_EXT_MPXY) <= 0) {
+		dev_info(dev, "SBI MPXY extension not available\n");
+		return -ENODEV;
+	}
+
+	/* Find-out shared memory size */
+	mpxy_shmem_size = mpxy_get_shmem_size();
+
+	/*
+	 * Setup MPXY shared memory on each CPU
+	 *
+	 * Note: Don't cleanup MPXY shared memory upon CPU power-down
+	 * because the RPMI System MSI irqchip driver needs it to be
+	 * available when migrating IRQs in CPU power-down path.
+	 */
+	cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "riscv/sbi-mpxy-shmem",
+			  mpxy_setup_shmem, NULL);
+
+	/* Mark as MPXY shared memory initialization done */
+	mpxy_shmem_init_done = true;
+
+	/* Allocate mailbox instance */
+	mbox = devm_kzalloc(dev, sizeof(*mbox), GFP_KERNEL);
+	if (!mbox)
+		return -ENOMEM;
+	mbox->dev = dev;
+	platform_set_drvdata(pdev, mbox);
+
+	/* Find-out of number of channels */
+	rc = mpxy_get_channel_count(&mbox->channel_count);
+	if (rc)
+		return dev_err_probe(dev, rc, "failed to get number of MPXY channels\n");
+	if (!mbox->channel_count)
+		dev_err_probe(dev, -ENODEV, "no MPXY channels available\n");
+
+	/* Allocate and fetch all channel IDs */
+	channel_ids = devm_kcalloc(dev, mbox->channel_count,
+				   sizeof(*channel_ids), GFP_KERNEL);
+	if (!channel_ids)
+		return -ENOMEM;
+	rc = mpxy_get_channel_ids(mbox->channel_count, channel_ids);
+	if (rc)
+		return dev_err_probe(dev, rc, "failed to MPXY channel IDs\n");
+
+	/* Populate all channels */
+	mbox->channels = devm_kcalloc(dev, mbox->channel_count,
+				      sizeof(*mbox->channels), GFP_KERNEL);
+	if (!mbox->channels)
+		return -ENOMEM;
+	for (i = 0; i < mbox->channel_count; i++) {
+		mchan = &mbox->channels[i];
+		mchan->mbox = mbox;
+		mchan->channel_id = channel_ids[i];
+
+		rc = mpxy_read_attrs(mchan->channel_id, SBI_MPXY_ATTR_MSG_PROT_ID,
+				     sizeof(mchan->attrs) / sizeof(u32),
+				     (u32 *)&mchan->attrs);
+		if (rc) {
+			return dev_err_probe(dev, rc,
+					     "MPXY channel 0x%x read attrs failed\n",
+					     mchan->channel_id);
+		}
+
+		if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID) {
+			rc = mpxy_mbox_read_rpmi_attrs(mchan);
+			if (rc) {
+				return dev_err_probe(dev, rc,
+						     "MPXY channel 0x%x read RPMI attrs failed\n",
+						     mchan->channel_id);
+			}
+		}
+
+		mchan->notif = devm_kzalloc(dev, mpxy_shmem_size, GFP_KERNEL);
+		if (!mchan->notif)
+			return -ENOMEM;
+
+		mchan->max_xfer_len = min(mpxy_shmem_size, mchan->attrs.msg_max_len);
+
+		if ((mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS) &&
+		    (mchan->attrs.capability & SBI_MPXY_CHAN_CAP_EVENTS_STATE))
+			mchan->have_events_state = true;
+
+		if ((mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS) &&
+		    (mchan->attrs.capability & SBI_MPXY_CHAN_CAP_MSI))
+			mchan->msi_index = mbox->msi_count++;
+		else
+			mchan->msi_index = U32_MAX;
+		mchan->msi_irq = U32_MAX;
+	}
+
+	/* Free-up channel IDs */
+	devm_kfree(dev, channel_ids);
+
+	/* Initialize mailbox controller */
+	mbox->controller.txdone_irq = false;
+	mbox->controller.txdone_poll = false;
+	mbox->controller.ops = &mpxy_mbox_ops;
+	mbox->controller.dev = dev;
+	mbox->controller.num_chans = mbox->channel_count;
+	mbox->controller.fw_xlate = mpxy_mbox_fw_xlate;
+	mbox->controller.chans = devm_kcalloc(dev, mbox->channel_count,
+					      sizeof(*mbox->controller.chans),
+					      GFP_KERNEL);
+	if (!mbox->controller.chans)
+		return -ENOMEM;
+	for (i = 0; i < mbox->channel_count; i++)
+		mbox->controller.chans[i].con_priv = &mbox->channels[i];
+
+	/* Set the MSI domain if not available */
+	if (!dev_get_msi_domain(dev)) {
+		/*
+		 * The device MSI domain for OF devices is only set at the
+		 * time of populating/creating OF device. If the device MSI
+		 * domain is discovered later after the OF device is created
+		 * then we need to set it explicitly before using any platform
+		 * MSI functions.
+		 */
+		if (is_of_node(dev_fwnode(dev)))
+			of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
+	}
+
+	/* Setup MSIs for mailbox (if required) */
+	if (mbox->msi_count) {
+		mbox->msi_index_to_channel = devm_kcalloc(dev, mbox->msi_count,
+							  sizeof(*mbox->msi_index_to_channel),
+							  GFP_KERNEL);
+		if (!mbox->msi_index_to_channel)
+			return -ENOMEM;
+
+		for (msi_idx = 0; msi_idx < mbox->msi_count; msi_idx++) {
+			for (i = 0; i < mbox->channel_count; i++) {
+				mchan = &mbox->channels[i];
+				if (mchan->msi_index == msi_idx) {
+					mbox->msi_index_to_channel[msi_idx] = mchan;
+					break;
+				}
+			}
+		}
+
+		rc = platform_device_msi_init_and_alloc_irqs(dev, mbox->msi_count,
+							     mpxy_mbox_msi_write);
+		if (rc) {
+			return dev_err_probe(dev, rc, "Failed to allocate %d MSIs\n",
+					     mbox->msi_count);
+		}
+
+		for (i = 0; i < mbox->channel_count; i++) {
+			mchan = &mbox->channels[i];
+			if (mchan->msi_index == U32_MAX)
+				continue;
+			mchan->msi_irq = msi_get_virq(dev, mchan->msi_index);
+		}
+	}
+
+	/* Register mailbox controller */
+	rc = devm_mbox_controller_register(dev, &mbox->controller);
+	if (rc) {
+		dev_err_probe(dev, rc, "Registering SBI MPXY mailbox failed\n");
+		if (mbox->msi_count)
+			platform_device_msi_free_irqs_all(dev);
+		return rc;
+	}
+
+	dev_info(dev, "mailbox registered with %d channels\n",
+		 mbox->channel_count);
+	return 0;
+}
+
+static void mpxy_mbox_remove(struct platform_device *pdev)
+{
+	struct mpxy_mbox *mbox = platform_get_drvdata(pdev);
+
+	if (mbox->msi_count)
+		platform_device_msi_free_irqs_all(mbox->dev);
+}
+
+static const struct of_device_id mpxy_mbox_of_match[] = {
+	{.compatible = "riscv,sbi-mpxy-mbox", },
+	{},
+};
+MODULE_DEVICE_TABLE(of, mpxy_mbox_of_match);
+
+static struct platform_driver mpxy_mbox_driver = {
+	.driver = {
+		.name = "riscv-sbi-mpxy-mbox",
+		.of_match_table = mpxy_mbox_of_match,
+	},
+	.probe = mpxy_mbox_probe,
+	.remove = mpxy_mbox_remove,
+};
+module_platform_driver(mpxy_mbox_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Anup Patel <apatel@ventanamicro.com>");
+MODULE_DESCRIPTION("RISC-V SBI MPXY mailbox controller driver");
-- 
2.43.0

[PATCH v4 08/23] dt-bindings: clock: Add RPMI clock service message proxy bindings

[Anup Patel]

Add device tree bindings for the RPMI clock service group based
message proxy implemented by the SBI implementation (machine mode
firmware or hypervisor).

The RPMI clock service group is defined by the RISC-V platform
management interface (RPMI) specification.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 .../bindings/clock/riscv,rpmi-mpxy-clock.yaml | 64 +++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/clock/riscv,rpmi-mpxy-clock.yaml

diff --git a/Documentation/devicetree/bindings/clock/riscv,rpmi-mpxy-clock.yaml b/Documentation/devicetree/bindings/clock/riscv,rpmi-mpxy-clock.yaml
new file mode 100644
index 000000000000..39db52de86b3
--- /dev/null
+++ b/Documentation/devicetree/bindings/clock/riscv,rpmi-mpxy-clock.yaml
@@ -0,0 +1,64 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/clock/riscv,rpmi-mpxy-clock.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: RISC-V RPMI clock service group based message proxy
+
+maintainers:
+  - Anup Patel <anup@brainfault.org>
+
+description: |
+  The RISC-V Platform Management Interface (RPMI) [1] defines a
+  messaging protocol which is modular and extensible. The supervisor
+  software can send/receive RPMI messages via SBI MPXY extension [2]
+  or some dedicated supervisor-mode RPMI transport.
+
+  The RPMI specification [1] defines clock service group for accessing
+  system clocks managed by a platform microcontroller. The SBI implementation
+  (machine mode firmware or hypervisor) can implement an SBI MPXY channel
+  to allow RPMI clock service group access to the supervisor software.
+
+  ===========================================
+  References
+  ===========================================
+
+  [1] RISC-V Platform Management Interface (RPMI)
+      https://github.com/riscv-non-isa/riscv-rpmi/releases
+
+  [2] RISC-V Supervisor Binary Interface (SBI)
+      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
+
+properties:
+  compatible:
+    description:
+      Intended for use by the SBI implementation.
+    const: riscv,rpmi-mpxy-clock
+
+  mboxes:
+    maxItems: 1
+    description:
+      Mailbox channel of the underlying RPMI transport.
+
+  riscv,sbi-mpxy-channel-id:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    description:
+      The SBI MPXY channel id to be used for providing RPMI access to
+      the supervisor software.
+
+required:
+  - compatible
+  - mboxes
+  - riscv,sbi-mpxy-channel-id
+
+additionalProperties: false
+
+examples:
+  - |
+    clock-controller {
+        compatible = "riscv,rpmi-mpxy-clock";
+        mboxes = <&rpmi_shmem_mbox 0x8>;
+        riscv,sbi-mpxy-channel-id = <0x1000>;
+    };
+...
-- 
2.43.0

[PATCH v4 09/23] dt-bindings: clock: Add RPMI clock service controller bindings

[Anup Patel]

Add device tree bindings for the RPMI clock service group based
controller for the supervisor software.

The RPMI clock service group is defined by the RISC-V platform
management interface (RPMI) specification.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 .../bindings/clock/riscv,rpmi-clock.yaml      | 61 +++++++++++++++++++
 1 file changed, 61 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml

diff --git a/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml b/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
new file mode 100644
index 000000000000..9c672a38595a
--- /dev/null
+++ b/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
@@ -0,0 +1,61 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/clock/riscv,rpmi-clock.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: RISC-V RPMI clock service group based clock controller
+
+maintainers:
+  - Anup Patel <anup@brainfault.org>
+
+description: |
+  The RISC-V Platform Management Interface (RPMI) [1] defines a
+  messaging protocol which is modular and extensible. The supervisor
+  software can send/receive RPMI messages via SBI MPXY extension [2]
+  or some dedicated supervisor-mode RPMI transport.
+
+  The RPMI specification [1] defines clock service group for accessing
+  system clocks managed by a platform microcontroller. The supervisor
+  software can access RPMI clock service group via SBI MPXY channel or
+  some dedicated supervisor-mode RPMI transport.
+
+  ===========================================
+  References
+  ===========================================
+
+  [1] RISC-V Platform Management Interface (RPMI)
+      https://github.com/riscv-non-isa/riscv-rpmi/releases
+
+  [2] RISC-V Supervisor Binary Interface (SBI)
+      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
+
+properties:
+  compatible:
+    description:
+      Intended for use by the supervisor software.
+    const: riscv,rpmi-clock
+
+  mboxes:
+    maxItems: 1
+    description:
+      Mailbox channel of the underlying RPMI transport or SBI message proxy channel.
+
+  "#clock-cells":
+    const: 1
+
+required:
+  - compatible
+  - mboxes
+  - "#clock-cells"
+
+additionalProperties: false
+
+examples:
+  - |
+    clock-controller {
+        compatible = "riscv,rpmi-clock";
+        mboxes = <&mpxy_mbox 0x1000 0x0>;
+        #clock-cells = <1>;
+    };
+...
-- 
2.43.0

[PATCH v4 10/23] clk: Add clock driver for the RISC-V RPMI clock service group

[Anup Patel]

From: Rahul Pathak <rpathak@ventanamicro.com>

The RPMI specification defines a clock service group which can be
accessed via SBI MPXY extension or dedicated S-mode RPMI transport.

Add mailbox client based clock driver for the RISC-V RPMI clock
service group.

Co-developed-by: Anup Patel <apatel@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Signed-off-by: Rahul Pathak <rpathak@ventanamicro.com>
---
 drivers/clk/Kconfig                        |   8 +
 drivers/clk/Makefile                       |   1 +
 drivers/clk/clk-rpmi.c                     | 589 +++++++++++++++++++++
 include/linux/mailbox/riscv-rpmi-message.h |  16 +
 include/linux/wordpart.h                   |   8 +
 5 files changed, 622 insertions(+)
 create mode 100644 drivers/clk/clk-rpmi.c

diff --git a/drivers/clk/Kconfig b/drivers/clk/Kconfig
index 713573b6c86c..d89308c7f75c 100644
--- a/drivers/clk/Kconfig
+++ b/drivers/clk/Kconfig
@@ -493,6 +493,14 @@ config COMMON_CLK_SP7021
 	  Not all features of the PLL are currently supported
 	  by the driver.
 
+config COMMON_CLK_RPMI
+	tristate "Clock driver based on RISC-V RPMI"
+	depends on MAILBOX
+	default RISCV
+	help
+	  Support for clocks based on the clock service group defined by
+	  the RISC-V platform management interface (RPMI) specification.
+
 source "drivers/clk/actions/Kconfig"
 source "drivers/clk/analogbits/Kconfig"
 source "drivers/clk/baikal-t1/Kconfig"
diff --git a/drivers/clk/Makefile b/drivers/clk/Makefile
index bf4bd45adc3a..b8588ab789c3 100644
--- a/drivers/clk/Makefile
+++ b/drivers/clk/Makefile
@@ -84,6 +84,7 @@ obj-$(CONFIG_CLK_LS1028A_PLLDIG)	+= clk-plldig.o
 obj-$(CONFIG_COMMON_CLK_PWM)		+= clk-pwm.o
 obj-$(CONFIG_CLK_QORIQ)			+= clk-qoriq.o
 obj-$(CONFIG_COMMON_CLK_RK808)		+= clk-rk808.o
+obj-$(CONFIG_COMMON_CLK_RPMI)		+= clk-rpmi.o
 obj-$(CONFIG_COMMON_CLK_HI655X)		+= clk-hi655x.o
 obj-$(CONFIG_COMMON_CLK_S2MPS11)	+= clk-s2mps11.o
 obj-$(CONFIG_COMMON_CLK_SCMI)           += clk-scmi.o
diff --git a/drivers/clk/clk-rpmi.c b/drivers/clk/clk-rpmi.c
new file mode 100644
index 000000000000..0ab2d916c8be
--- /dev/null
+++ b/drivers/clk/clk-rpmi.c
@@ -0,0 +1,589 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * RISC-V MPXY Based Clock Driver
+ *
+ * Copyright (C) 2025 Ventana Micro Systems Ltd.
+ */
+
+#include <linux/err.h>
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/wordpart.h>
+#include <linux/clk-provider.h>
+#include <linux/mailbox_client.h>
+#include <linux/platform_device.h>
+#include <linux/mailbox/riscv-rpmi-message.h>
+
+#define RPMI_CLK_DISCRETE_MAX_NUM_RATES		16
+#define RPMI_CLK_NAME_LEN			16
+
+#define to_rpmi_clk(clk)	container_of(clk, struct rpmi_clk, hw)
+
+#define rpmi_clkrate_u64(hi, lo)	get_u64_from_u32(hi, lo)
+
+enum rpmi_clk_config {
+	RPMI_CLK_DISABLE = 0,
+	RPMI_CLK_ENABLE = 1
+};
+
+enum rpmi_clk_type {
+	RPMI_CLK_DISCRETE = 0,
+	RPMI_CLK_LINEAR = 1,
+	RPMI_CLK_TYPE_MAX_IDX
+};
+
+struct rpmi_clk_context {
+	struct device *dev;
+	struct mbox_chan *chan;
+	struct mbox_client client;
+	u32 max_msg_data_size;
+};
+
+union rpmi_clk_rates {
+	u64 discrete[RPMI_CLK_DISCRETE_MAX_NUM_RATES];
+	struct {
+		u64 min;
+		u64 max;
+		u64 step;
+	} linear;
+};
+
+struct rpmi_clk {
+	struct rpmi_clk_context *context;
+	u32 id;
+	u32 num_rates;
+	u32 transition_latency;
+	enum rpmi_clk_type type;
+	union rpmi_clk_rates *rates;
+	char name[RPMI_CLK_NAME_LEN];
+	struct clk_hw hw;
+};
+
+struct rpmi_clk_rate_discrete {
+	__le32 lo;
+	__le32 hi;
+};
+
+struct rpmi_clk_rate_linear {
+	__le32 min_lo;
+	__le32 min_hi;
+	__le32 max_lo;
+	__le32 max_hi;
+	__le32 step_lo;
+	__le32 step_hi;
+};
+
+struct rpmi_get_num_clocks_rx {
+	__le32 status;
+	__le32 num_clocks;
+};
+
+struct rpmi_get_attrs_tx {
+	__le32 clkid;
+};
+
+struct rpmi_get_attrs_rx {
+	__le32 status;
+	__le32 flags;
+	__le32 num_rates;
+	__le32 transition_latency;
+	char name[RPMI_CLK_NAME_LEN];
+};
+
+struct rpmi_get_supp_rates_tx {
+	__le32 clkid;
+	__le32 clk_rate_idx;
+};
+
+struct rpmi_get_supp_rates_rx {
+	__le32 status;
+	__le32 flags;
+	__le32 remaining;
+	__le32 returned;
+	__le32 rates[];
+};
+
+struct rpmi_get_rate_tx {
+	__le32 clkid;
+};
+
+struct rpmi_get_rate_rx {
+	__le32 status;
+	__le32 lo;
+	__le32 hi;
+};
+
+struct rpmi_set_rate_tx {
+	__le32 clkid;
+	__le32 flags;
+	__le32 lo;
+	__le32 hi;
+};
+
+struct rpmi_set_rate_rx {
+	__le32 status;
+};
+
+struct rpmi_set_config_tx {
+	__le32 clkid;
+	__le32 config;
+};
+
+struct rpmi_set_config_rx {
+	__le32 status;
+};
+
+static u32 rpmi_clk_get_num_clocks(struct rpmi_clk_context *context)
+{
+	struct rpmi_get_num_clocks_rx rx;
+	struct rpmi_mbox_message msg;
+	int ret;
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_GET_NUM_CLOCKS,
+					  NULL, 0, &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+
+	if (ret || rx.status)
+		return 0;
+
+	return le32_to_cpu(rx.num_clocks);
+}
+
+static int rpmi_clk_get_attrs(u32 clkid, struct rpmi_clk *rpmi_clk)
+{
+	struct rpmi_clk_context *context = rpmi_clk->context;
+	struct rpmi_mbox_message msg;
+	struct rpmi_get_attrs_tx tx;
+	struct rpmi_get_attrs_rx rx;
+	u8 format;
+	int ret;
+
+	tx.clkid = cpu_to_le32(clkid);
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_GET_ATTRIBUTES,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	rpmi_clk->id = clkid;
+	rpmi_clk->num_rates = le32_to_cpu(rx.num_rates);
+	rpmi_clk->transition_latency = le32_to_cpu(rx.transition_latency);
+	strscpy(rpmi_clk->name, rx.name, RPMI_CLK_NAME_LEN);
+
+	format = le32_to_cpu(rx.flags) & 3U;
+	if (format >= RPMI_CLK_TYPE_MAX_IDX)
+		return -EINVAL;
+
+	rpmi_clk->type = format;
+
+	return 0;
+}
+
+static int rpmi_clk_get_supported_rates(u32 clkid, struct rpmi_clk *rpmi_clk)
+{
+	struct rpmi_clk_context *context = rpmi_clk->context;
+	struct rpmi_clk_rate_discrete *rate_discrete;
+	struct rpmi_clk_rate_linear *rate_linear;
+	struct rpmi_get_supp_rates_rx *rx __free(kfree) = NULL;
+	struct rpmi_get_supp_rates_tx tx;
+	struct rpmi_mbox_message msg;
+	size_t clk_rate_idx = 0;
+	int ret, rateidx, j;
+
+	tx.clkid = cpu_to_le32(clkid);
+	tx.clk_rate_idx = 0;
+
+	/*
+	 * Make sure we allocate rx buffer sufficient to be accommodate all
+	 * the rates sent in one RPMI message.
+	 */
+	rx = kzalloc(context->max_msg_data_size, GFP_KERNEL);
+	if (!rx)
+		return -ENOMEM;
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_GET_SUPPORTED_RATES,
+					  &tx, sizeof(tx), rx, context->max_msg_data_size);
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx->status)
+		return rpmi_to_linux_error(le32_to_cpu(rx->status));
+	if (!le32_to_cpu(rx->returned))
+		return -EINVAL;
+
+	if (rpmi_clk->type == RPMI_CLK_DISCRETE) {
+		rate_discrete = (struct rpmi_clk_rate_discrete *)rx->rates;
+
+		for (rateidx = 0; rateidx < le32_to_cpu(rx->returned); rateidx++) {
+			rpmi_clk->rates->discrete[rateidx] =
+				rpmi_clkrate_u64(le32_to_cpu(rate_discrete[rateidx].hi),
+						 le32_to_cpu(rate_discrete[rateidx].lo));
+		}
+
+		/*
+		 * Keep sending the request message until all
+		 * the rates are received.
+		 */
+		while (le32_to_cpu(rx->remaining)) {
+			clk_rate_idx += le32_to_cpu(rx->returned);
+			tx.clk_rate_idx = cpu_to_le32(clk_rate_idx);
+
+			rpmi_mbox_init_send_with_response(&msg,
+							  RPMI_CLK_SRV_GET_SUPPORTED_RATES,
+							  &tx, sizeof(tx),
+							  rx, context->max_msg_data_size);
+			ret = rpmi_mbox_send_message(context->chan, &msg);
+			if (ret)
+				return ret;
+			if (rx->status)
+				return rpmi_to_linux_error(le32_to_cpu(rx->status));
+			if (!le32_to_cpu(rx->returned))
+				return -EINVAL;
+
+			for (j = 0; j < le32_to_cpu(rx->returned); j++) {
+				if (rateidx >= clk_rate_idx + le32_to_cpu(rx->returned))
+					break;
+				rpmi_clk->rates->discrete[rateidx++] =
+					rpmi_clkrate_u64(le32_to_cpu(rate_discrete[j].hi),
+							 le32_to_cpu(rate_discrete[j].lo));
+			}
+		}
+	} else if (rpmi_clk->type == RPMI_CLK_LINEAR) {
+		rate_linear = (struct rpmi_clk_rate_linear *)rx->rates;
+
+		rpmi_clk->rates->linear.min = rpmi_clkrate_u64(le32_to_cpu(rate_linear->min_hi),
+							       le32_to_cpu(rate_linear->min_lo));
+		rpmi_clk->rates->linear.max = rpmi_clkrate_u64(le32_to_cpu(rate_linear->max_hi),
+							       le32_to_cpu(rate_linear->max_lo));
+		rpmi_clk->rates->linear.step = rpmi_clkrate_u64(le32_to_cpu(rate_linear->step_hi),
+								le32_to_cpu(rate_linear->step_lo));
+	}
+
+	return 0;
+}
+
+static unsigned long rpmi_clk_recalc_rate(struct clk_hw *hw,
+					  unsigned long parent_rate)
+{
+	struct rpmi_clk *rpmi_clk = to_rpmi_clk(hw);
+	struct rpmi_clk_context *context = rpmi_clk->context;
+	struct rpmi_mbox_message msg;
+	struct rpmi_get_rate_tx tx;
+	struct rpmi_get_rate_rx rx;
+	int ret;
+
+	tx.clkid = cpu_to_le32(rpmi_clk->id);
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_GET_RATE,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	return rpmi_clkrate_u64(le32_to_cpu(rx.hi), le32_to_cpu(rx.lo));
+}
+
+static int rpmi_clk_determine_rate(struct clk_hw *hw,
+				   struct clk_rate_request *req)
+{
+	struct rpmi_clk *rpmi_clk = to_rpmi_clk(hw);
+	u64 fmin, fmax, ftmp;
+
+	/*
+	 * Keep the requested rate if the clock format
+	 * is of discrete type. Let the platform which
+	 * is actually controlling the clock handle that.
+	 */
+	if (rpmi_clk->type == RPMI_CLK_DISCRETE)
+		return 0;
+
+	fmin = rpmi_clk->rates->linear.min;
+	fmax = rpmi_clk->rates->linear.max;
+
+	if (req->rate <= fmin) {
+		req->rate = fmin;
+		return 0;
+	} else if (req->rate >= fmax) {
+		req->rate = fmax;
+		return 0;
+	}
+
+	ftmp = req->rate - fmin;
+	ftmp += rpmi_clk->rates->linear.step - 1;
+	do_div(ftmp, rpmi_clk->rates->linear.step);
+
+	req->rate = ftmp * rpmi_clk->rates->linear.step + fmin;
+
+	return 0;
+}
+
+static int rpmi_clk_set_rate(struct clk_hw *hw, unsigned long rate,
+			     unsigned long parent_rate)
+{
+	struct rpmi_clk *rpmi_clk = to_rpmi_clk(hw);
+	struct rpmi_clk_context *context = rpmi_clk->context;
+	struct rpmi_mbox_message msg;
+	struct rpmi_set_rate_tx tx;
+	struct rpmi_set_rate_rx rx;
+	int ret;
+
+	tx.clkid = cpu_to_le32(rpmi_clk->id);
+	tx.lo = cpu_to_le32(lower_32_bits(rate));
+	tx.hi = cpu_to_le32(upper_32_bits(rate));
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_SET_RATE,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	return 0;
+}
+
+static int rpmi_clk_enable(struct clk_hw *hw)
+{
+	struct rpmi_clk *rpmi_clk = to_rpmi_clk(hw);
+	struct rpmi_clk_context *context = rpmi_clk->context;
+	struct rpmi_mbox_message msg;
+	struct rpmi_set_config_tx tx;
+	struct rpmi_set_config_rx rx;
+	int ret;
+
+	tx.config = cpu_to_le32(RPMI_CLK_ENABLE);
+	tx.clkid = cpu_to_le32(rpmi_clk->id);
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_SET_CONFIG,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	return 0;
+}
+
+static void rpmi_clk_disable(struct clk_hw *hw)
+{
+	struct rpmi_clk *rpmi_clk = to_rpmi_clk(hw);
+	struct rpmi_clk_context *context = rpmi_clk->context;
+	struct rpmi_mbox_message msg;
+	struct rpmi_set_config_tx tx;
+	struct rpmi_set_config_rx rx;
+	int ret;
+
+	tx.config = cpu_to_le32(RPMI_CLK_DISABLE);
+	tx.clkid = cpu_to_le32(rpmi_clk->id);
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_CLK_SRV_SET_CONFIG,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret || rx.status)
+		pr_err("Failed to disable clk-%u\n", rpmi_clk->id);
+}
+
+static const struct clk_ops rpmi_clk_ops = {
+	.recalc_rate = rpmi_clk_recalc_rate,
+	.determine_rate = rpmi_clk_determine_rate,
+	.set_rate = rpmi_clk_set_rate,
+	.prepare = rpmi_clk_enable,
+	.unprepare = rpmi_clk_disable,
+};
+
+static struct clk_hw *rpmi_clk_enumerate(struct rpmi_clk_context *context, u32 clkid)
+{
+	struct device *dev = context->dev;
+	unsigned long min_rate, max_rate;
+	union rpmi_clk_rates *rates;
+	struct rpmi_clk *rpmi_clk;
+	struct clk_init_data init = {};
+	struct clk_hw *clk_hw;
+	int ret;
+
+	rates = devm_kzalloc(dev, sizeof(*rates), GFP_KERNEL);
+	if (!rates)
+		return ERR_PTR(-ENOMEM);
+
+	rpmi_clk = devm_kzalloc(dev, sizeof(*rpmi_clk), GFP_KERNEL);
+	if (!rpmi_clk)
+		return ERR_PTR(-ENOMEM);
+
+	rpmi_clk->context = context;
+	rpmi_clk->rates = rates;
+
+	ret = rpmi_clk_get_attrs(clkid, rpmi_clk);
+	if (ret)
+		return dev_err_ptr_probe(dev, ret,
+					 "Failed to get clk-%u attributes\n",
+					 clkid);
+
+	ret = rpmi_clk_get_supported_rates(clkid, rpmi_clk);
+	if (ret)
+		return dev_err_ptr_probe(dev, ret,
+					 "Get supported rates failed for clk-%u\n",
+					 clkid);
+
+	init.flags = CLK_GET_RATE_NOCACHE;
+	init.num_parents = 0;
+	init.ops = &rpmi_clk_ops;
+	init.name = rpmi_clk->name;
+	clk_hw = &rpmi_clk->hw;
+	clk_hw->init = &init;
+
+	ret = devm_clk_hw_register(dev, clk_hw);
+	if (ret)
+		return dev_err_ptr_probe(dev, ret,
+					 "Unable to register clk-%u\n",
+					 clkid);
+
+	if (rpmi_clk->type == RPMI_CLK_DISCRETE) {
+		min_rate = rpmi_clk->rates->discrete[0];
+		max_rate = rpmi_clk->rates->discrete[rpmi_clk->num_rates -  1];
+	} else {
+		min_rate = rpmi_clk->rates->linear.min;
+		max_rate = rpmi_clk->rates->linear.max;
+	}
+
+	clk_hw_set_rate_range(clk_hw, min_rate, max_rate);
+
+	return clk_hw;
+}
+
+static int rpmi_clk_probe(struct platform_device *pdev)
+{
+	int ret;
+	unsigned int num_clocks, i;
+	struct clk_hw_onecell_data *clk_data;
+	struct rpmi_clk_context *context;
+	struct rpmi_mbox_message msg;
+	struct clk_hw *hw_ptr;
+	struct device *dev = &pdev->dev;
+
+	context = devm_kzalloc(dev, sizeof(*context), GFP_KERNEL);
+	if (!context)
+		return -ENOMEM;
+	context->dev = dev;
+	platform_set_drvdata(pdev, context);
+
+	context->client.dev		= context->dev;
+	context->client.rx_callback	= NULL;
+	context->client.tx_block	= false;
+	context->client.knows_txdone	= true;
+	context->client.tx_tout		= 0;
+
+	context->chan = mbox_request_channel(&context->client, 0);
+	if (IS_ERR(context->chan))
+		return PTR_ERR(context->chan);
+
+	rpmi_mbox_init_get_attribute(&msg, RPMI_MBOX_ATTR_SPEC_VERSION);
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, ret, "Failed to get spec version\n");
+	}
+	if (msg.attr.value < RPMI_MKVER(1, 0)) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, -EINVAL,
+				     "msg protocol version mismatch, expected 0x%x, found 0x%x\n",
+				     RPMI_MKVER(1, 0), msg.attr.value);
+	}
+
+	rpmi_mbox_init_get_attribute(&msg, RPMI_MBOX_ATTR_SERVICEGROUP_ID);
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, ret, "Failed to get service group ID\n");
+	}
+	if (msg.attr.value != RPMI_SRVGRP_CLOCK) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, -EINVAL,
+				     "service group match failed, expected 0x%x, found 0x%x\n",
+				     RPMI_SRVGRP_CLOCK, msg.attr.value);
+	}
+
+	rpmi_mbox_init_get_attribute(&msg, RPMI_MBOX_ATTR_SERVICEGROUP_VERSION);
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, ret, "Failed to get service group version\n");
+	}
+	if (msg.attr.value < RPMI_MKVER(1, 0)) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, -EINVAL,
+				     "service group version failed, expected 0x%x, found 0x%x\n",
+				     RPMI_MKVER(1, 0), msg.attr.value);
+	}
+
+	rpmi_mbox_init_get_attribute(&msg, RPMI_MBOX_ATTR_MAX_MSG_DATA_SIZE);
+	ret = rpmi_mbox_send_message(context->chan, &msg);
+	if (ret) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, ret, "Failed to get max message data size\n");
+	}
+
+	context->max_msg_data_size = msg.attr.value;
+	num_clocks = rpmi_clk_get_num_clocks(context);
+	if (!num_clocks) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, -ENODEV, "No clocks found\n");
+	}
+
+	clk_data = devm_kzalloc(dev, struct_size(clk_data, hws, num_clocks),
+				GFP_KERNEL);
+	if (!clk_data) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, -ENOMEM, "No memory for clock data\n");
+	}
+	clk_data->num = num_clocks;
+
+	for (i = 0; i < clk_data->num; i++) {
+		hw_ptr = rpmi_clk_enumerate(context, i);
+		if (IS_ERR(hw_ptr)) {
+			mbox_free_channel(context->chan);
+			return dev_err_probe(dev, PTR_ERR(hw_ptr),
+					     "failed to register clk-%d\n", i);
+		}
+		clk_data->hws[i] = hw_ptr;
+	}
+
+	ret = devm_of_clk_add_hw_provider(dev, of_clk_hw_onecell_get, clk_data);
+	if (ret) {
+		mbox_free_channel(context->chan);
+		return dev_err_probe(dev, ret, "failed to register clock HW provider\n");
+	}
+
+	return 0;
+}
+
+static void rpmi_clk_remove(struct platform_device *pdev)
+{
+	struct rpmi_clk_context *context = platform_get_drvdata(pdev);
+
+	mbox_free_channel(context->chan);
+}
+
+static const struct of_device_id rpmi_clk_of_match[] = {
+	{ .compatible = "riscv,rpmi-clock" },
+	{ }
+};
+MODULE_DEVICE_TABLE(of, rpmi_clk_of_match);
+
+static struct platform_driver rpmi_clk_driver = {
+	.driver = {
+		.name = "riscv-rpmi-clock",
+		.of_match_table = rpmi_clk_of_match,
+	},
+	.probe = rpmi_clk_probe,
+	.remove = rpmi_clk_remove,
+};
+module_platform_driver(rpmi_clk_driver);
+
+MODULE_AUTHOR("Rahul Pathak <rpathak@ventanamicro.com>");
+MODULE_DESCRIPTION("Clock Driver based on RPMI message protocol");
+MODULE_LICENSE("GPL");
diff --git a/include/linux/mailbox/riscv-rpmi-message.h b/include/linux/mailbox/riscv-rpmi-message.h
index 6201d4dd6047..18b606e0023c 100644
--- a/include/linux/mailbox/riscv-rpmi-message.h
+++ b/include/linux/mailbox/riscv-rpmi-message.h
@@ -87,6 +87,22 @@ static inline int rpmi_to_linux_error(int rpmi_error)
 	}
 }
 
+/* RPMI service group IDs */
+#define RPMI_SRVGRP_CLOCK		0x00008
+
+/* RPMI clock service IDs */
+enum rpmi_clock_service_id {
+	RPMI_CLK_SRV_ENABLE_NOTIFICATION = 0x01,
+	RPMI_CLK_SRV_GET_NUM_CLOCKS = 0x02,
+	RPMI_CLK_SRV_GET_ATTRIBUTES = 0x03,
+	RPMI_CLK_SRV_GET_SUPPORTED_RATES = 0x04,
+	RPMI_CLK_SRV_SET_CONFIG = 0x05,
+	RPMI_CLK_SRV_GET_CONFIG = 0x06,
+	RPMI_CLK_SRV_SET_RATE = 0x07,
+	RPMI_CLK_SRV_GET_RATE = 0x08,
+	RPMI_CLK_SRV_ID_MAX_COUNT
+};
+
 /* RPMI linux mailbox attribute IDs */
 enum rpmi_mbox_attribute_id {
 	RPMI_MBOX_ATTR_SPEC_VERSION = 0,
diff --git a/include/linux/wordpart.h b/include/linux/wordpart.h
index 5a7b97bb7c95..1ab72ebd6878 100644
--- a/include/linux/wordpart.h
+++ b/include/linux/wordpart.h
@@ -31,6 +31,14 @@
  */
 #define lower_16_bits(n) ((u16)((n) & 0xffff))
 
+/**
+ * get_u64_from_u32(hi, lo) - return u64 number by combining
+ * two u32 numbers.
+ * @hi: upper 32 bit number
+ * @lo: lower 32 bit number
+ */
+#define get_u64_from_u32(hi, lo)	(((u64)(hi) << 32) | (u32)(lo))
+
 /**
  * REPEAT_BYTE - repeat the value @x multiple times as an unsigned long value
  * @x: value to repeat
-- 
2.43.0

[PATCH v4 11/23] dt-bindings: Add RPMI system MSI message proxy bindings

[Anup Patel]

Add device tree bindings for the RPMI system MSI service group
based message proxy implemented by the SBI implementation (machine
mode firmware or hypervisor).

The RPMI system MSI service group is defined by the RISC-V
platform management interface (RPMI) specification.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 .../riscv,rpmi-mpxy-system-msi.yaml           | 67 +++++++++++++++++++
 1 file changed, 67 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml

diff --git a/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
new file mode 100644
index 000000000000..26dd13731350
--- /dev/null
+++ b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
@@ -0,0 +1,67 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: RISC-V RPMI system MSI service group based message proxy
+
+maintainers:
+  - Anup Patel <anup@brainfault.org>
+
+description: |
+  The RISC-V Platform Management Interface (RPMI) [1] defines a
+  messaging protocol which is modular and extensible. The supervisor
+  software can send/receive RPMI messages via SBI MPXY extension [2]
+  or some dedicated supervisor-mode RPMI transport.
+
+  The RPMI specification [1] defines system MSI service group which
+  allow application processors to receive MSIs upon system events
+  such as P2A doorbell, graceful shutdown/reboot request, CPU hotplug
+  event, memory hotplug event, etc from the platform microcontroller.
+  The SBI implementation machine mode firmware or hypervisor) can
+  implement an SBI MPXY channel to allow RPMI system MSI service
+  group access to the supervisor software.
+
+  ===========================================
+  References
+  ===========================================
+
+  [1] RISC-V Platform Management Interface (RPMI)
+      https://github.com/riscv-non-isa/riscv-rpmi/releases
+
+  [2] RISC-V Supervisor Binary Interface (SBI)
+      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
+
+properties:
+  compatible:
+    description:
+      Intended for use by the SBI implementation.
+    const: riscv,rpmi-mpxy-system-msi
+
+  mboxes:
+    maxItems: 1
+    description:
+      Mailbox channel of the underlying RPMI transport.
+
+  riscv,sbi-mpxy-channel-id:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    description:
+      The SBI MPXY channel id to be used for providing RPMI access to
+      the supervisor software.
+
+required:
+  - compatible
+  - mboxes
+  - riscv,sbi-mpxy-channel-id
+
+additionalProperties: false
+
+examples:
+  - |
+    interrupt-controller {
+        compatible = "riscv,rpmi-mpxy-system-msi";
+        mboxes = <&rpmi_shmem_mbox 0x2>;
+        riscv,sbi-mpxy-channel-id = <0x2000>;
+    };
+...
-- 
2.43.0

[PATCH v4 12/23] dt-bindings: Add RPMI system MSI interrupt controller bindings

[Anup Patel]

Add device tree bindings for the RPMI system MSI service group
based interrupt controller for the supervisor software.

The RPMI system MSI service group is defined by the RISC-V
platform management interface (RPMI) specification.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 .../riscv,rpmi-system-msi.yaml                | 74 +++++++++++++++++++
 1 file changed, 74 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml

diff --git a/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
new file mode 100644
index 000000000000..ac13cec0666e
--- /dev/null
+++ b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
@@ -0,0 +1,74 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/interrupt-controller/riscv,rpmi-system-msi.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: RISC-V RPMI system MSI service group based interrupt controller
+
+maintainers:
+  - Anup Patel <anup@brainfault.org>
+
+description: |
+  The RISC-V Platform Management Interface (RPMI) [1] defines a
+  messaging protocol which is modular and extensible. The supervisor
+  software can send/receive RPMI messages via SBI MPXY extension [2]
+  or some dedicated supervisor-mode RPMI transport.
+
+  The RPMI specification [1] defines system MSI service group which
+  allow application processors to receive MSIs upon system events
+  such as P2A doorbell, graceful shutdown/reboot request, CPU hotplug
+  event, memory hotplug event, etc from the platform microcontroller.
+  The supervisor software can access RPMI system MSI service group via
+  SBI MPXY channel or some dedicated supervisor-mode RPMI transport.
+
+  ===========================================
+  References
+  ===========================================
+
+  [1] RISC-V Platform Management Interface (RPMI)
+      https://github.com/riscv-non-isa/riscv-rpmi/releases
+
+  [2] RISC-V Supervisor Binary Interface (SBI)
+      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
+
+allOf:
+  - $ref: /schemas/interrupt-controller.yaml#
+
+properties:
+  compatible:
+    description:
+      Intended for use by the supervisor software.
+    const: riscv,rpmi-system-msi
+
+  mboxes:
+    maxItems: 1
+    description:
+      Mailbox channel of the underlying RPMI transport or SBI message proxy channel.
+
+  msi-parent: true
+
+  interrupt-controller: true
+
+  "#interrupt-cells":
+    const: 1
+
+required:
+  - compatible
+  - mboxes
+  - msi-parent
+  - interrupt-controller
+  - "#interrupt-cells"
+
+additionalProperties: false
+
+examples:
+  - |
+    interrupt-controller {
+        compatible = "riscv,rpmi-system-msi";
+        mboxes = <&mpxy_mbox 0x2000 0x0>;
+        msi-parent = <&imsic_slevel>;
+        interrupt-controller;
+        #interrupt-cells = <1>;
+    };
+...
-- 
2.43.0

[PATCH v4 13/23] irqchip: Add driver for the RPMI system MSI service group

[Anup Patel]

The RPMI specification defines a system MSI service group which
allows application processors to receive MSIs upon system events
such as graceful shutdown/reboot request, CPU hotplug event, memory
hotplug event, etc.

Add an irqchip driver for the RISC-V RPMI system MSI service group
to directly receive system MSIs in Linux kernel.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 drivers/irqchip/Kconfig                    |   7 +
 drivers/irqchip/Makefile                   |   1 +
 drivers/irqchip/irq-riscv-rpmi-sysmsi.c    | 284 +++++++++++++++++++++
 include/linux/mailbox/riscv-rpmi-message.h |  13 +
 4 files changed, 305 insertions(+)
 create mode 100644 drivers/irqchip/irq-riscv-rpmi-sysmsi.c

diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
index 08bb3b031f23..83700fc2ddc9 100644
--- a/drivers/irqchip/Kconfig
+++ b/drivers/irqchip/Kconfig
@@ -612,6 +612,13 @@ config RISCV_IMSIC
 	select GENERIC_MSI_IRQ
 	select IRQ_MSI_LIB
 
+config RISCV_RPMI_SYSMSI
+	bool
+	depends on MAILBOX
+	select IRQ_DOMAIN_HIERARCHY
+	select GENERIC_MSI_IRQ
+	default RISCV
+
 config SIFIVE_PLIC
 	bool
 	depends on RISCV
diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
index 365bcea9a61f..515280da499c 100644
--- a/drivers/irqchip/Makefile
+++ b/drivers/irqchip/Makefile
@@ -102,6 +102,7 @@ obj-$(CONFIG_RISCV_INTC)		+= irq-riscv-intc.o
 obj-$(CONFIG_RISCV_APLIC)		+= irq-riscv-aplic-main.o irq-riscv-aplic-direct.o
 obj-$(CONFIG_RISCV_APLIC_MSI)		+= irq-riscv-aplic-msi.o
 obj-$(CONFIG_RISCV_IMSIC)		+= irq-riscv-imsic-state.o irq-riscv-imsic-early.o irq-riscv-imsic-platform.o
+obj-$(CONFIG_RISCV_RPMI_SYSMSI)		+= irq-riscv-rpmi-sysmsi.o
 obj-$(CONFIG_SIFIVE_PLIC)		+= irq-sifive-plic.o
 obj-$(CONFIG_STARFIVE_JH8100_INTC)	+= irq-starfive-jh8100-intc.o
 obj-$(CONFIG_THEAD_C900_ACLINT_SSWI)	+= irq-thead-c900-aclint-sswi.o
diff --git a/drivers/irqchip/irq-riscv-rpmi-sysmsi.c b/drivers/irqchip/irq-riscv-rpmi-sysmsi.c
new file mode 100644
index 000000000000..d7b19fb7bf4e
--- /dev/null
+++ b/drivers/irqchip/irq-riscv-rpmi-sysmsi.c
@@ -0,0 +1,284 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2025 Ventana Micro Systems Inc. */
+
+#include <linux/device.h>
+#include <linux/dev_printk.h>
+#include <linux/irq.h>
+#include <linux/irqdomain.h>
+#include <linux/mailbox_client.h>
+#include <linux/mailbox/riscv-rpmi-message.h>
+#include <linux/module.h>
+#include <linux/msi.h>
+#include <linux/of_irq.h>
+#include <linux/platform_device.h>
+#include <linux/types.h>
+#include <vdso/bits.h>
+
+struct rpmi_sysmsi_get_attrs_rx {
+	__le32 status;
+	__le32 sys_num_msi;
+	__le32 flag0;
+	__le32 flag1;
+};
+
+#define RPMI_SYSMSI_MSI_ATTRIBUTES_FLAG0_PREF_PRIV	BIT(0)
+
+struct rpmi_sysmsi_set_msi_state_tx {
+	__le32 sys_msi_index;
+	__le32 sys_msi_state;
+};
+
+struct rpmi_sysmsi_set_msi_state_rx {
+	__le32 status;
+};
+
+#define RPMI_SYSMSI_MSI_STATE_ENABLE			BIT(0)
+#define RPMI_SYSMSI_MSI_STATE_PENDING			BIT(1)
+
+struct rpmi_sysmsi_set_msi_target_tx {
+	__le32 sys_msi_index;
+	__le32 sys_msi_address_low;
+	__le32 sys_msi_address_high;
+	__le32 sys_msi_data;
+};
+
+struct rpmi_sysmsi_set_msi_target_rx {
+	__le32 status;
+};
+
+struct rpmi_sysmsi_priv {
+	struct device		*dev;
+	struct mbox_client	client;
+	struct mbox_chan	*chan;
+	u32			nr_irqs;
+	u32			gsi_base;
+};
+
+static int rpmi_sysmsi_get_num_msi(struct rpmi_sysmsi_priv *priv)
+{
+	struct rpmi_sysmsi_get_attrs_rx rx;
+	struct rpmi_mbox_message msg;
+	int ret;
+
+	rpmi_mbox_init_send_with_response(&msg, RPMI_SYSMSI_SRV_GET_ATTRIBUTES,
+					  NULL, 0, &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(priv->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	return le32_to_cpu(rx.sys_num_msi);
+}
+
+static int rpmi_sysmsi_set_msi_state(struct rpmi_sysmsi_priv *priv,
+				     u32 sys_msi_index, u32 sys_msi_state)
+{
+	struct rpmi_sysmsi_set_msi_state_tx tx;
+	struct rpmi_sysmsi_set_msi_state_rx rx;
+	struct rpmi_mbox_message msg;
+	int ret;
+
+	tx.sys_msi_index = cpu_to_le32(sys_msi_index);
+	tx.sys_msi_state = cpu_to_le32(sys_msi_state);
+	rpmi_mbox_init_send_with_response(&msg, RPMI_SYSMSI_SRV_SET_MSI_STATE,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(priv->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	return 0;
+}
+
+static int rpmi_sysmsi_set_msi_target(struct rpmi_sysmsi_priv *priv,
+				      u32 sys_msi_index, struct msi_msg *m)
+{
+	struct rpmi_sysmsi_set_msi_target_tx tx;
+	struct rpmi_sysmsi_set_msi_target_rx rx;
+	struct rpmi_mbox_message msg;
+	int ret;
+
+	tx.sys_msi_index = cpu_to_le32(sys_msi_index);
+	tx.sys_msi_address_low = cpu_to_le32(m->address_lo);
+	tx.sys_msi_address_high = cpu_to_le32(m->address_hi);
+	tx.sys_msi_data = cpu_to_le32(m->data);
+	rpmi_mbox_init_send_with_response(&msg, RPMI_SYSMSI_SRV_SET_MSI_TARGET,
+					  &tx, sizeof(tx), &rx, sizeof(rx));
+	ret = rpmi_mbox_send_message(priv->chan, &msg);
+	if (ret)
+		return ret;
+	if (rx.status)
+		return rpmi_to_linux_error(le32_to_cpu(rx.status));
+
+	return 0;
+}
+
+static void rpmi_sysmsi_irq_mask(struct irq_data *d)
+{
+	struct rpmi_sysmsi_priv *priv = irq_data_get_irq_chip_data(d);
+	int ret;
+
+	ret = rpmi_sysmsi_set_msi_state(priv, irqd_to_hwirq(d), 0);
+	if (ret) {
+		dev_warn(priv->dev, "Failed to mask hwirq %lu (error %d)\n",
+			 irqd_to_hwirq(d), ret);
+	}
+	irq_chip_mask_parent(d);
+}
+
+static void rpmi_sysmsi_irq_unmask(struct irq_data *d)
+{
+	struct rpmi_sysmsi_priv *priv = irq_data_get_irq_chip_data(d);
+	int ret;
+
+	irq_chip_unmask_parent(d);
+	ret = rpmi_sysmsi_set_msi_state(priv, irqd_to_hwirq(d), RPMI_SYSMSI_MSI_STATE_ENABLE);
+	if (ret) {
+		dev_warn(priv->dev, "Failed to unmask hwirq %lu (error %d)\n",
+			 irqd_to_hwirq(d), ret);
+	}
+}
+
+static void rpmi_sysmsi_write_msg(struct irq_data *d, struct msi_msg *msg)
+{
+	struct rpmi_sysmsi_priv *priv = irq_data_get_irq_chip_data(d);
+	int ret;
+
+	/* For zeroed MSI, do nothing as of now */
+	if (!msg->address_hi && !msg->address_lo && !msg->data)
+		return;
+
+	ret = rpmi_sysmsi_set_msi_target(priv, irqd_to_hwirq(d), msg);
+	if (ret) {
+		dev_warn(priv->dev, "Failed to set target for hwirq %lu (error %d)\n",
+			 irqd_to_hwirq(d), ret);
+	}
+}
+
+static void rpmi_sysmsi_set_desc(msi_alloc_info_t *arg, struct msi_desc *desc)
+{
+	arg->desc = desc;
+	arg->hwirq = (u32)desc->data.icookie.value;
+}
+
+static int rpmi_sysmsi_translate(struct irq_domain *d, struct irq_fwspec *fwspec,
+				 unsigned long *hwirq, unsigned int *type)
+{
+	struct msi_domain_info *info = d->host_data;
+	struct rpmi_sysmsi_priv *priv = info->data;
+
+	if (WARN_ON(fwspec->param_count < 1))
+		return -EINVAL;
+
+	/* For DT, gsi_base is always zero. */
+	*hwirq = fwspec->param[0] - priv->gsi_base;
+	*type = IRQ_TYPE_NONE;
+	return 0;
+}
+
+static const struct msi_domain_template rpmi_sysmsi_template = {
+	.chip = {
+		.name			= "RPMI-SYSMSI",
+		.irq_mask		= rpmi_sysmsi_irq_mask,
+		.irq_unmask		= rpmi_sysmsi_irq_unmask,
+#ifdef CONFIG_SMP
+		.irq_set_affinity	= irq_chip_set_affinity_parent,
+#endif
+		.irq_write_msi_msg	= rpmi_sysmsi_write_msg,
+		.flags			= IRQCHIP_SET_TYPE_MASKED |
+					  IRQCHIP_SKIP_SET_WAKE |
+					  IRQCHIP_MASK_ON_SUSPEND,
+	},
+
+	.ops = {
+		.set_desc		= rpmi_sysmsi_set_desc,
+		.msi_translate		= rpmi_sysmsi_translate,
+	},
+
+	.info = {
+		.bus_token		= DOMAIN_BUS_WIRED_TO_MSI,
+		.flags			= MSI_FLAG_USE_DEV_FWNODE,
+		.handler		= handle_simple_irq,
+		.handler_name		= "simple",
+	},
+};
+
+static int rpmi_sysmsi_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct rpmi_sysmsi_priv *priv;
+	int rc;
+
+	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	priv->dev = dev;
+	platform_set_drvdata(pdev, priv);
+
+	/* Setup mailbox client */
+	priv->client.dev		= priv->dev;
+	priv->client.rx_callback	= NULL;
+	priv->client.tx_block		= false;
+	priv->client.knows_txdone	= true;
+	priv->client.tx_tout		= 0;
+
+	/* Request mailbox channel */
+	priv->chan = mbox_request_channel(&priv->client, 0);
+	if (IS_ERR(priv->chan))
+		return PTR_ERR(priv->chan);
+
+	/* Get number of system MSIs */
+	rc = rpmi_sysmsi_get_num_msi(priv);
+	if (rc < 1) {
+		mbox_free_channel(priv->chan);
+		if (rc)
+			return dev_err_probe(dev, rc, "Failed to get number of system MSIs\n");
+		else
+			return dev_err_probe(dev, -ENODEV, "No system MSIs found\n");
+	}
+	priv->nr_irqs = rc;
+
+	/* Set the device MSI domain if not available */
+	if (!dev_get_msi_domain(dev)) {
+		/*
+		 * The device MSI domain for OF devices is only set at the
+		 * time of populating/creating OF device. If the device MSI
+		 * domain is discovered later after the OF device is created
+		 * then we need to set it explicitly before using any platform
+		 * MSI functions.
+		 */
+		if (is_of_node(dev_fwnode(dev)))
+			of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
+
+		if (!dev_get_msi_domain(dev)) {
+			mbox_free_channel(priv->chan);
+			return -EPROBE_DEFER;
+		}
+	}
+
+	if (!msi_create_device_irq_domain(dev, MSI_DEFAULT_DOMAIN,
+					  &rpmi_sysmsi_template,
+					  priv->nr_irqs, priv, priv)) {
+		mbox_free_channel(priv->chan);
+		return dev_err_probe(dev, -ENOMEM, "failed to create MSI irq domain\n");
+	}
+
+	dev_info(dev, "%u system MSIs registered\n", priv->nr_irqs);
+	return 0;
+}
+
+static const struct of_device_id rpmi_sysmsi_match[] = {
+	{ .compatible = "riscv,rpmi-system-msi" },
+	{}
+};
+
+static struct platform_driver rpmi_sysmsi_driver = {
+	.driver = {
+		.name		= "rpmi-sysmsi",
+		.of_match_table	= rpmi_sysmsi_match,
+	},
+	.probe = rpmi_sysmsi_probe,
+};
+builtin_platform_driver(rpmi_sysmsi_driver);
diff --git a/include/linux/mailbox/riscv-rpmi-message.h b/include/linux/mailbox/riscv-rpmi-message.h
index 18b606e0023c..90eacb37fdd8 100644
--- a/include/linux/mailbox/riscv-rpmi-message.h
+++ b/include/linux/mailbox/riscv-rpmi-message.h
@@ -88,6 +88,7 @@ static inline int rpmi_to_linux_error(int rpmi_error)
 }
 
 /* RPMI service group IDs */
+#define RPMI_SRVGRP_SYSTEM_MSI		0x00002
 #define RPMI_SRVGRP_CLOCK		0x00008
 
 /* RPMI clock service IDs */
@@ -103,6 +104,18 @@ enum rpmi_clock_service_id {
 	RPMI_CLK_SRV_ID_MAX_COUNT
 };
 
+/* RPMI system MSI service IDs */
+enum rpmi_sysmsi_service_id {
+	RPMI_SYSMSI_SRV_ENABLE_NOTIFICATION = 0x01,
+	RPMI_SYSMSI_SRV_GET_ATTRIBUTES = 0x02,
+	RPMI_SYSMSI_SRV_GET_MSI_ATTRIBUTES = 0x03,
+	RPMI_SYSMSI_SRV_SET_MSI_STATE = 0x04,
+	RPMI_SYSMSI_SRV_GET_MSI_STATE = 0x05,
+	RPMI_SYSMSI_SRV_SET_MSI_TARGET = 0x06,
+	RPMI_SYSMSI_SRV_GET_MSI_TARGET = 0x07,
+	RPMI_SYSMSI_SRV_ID_MAX_COUNT
+};
+
 /* RPMI linux mailbox attribute IDs */
 enum rpmi_mbox_attribute_id {
 	RPMI_MBOX_ATTR_SPEC_VERSION = 0,
-- 
2.43.0

[PATCH v4 14/23] ACPI: property: Refactor acpi_fwnode_get_reference_args()

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

Currently acpi_fwnode_get_reference_args() calls the public function
__acpi_node_get_property_reference() which ignores the nargs_prop
parameter. To fix this, make __acpi_node_get_property_reference() to
call the static acpi_fwnode_get_reference() so that callers of
fwnode_get_reference_args() can still pass a valid property name to
fetch the number of arguments.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
---
 drivers/acpi/property.c | 101 ++++++++++++++++++++--------------------
 1 file changed, 50 insertions(+), 51 deletions(-)

diff --git a/drivers/acpi/property.c b/drivers/acpi/property.c
index 436019d96027..d4863746fb11 100644
--- a/drivers/acpi/property.c
+++ b/drivers/acpi/property.c
@@ -882,45 +882,10 @@ static struct fwnode_handle *acpi_parse_string_ref(const struct fwnode_handle *f
 	return &dn->fwnode;
 }
 
-/**
- * __acpi_node_get_property_reference - returns handle to the referenced object
- * @fwnode: Firmware node to get the property from
- * @propname: Name of the property
- * @index: Index of the reference to return
- * @num_args: Maximum number of arguments after each reference
- * @args: Location to store the returned reference with optional arguments
- *	  (may be NULL)
- *
- * Find property with @name, verifify that it is a package containing at least
- * one object reference and if so, store the ACPI device object pointer to the
- * target object in @args->adev.  If the reference includes arguments, store
- * them in the @args->args[] array.
- *
- * If there's more than one reference in the property value package, @index is
- * used to select the one to return.
- *
- * It is possible to leave holes in the property value set like in the
- * example below:
- *
- * Package () {
- *     "cs-gpios",
- *     Package () {
- *        ^GPIO, 19, 0, 0,
- *        ^GPIO, 20, 0, 0,
- *        0,
- *        ^GPIO, 21, 0, 0,
- *     }
- * }
- *
- * Calling this function with index %2 or index %3 return %-ENOENT. If the
- * property does not contain any more values %-ENOENT is returned. The NULL
- * entry must be single integer and preferably contain value %0.
- *
- * Return: %0 on success, negative error code on failure.
- */
-int __acpi_node_get_property_reference(const struct fwnode_handle *fwnode,
-	const char *propname, size_t index, size_t num_args,
-	struct fwnode_reference_args *args)
+static int acpi_fwnode_get_reference_args(const struct fwnode_handle *fwnode,
+					  const char *propname, const char *nargs_prop,
+					  unsigned int args_count, unsigned int index,
+					  struct fwnode_reference_args *args)
 {
 	const union acpi_object *element, *end;
 	const union acpi_object *obj;
@@ -999,7 +964,7 @@ int __acpi_node_get_property_reference(const struct fwnode_handle *fwnode,
 
 			ret = acpi_get_ref_args(idx == index ? args : NULL,
 						acpi_fwnode_handle(device),
-						&element, end, num_args);
+						&element, end, args_count);
 			if (ret < 0)
 				return ret;
 
@@ -1017,7 +982,7 @@ int __acpi_node_get_property_reference(const struct fwnode_handle *fwnode,
 
 			ret = acpi_get_ref_args(idx == index ? args : NULL,
 						ref_fwnode, &element, end,
-						num_args);
+						args_count);
 			if (ret < 0)
 				return ret;
 
@@ -1039,6 +1004,50 @@ int __acpi_node_get_property_reference(const struct fwnode_handle *fwnode,
 
 	return -ENOENT;
 }
+
+/**
+ * __acpi_node_get_property_reference - returns handle to the referenced object
+ * @fwnode: Firmware node to get the property from
+ * @propname: Name of the property
+ * @index: Index of the reference to return
+ * @num_args: Maximum number of arguments after each reference
+ * @args: Location to store the returned reference with optional arguments
+ *	  (may be NULL)
+ *
+ * Find property with @name, verifify that it is a package containing at least
+ * one object reference and if so, store the ACPI device object pointer to the
+ * target object in @args->adev.  If the reference includes arguments, store
+ * them in the @args->args[] array.
+ *
+ * If there's more than one reference in the property value package, @index is
+ * used to select the one to return.
+ *
+ * It is possible to leave holes in the property value set like in the
+ * example below:
+ *
+ * Package () {
+ *     "cs-gpios",
+ *     Package () {
+ *        ^GPIO, 19, 0, 0,
+ *        ^GPIO, 20, 0, 0,
+ *        0,
+ *        ^GPIO, 21, 0, 0,
+ *     }
+ * }
+ *
+ * Calling this function with index %2 or index %3 return %-ENOENT. If the
+ * property does not contain any more values %-ENOENT is returned. The NULL
+ * entry must be single integer and preferably contain value %0.
+ *
+ * Return: %0 on success, negative error code on failure.
+ */
+int __acpi_node_get_property_reference(const struct fwnode_handle *fwnode,
+				       const char *propname, size_t index,
+				       size_t num_args,
+				       struct fwnode_reference_args *args)
+{
+	return acpi_fwnode_get_reference_args(fwnode, propname, NULL, index, num_args, args);
+}
 EXPORT_SYMBOL_GPL(__acpi_node_get_property_reference);
 
 static int acpi_data_prop_read_single(const struct acpi_device_data *data,
@@ -1558,16 +1567,6 @@ acpi_fwnode_property_read_string_array(const struct fwnode_handle *fwnode,
 				   val, nval);
 }
 
-static int
-acpi_fwnode_get_reference_args(const struct fwnode_handle *fwnode,
-			       const char *prop, const char *nargs_prop,
-			       unsigned int args_count, unsigned int index,
-			       struct fwnode_reference_args *args)
-{
-	return __acpi_node_get_property_reference(fwnode, prop, index,
-						  args_count, args);
-}
-
 static const char *acpi_fwnode_get_name(const struct fwnode_handle *fwnode)
 {
 	const struct acpi_device *adev;
-- 
2.43.0

[PATCH v4 15/23] ACPI: property: Add support for cells property

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

Currently, ACPI doesn't support cells property when
fwnode_property_get_reference_args() is called. ACPI always expects
the number of arguments to be passed. However, the above mentioned
call being a common interface for OF and ACPI, it is better to have
single calling convention which works for both. Hence, add support
for cells property on the reference device to get the number of
arguments dynamically.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
---
 drivers/acpi/property.c | 22 ++++++++++++++++++++++
 drivers/base/property.c |  2 +-
 2 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/drivers/acpi/property.c b/drivers/acpi/property.c
index d4863746fb11..c9c3d6920326 100644
--- a/drivers/acpi/property.c
+++ b/drivers/acpi/property.c
@@ -882,6 +882,20 @@ static struct fwnode_handle *acpi_parse_string_ref(const struct fwnode_handle *f
 	return &dn->fwnode;
 }
 
+static unsigned int acpi_fwnode_get_args_count(const struct acpi_device *device,
+					       const char *nargs_prop)
+{
+	const union acpi_object *obj;
+
+	if (!nargs_prop)
+		return 0;
+
+	if (acpi_dev_get_property(device, nargs_prop, ACPI_TYPE_INTEGER, &obj))
+		return 0;
+
+	return obj->integer.value;
+}
+
 static int acpi_fwnode_get_reference_args(const struct fwnode_handle *fwnode,
 					  const char *propname, const char *nargs_prop,
 					  unsigned int args_count, unsigned int index,
@@ -960,6 +974,9 @@ static int acpi_fwnode_get_reference_args(const struct fwnode_handle *fwnode,
 			if (!device)
 				return -EINVAL;
 
+			if (nargs_prop)
+				args_count = acpi_fwnode_get_args_count(device, nargs_prop);
+
 			element++;
 
 			ret = acpi_get_ref_args(idx == index ? args : NULL,
@@ -978,6 +995,11 @@ static int acpi_fwnode_get_reference_args(const struct fwnode_handle *fwnode,
 			if (!ref_fwnode)
 				return -EINVAL;
 
+			if (nargs_prop) {
+				device = to_acpi_device_node(ref_fwnode);
+				args_count = acpi_fwnode_get_args_count(device, nargs_prop);
+			}
+
 			element++;
 
 			ret = acpi_get_ref_args(idx == index ? args : NULL,
diff --git a/drivers/base/property.c b/drivers/base/property.c
index c1392743df9c..172dfb950328 100644
--- a/drivers/base/property.c
+++ b/drivers/base/property.c
@@ -578,7 +578,7 @@ EXPORT_SYMBOL_GPL(fwnode_property_match_property_string);
  * @prop:	The name of the property
  * @nargs_prop:	The name of the property telling the number of
  *		arguments in the referred node. NULL if @nargs is known,
- *		otherwise @nargs is ignored. Only relevant on OF.
+ *		otherwise @nargs is ignored.
  * @nargs:	Number of arguments. Ignored if @nargs_prop is non-NULL.
  * @index:	Index of the reference, from zero onwards.
  * @args:	Result structure with reference and integer arguments.
-- 
2.43.0

[PATCH v4 16/23] ACPI: scan: Update honor list for RPMI System MSI

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

The RPMI System MSI interrupt controller (just like PLIC and APLIC)
needs to probed prior to devices like GED which use interrupts provided
by it. Also, it has dependency on the SBI MPXY mailbox device.

Add HIDs of RPMI System MSI and SBI MPXY mailbox devices to the honor
list so that those dependencies are handled.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 drivers/acpi/scan.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index fb1fe9f3b1a3..54181b03b345 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -858,6 +858,8 @@ static const char * const acpi_honor_dep_ids[] = {
 	"INTC10CF", /* IVSC (MTL) driver must be loaded to allow i2c access to camera sensors */
 	"RSCV0001", /* RISC-V PLIC */
 	"RSCV0002", /* RISC-V APLIC */
+	"RSCV0005", /* RISC-V SBI MPXY MBOX */
+	"RSCV0006", /* RISC-V RPMI SYSMSI */
 	"PNP0C0F",  /* PCI Link Device */
 	NULL
 };
-- 
2.43.0

[PATCH v4 17/23] ACPI: RISC-V: Create interrupt controller list in sorted order

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

Currently, the interrupt controller list is created without any order.
Create the list sorted with the GSI base of the interrupt controllers.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
---
 drivers/acpi/riscv/irq.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/acpi/riscv/irq.c b/drivers/acpi/riscv/irq.c
index cced960c2aef..95b4d7574ce3 100644
--- a/drivers/acpi/riscv/irq.c
+++ b/drivers/acpi/riscv/irq.c
@@ -115,7 +115,7 @@ struct fwnode_handle *riscv_acpi_get_gsi_domain_id(u32 gsi)
 static int __init riscv_acpi_register_ext_intc(u32 gsi_base, u32 nr_irqs, u32 nr_idcs,
 					       u32 id, u32 type)
 {
-	struct riscv_ext_intc_list *ext_intc_element;
+	struct riscv_ext_intc_list *ext_intc_element, *node;
 
 	ext_intc_element = kzalloc(sizeof(*ext_intc_element), GFP_KERNEL);
 	if (!ext_intc_element)
@@ -125,7 +125,17 @@ static int __init riscv_acpi_register_ext_intc(u32 gsi_base, u32 nr_irqs, u32 nr
 	ext_intc_element->nr_irqs = nr_irqs;
 	ext_intc_element->nr_idcs = nr_idcs;
 	ext_intc_element->id = id;
-	list_add_tail(&ext_intc_element->list, &ext_intc_list);
+	if (list_empty(&ext_intc_list)) {
+		list_add(&ext_intc_element->list, &ext_intc_list);
+		return 0;
+	}
+
+	list_for_each_entry(node, &ext_intc_list, list) {
+		if (node->gsi_base < ext_intc_element->gsi_base)
+			break;
+	}
+
+	__list_add(&ext_intc_element->list, node->list.prev, &node->list);
 	return 0;
 }
 
-- 
2.43.0

[PATCH v4 18/23] ACPI: RISC-V: Add support to update gsi range

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

Some RISC-V interrupt controllers like RPMI based system MSI interrupt
controllers do not have MADT entry defined. These interrupt controllers
exist only in the namespace. ACPI spec defines _GSB method to get the
GSI base of the interrupt controller, However, there is no such standard
method to get the GSI range. To support such interrupt controllers, set
the GSI range of such interrupt controllers to non-overlapping range and
provide API for interrupt controller driver to update it with proper
value.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
---
 arch/riscv/include/asm/irq.h |  5 +++++
 drivers/acpi/riscv/irq.c     | 38 ++++++++++++++++++++++++++++++++++--
 2 files changed, 41 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/include/asm/irq.h b/arch/riscv/include/asm/irq.h
index 7b038f3b7cb0..2caf049f09c8 100644
--- a/arch/riscv/include/asm/irq.h
+++ b/arch/riscv/include/asm/irq.h
@@ -40,6 +40,7 @@ unsigned long acpi_rintc_ext_parent_to_hartid(unsigned int plic_id, unsigned int
 unsigned int acpi_rintc_get_plic_nr_contexts(unsigned int plic_id);
 unsigned int acpi_rintc_get_plic_context(unsigned int plic_id, unsigned int ctxt_idx);
 int __init acpi_rintc_get_imsic_mmio_info(u32 index, struct resource *res);
+int riscv_acpi_update_gsi_range(u32 gsi_base, u32 nr_irqs);
 
 #else
 static inline int riscv_acpi_get_gsi_info(struct fwnode_handle *fwnode, u32 *gsi_base,
@@ -74,6 +75,10 @@ static inline int __init acpi_rintc_get_imsic_mmio_info(u32 index, struct resour
 	return 0;
 }
 
+static inline int riscv_acpi_update_gsi_range(u32 gsi_base, u32 nr_irqs)
+{
+	return -ENODEV;
+}
 #endif /* CONFIG_ACPI */
 
 #endif /* _ASM_RISCV_IRQ_H */
diff --git a/drivers/acpi/riscv/irq.c b/drivers/acpi/riscv/irq.c
index 95b4d7574ce3..c881cf037f0a 100644
--- a/drivers/acpi/riscv/irq.c
+++ b/drivers/acpi/riscv/irq.c
@@ -10,6 +10,8 @@
 
 #include "init.h"
 
+#define RISCV_ACPI_INTC_FLAG_PENDING BIT(0)
+
 struct riscv_ext_intc_list {
 	acpi_handle		handle;
 	u32			gsi_base;
@@ -17,6 +19,7 @@ struct riscv_ext_intc_list {
 	u32			nr_idcs;
 	u32			id;
 	u32			type;
+	u32			flag;
 	struct list_head	list;
 };
 
@@ -69,6 +72,22 @@ static acpi_status riscv_acpi_update_gsi_handle(u32 gsi_base, acpi_handle handle
 	return AE_NOT_FOUND;
 }
 
+int riscv_acpi_update_gsi_range(u32 gsi_base, u32 nr_irqs)
+{
+	struct riscv_ext_intc_list *ext_intc_element;
+
+	list_for_each_entry(ext_intc_element, &ext_intc_list, list) {
+		if (gsi_base == ext_intc_element->gsi_base &&
+		    (ext_intc_element->flag & RISCV_ACPI_INTC_FLAG_PENDING)) {
+			ext_intc_element->nr_irqs = nr_irqs;
+			ext_intc_element->flag &= ~RISCV_ACPI_INTC_FLAG_PENDING;
+			return 0;
+		}
+	}
+
+	return -ENODEV;
+}
+
 int riscv_acpi_get_gsi_info(struct fwnode_handle *fwnode, u32 *gsi_base,
 			    u32 *id, u32 *nr_irqs, u32 *nr_idcs)
 {
@@ -115,14 +134,22 @@ struct fwnode_handle *riscv_acpi_get_gsi_domain_id(u32 gsi)
 static int __init riscv_acpi_register_ext_intc(u32 gsi_base, u32 nr_irqs, u32 nr_idcs,
 					       u32 id, u32 type)
 {
-	struct riscv_ext_intc_list *ext_intc_element, *node;
+	struct riscv_ext_intc_list *ext_intc_element, *node, *prev;
 
 	ext_intc_element = kzalloc(sizeof(*ext_intc_element), GFP_KERNEL);
 	if (!ext_intc_element)
 		return -ENOMEM;
 
 	ext_intc_element->gsi_base = gsi_base;
-	ext_intc_element->nr_irqs = nr_irqs;
+
+	/* If nr_irqs is zero, indicate it in flag and set to max range possible */
+	if (nr_irqs) {
+		ext_intc_element->nr_irqs = nr_irqs;
+	} else {
+		ext_intc_element->flag |= RISCV_ACPI_INTC_FLAG_PENDING;
+		ext_intc_element->nr_irqs = U32_MAX - ext_intc_element->gsi_base;
+	}
+
 	ext_intc_element->nr_idcs = nr_idcs;
 	ext_intc_element->id = id;
 	if (list_empty(&ext_intc_list)) {
@@ -135,6 +162,13 @@ static int __init riscv_acpi_register_ext_intc(u32 gsi_base, u32 nr_irqs, u32 nr
 			break;
 	}
 
+	/* Adjust the previous node's GSI range if that has pending registration */
+	prev = list_prev_entry(node, list);
+	if (!list_entry_is_head(prev, &ext_intc_list, list)) {
+		if (prev->flag & RISCV_ACPI_INTC_FLAG_PENDING)
+			prev->nr_irqs = ext_intc_element->gsi_base - prev->gsi_base;
+	}
+
 	__list_add(&ext_intc_element->list, node->list.prev, &node->list);
 	return 0;
 }
-- 
2.43.0

[PATCH v4 19/23] ACPI: RISC-V: Add RPMI System MSI to GSI mapping

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

The RPMI System MSI device will provide GSIs to downstream devices
(such as GED) so add it to the RISC-V GSI to fwnode mapping.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 arch/riscv/include/asm/irq.h |  1 +
 drivers/acpi/riscv/irq.c     | 30 ++++++++++++++++++++++++++++++
 2 files changed, 31 insertions(+)

diff --git a/arch/riscv/include/asm/irq.h b/arch/riscv/include/asm/irq.h
index 2caf049f09c8..9c9d22f5165e 100644
--- a/arch/riscv/include/asm/irq.h
+++ b/arch/riscv/include/asm/irq.h
@@ -30,6 +30,7 @@ enum riscv_irqchip_type {
 	ACPI_RISCV_IRQCHIP_IMSIC	= 0x01,
 	ACPI_RISCV_IRQCHIP_PLIC		= 0x02,
 	ACPI_RISCV_IRQCHIP_APLIC	= 0x03,
+	ACPI_RISCV_IRQCHIP_SMSI		= 0x04,
 };
 
 int riscv_acpi_get_gsi_info(struct fwnode_handle *fwnode, u32 *gsi_base,
diff --git a/drivers/acpi/riscv/irq.c b/drivers/acpi/riscv/irq.c
index c881cf037f0a..f5e91972b5fa 100644
--- a/drivers/acpi/riscv/irq.c
+++ b/drivers/acpi/riscv/irq.c
@@ -173,6 +173,33 @@ static int __init riscv_acpi_register_ext_intc(u32 gsi_base, u32 nr_irqs, u32 nr
 	return 0;
 }
 
+static acpi_status __init riscv_acpi_create_gsi_map_smsi(acpi_handle handle, u32 level,
+							 void *context, void **return_value)
+{
+	acpi_status status;
+	u64 gbase;
+
+	if (!acpi_has_method(handle, "_GSB")) {
+		acpi_handle_err(handle, "_GSB method not found\n");
+		return AE_ERROR;
+	}
+
+	status = acpi_evaluate_integer(handle, "_GSB", NULL, &gbase);
+	if (ACPI_FAILURE(status)) {
+		acpi_handle_err(handle, "failed to evaluate _GSB method\n");
+		return status;
+	}
+
+	riscv_acpi_register_ext_intc(gbase, 0, 0, 0, ACPI_RISCV_IRQCHIP_SMSI);
+	status = riscv_acpi_update_gsi_handle((u32)gbase, handle);
+	if (ACPI_FAILURE(status)) {
+		acpi_handle_err(handle, "failed to find the GSI mapping entry\n");
+		return status;
+	}
+
+	return AE_OK;
+}
+
 static acpi_status __init riscv_acpi_create_gsi_map(acpi_handle handle, u32 level,
 						    void *context, void **return_value)
 {
@@ -227,6 +254,9 @@ void __init riscv_acpi_init_gsi_mapping(void)
 
 	if (acpi_table_parse_madt(ACPI_MADT_TYPE_APLIC, riscv_acpi_aplic_parse_madt, 0) > 0)
 		acpi_get_devices("RSCV0002", riscv_acpi_create_gsi_map, NULL, NULL);
+
+	/* Unlike PLIC/APLIC, SYSMSI doesn't have MADT */
+	acpi_get_devices("RSCV0006", riscv_acpi_create_gsi_map_smsi, NULL, NULL);
 }
 
 static acpi_handle riscv_acpi_get_gsi_handle(u32 gsi)
-- 
2.43.0

[PATCH v4 20/23] mailbox/riscv-sbi-mpxy: Add ACPI support

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

Add ACPI support for the RISC-V SBI message proxy (MPXY) based
mailbox driver.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 drivers/mailbox/riscv-sbi-mpxy-mbox.c | 28 +++++++++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)

diff --git a/drivers/mailbox/riscv-sbi-mpxy-mbox.c b/drivers/mailbox/riscv-sbi-mpxy-mbox.c
index 121ee5fd3d0d..3d71af6321c3 100644
--- a/drivers/mailbox/riscv-sbi-mpxy-mbox.c
+++ b/drivers/mailbox/riscv-sbi-mpxy-mbox.c
@@ -6,9 +6,11 @@
  */
 
 #include <asm/sbi.h>
+#include <linux/acpi.h>
 #include <linux/cpu.h>
 #include <linux/err.h>
 #include <linux/init.h>
+#include <linux/irqchip/riscv-imsic.h>
 #include <linux/jump_label.h>
 #include <linux/kernel.h>
 #include <linux/mailbox_controller.h>
@@ -767,6 +769,7 @@ static int mpxy_mbox_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
 	struct mpxy_mbox_channel *mchan;
+	struct irq_domain *msi_domain;
 	struct mpxy_mbox *mbox;
 	int i, msi_idx, rc;
 	u32 *channel_ids;
@@ -892,6 +895,8 @@ static int mpxy_mbox_probe(struct platform_device *pdev)
 
 	/* Set the MSI domain if not available */
 	if (!dev_get_msi_domain(dev)) {
+		struct fwnode_handle *fwnode = dev_fwnode(dev);
+
 		/*
 		 * The device MSI domain for OF devices is only set at the
 		 * time of populating/creating OF device. If the device MSI
@@ -899,8 +904,13 @@ static int mpxy_mbox_probe(struct platform_device *pdev)
 		 * then we need to set it explicitly before using any platform
 		 * MSI functions.
 		 */
-		if (is_of_node(dev_fwnode(dev)))
-			of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
+		if (is_of_node(fwnode)) {
+			of_msi_configure(dev, to_of_node(fwnode));
+		} else if (is_acpi_device_node(fwnode)) {
+			msi_domain = irq_find_matching_fwnode(imsic_acpi_get_fwnode(dev),
+							      DOMAIN_BUS_PLATFORM_MSI);
+			dev_set_msi_domain(dev, msi_domain);
+		}
 	}
 
 	/* Setup MSIs for mailbox (if required) */
@@ -945,6 +955,13 @@ static int mpxy_mbox_probe(struct platform_device *pdev)
 		return rc;
 	}
 
+#ifdef CONFIG_ACPI
+	struct acpi_device *adev = ACPI_COMPANION(dev);
+
+	if (adev)
+		acpi_dev_clear_dependencies(adev);
+#endif
+
 	dev_info(dev, "mailbox registered with %d channels\n",
 		 mbox->channel_count);
 	return 0;
@@ -964,10 +981,17 @@ static const struct of_device_id mpxy_mbox_of_match[] = {
 };
 MODULE_DEVICE_TABLE(of, mpxy_mbox_of_match);
 
+static const struct acpi_device_id mpxy_mbox_acpi_match[] = {
+	{ "RSCV0005" },
+	{}
+};
+MODULE_DEVICE_TABLE(acpi, mpxy_mbox_acpi_match);
+
 static struct platform_driver mpxy_mbox_driver = {
 	.driver = {
 		.name = "riscv-sbi-mpxy-mbox",
 		.of_match_table = mpxy_mbox_of_match,
+		.acpi_match_table = mpxy_mbox_acpi_match,
 	},
 	.probe = mpxy_mbox_probe,
 	.remove = mpxy_mbox_remove,
-- 
2.43.0

[PATCH v4 21/23] irqchip/riscv-rpmi-sysmsi: Add ACPI support

[Anup Patel]

From: Sunil V L <sunilvl@ventanamicro.com>

Add ACPI support for the RISC-V RPMI system MSI based irqchip driver.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 drivers/irqchip/Kconfig                 |  2 +-
 drivers/irqchip/irq-riscv-rpmi-sysmsi.c | 48 ++++++++++++++++++++++---
 2 files changed, 45 insertions(+), 5 deletions(-)

diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
index 83700fc2ddc9..132e56a596fc 100644
--- a/drivers/irqchip/Kconfig
+++ b/drivers/irqchip/Kconfig
@@ -614,7 +614,7 @@ config RISCV_IMSIC
 
 config RISCV_RPMI_SYSMSI
 	bool
-	depends on MAILBOX
+	depends on RISCV && MAILBOX
 	select IRQ_DOMAIN_HIERARCHY
 	select GENERIC_MSI_IRQ
 	default RISCV
diff --git a/drivers/irqchip/irq-riscv-rpmi-sysmsi.c b/drivers/irqchip/irq-riscv-rpmi-sysmsi.c
index d7b19fb7bf4e..6d27c036928a 100644
--- a/drivers/irqchip/irq-riscv-rpmi-sysmsi.c
+++ b/drivers/irqchip/irq-riscv-rpmi-sysmsi.c
@@ -1,9 +1,11 @@
 // SPDX-License-Identifier: GPL-2.0
 /* Copyright (C) 2025 Ventana Micro Systems Inc. */
 
+#include <linux/acpi.h>
 #include <linux/device.h>
 #include <linux/dev_printk.h>
 #include <linux/irq.h>
+#include <linux/irqchip/riscv-imsic.h>
 #include <linux/irqdomain.h>
 #include <linux/mailbox_client.h>
 #include <linux/mailbox/riscv-rpmi-message.h>
@@ -209,6 +211,9 @@ static int rpmi_sysmsi_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
 	struct rpmi_sysmsi_priv *priv;
+	struct irq_domain *msi_domain;
+	struct fwnode_handle *fwnode;
+	u32 id;
 	int rc;
 
 	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
@@ -240,6 +245,22 @@ static int rpmi_sysmsi_probe(struct platform_device *pdev)
 	}
 	priv->nr_irqs = rc;
 
+	fwnode = dev_fwnode(dev);
+	if (is_acpi_node(fwnode)) {
+		u32 nr_irqs;
+
+		rc = riscv_acpi_get_gsi_info(fwnode, &priv->gsi_base, &id,
+					     &nr_irqs, NULL);
+		if (rc) {
+			dev_err(dev, "failed to find GSI mapping\n");
+			return rc;
+		}
+
+		/* Update with actual GSI range */
+		if (nr_irqs != priv->nr_irqs)
+			riscv_acpi_update_gsi_range(priv->gsi_base, priv->nr_irqs);
+	}
+
 	/* Set the device MSI domain if not available */
 	if (!dev_get_msi_domain(dev)) {
 		/*
@@ -249,8 +270,13 @@ static int rpmi_sysmsi_probe(struct platform_device *pdev)
 		 * then we need to set it explicitly before using any platform
 		 * MSI functions.
 		 */
-		if (is_of_node(dev_fwnode(dev)))
-			of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
+		if (is_of_node(fwnode)) {
+			of_msi_configure(dev, to_of_node(fwnode));
+		} else if (is_acpi_device_node(fwnode)) {
+			msi_domain = irq_find_matching_fwnode(imsic_acpi_get_fwnode(dev),
+							      DOMAIN_BUS_PLATFORM_MSI);
+			dev_set_msi_domain(dev, msi_domain);
+		}
 
 		if (!dev_get_msi_domain(dev)) {
 			mbox_free_channel(priv->chan);
@@ -265,6 +291,13 @@ static int rpmi_sysmsi_probe(struct platform_device *pdev)
 		return dev_err_probe(dev, -ENOMEM, "failed to create MSI irq domain\n");
 	}
 
+#ifdef CONFIG_ACPI
+	struct acpi_device *adev = ACPI_COMPANION(dev);
+
+	if (adev)
+		acpi_dev_clear_dependencies(adev);
+#endif
+
 	dev_info(dev, "%u system MSIs registered\n", priv->nr_irqs);
 	return 0;
 }
@@ -274,10 +307,17 @@ static const struct of_device_id rpmi_sysmsi_match[] = {
 	{}
 };
 
+static const struct acpi_device_id acpi_rpmi_sysmsi_match[] = {
+	{ "RSCV0006" },
+	{}
+};
+MODULE_DEVICE_TABLE(acpi, acpi_rpmi_sysmsi_match);
+
 static struct platform_driver rpmi_sysmsi_driver = {
 	.driver = {
-		.name		= "rpmi-sysmsi",
-		.of_match_table	= rpmi_sysmsi_match,
+		.name			= "rpmi-sysmsi",
+		.of_match_table		= rpmi_sysmsi_match,
+		.acpi_match_table	= acpi_rpmi_sysmsi_match,
 	},
 	.probe = rpmi_sysmsi_probe,
 };
-- 
2.43.0

[PATCH v4 22/23] RISC-V: Enable GPIO keyboard and event device in RV64 defconfig

[Anup Patel]

The GPIO keyboard and event device can be used to receive graceful
shutdown or reboot input keys so let us enable it by default for
RV64 (just like ARM64).

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 arch/riscv/configs/defconfig | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/riscv/configs/defconfig b/arch/riscv/configs/defconfig
index 3c8e16d71e17..6aa330f4b67f 100644
--- a/arch/riscv/configs/defconfig
+++ b/arch/riscv/configs/defconfig
@@ -141,6 +141,8 @@ CONFIG_MICREL_PHY=y
 CONFIG_MICROSEMI_PHY=y
 CONFIG_MOTORCOMM_PHY=y
 CONFIG_INPUT_MOUSEDEV=y
+CONFIG_INPUT_EVDEV=y
+CONFIG_KEYBOARD_GPIO=y
 CONFIG_KEYBOARD_SUN4I_LRADC=m
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
-- 
2.43.0

[PATCH v4 23/23] MAINTAINERS: Add entry for RISC-V RPMI and MPXY drivers

[Anup Patel]

Add Rahul and myself as maintainers for RISC-V RPMI and MPXY drivers.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
---
 MAINTAINERS | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index d48dd6726fe6..f09b865a697e 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -20963,6 +20963,21 @@ F:	drivers/perf/riscv_pmu.c
 F:	drivers/perf/riscv_pmu_legacy.c
 F:	drivers/perf/riscv_pmu_sbi.c
 
+RISC-V RPMI AND MPXY DRIVERS
+M:	Rahul Pathak <rahul@summations.net>
+M:	Anup Patel <anup@brainfault.org>
+L:	linux-riscv@lists.infradead.org
+F:	Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
+F:	Documentation/devicetree/bindings/clock/riscv,rpmi-mpxy-clock.yaml
+F:	Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
+F:	Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
+F:	Documentation/devicetree/bindings/mailbox/riscv,rpmi-shmem-mbox.yaml
+F:	Documentation/devicetree/bindings/mailbox/riscv,sbi-mpxy-mbox.yaml
+F:	drivers/clk/clk-rpmi.c
+F:	drivers/irqchip/irq-riscv-rpmi-sysmsi.c
+F:	drivers/mailbox/riscv-sbi-mpxy-mbox.c
+F:	include/linux/mailbox/riscv-rpmi-message.h
+
 RISC-V SPACEMIT SoC Support
 M:	Yixun Lan <dlan@gentoo.org>
 L:	linux-riscv@lists.infradead.org
-- 
2.43.0

Re: [PATCH v4 04/23] RISC-V: Add defines for the SBI message proxy extension

[Andy Shevchenko]

On Sun, May 25, 2025 at 02:16:51PM +0530, Anup Patel wrote:
> Add defines for the new SBI message proxy extension which is part
> of the SBI v3.0 specification.

...

> +/** RPMI message protocol specific MPXY attributes */

If you do a kernel-doc, do it properly.

> +enum sbi_mpxy_rpmi_attribute_id {
> +	SBI_MPXY_RPMI_ATTR_SERVICEGROUP_ID = SBI_MPXY_ATTR_MSGPROTO_ATTR_START,
> +	SBI_MPXY_RPMI_ATTR_SERVICEGROUP_VERSION,

> +	SBI_MPXY_RPMI_ATTR_MAX_ID,

Remove trailing comma for the terminator.

> +};
> +
> +/* Encoding of MSG_PROT_VER attribute */
> +#define SBI_MPXY_MSG_PROT_VER_MAJOR(__ver)	(((__ver) >> 16) & 0xffff)
> +#define SBI_MPXY_MSG_PROT_VER_MINOR(__ver)	((__ver) & 0xffff)

This can utilise GENAMSK() or even upper_16_bits()/lower_16_bits().

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 05/23] mailbox: Add common header for RPMI messages sent via mailbox

[Andy Shevchenko]

On Sun, May 25, 2025 at 02:16:52PM +0530, Anup Patel wrote:
> The RPMI based mailbox controller drivers and mailbox clients need to
> share defines related to RPMI messages over mailbox interface so add
> a common header for this purpose.

...

> +#include <linux/mailbox_client.h>

This is not even closer to the list of the headers the header is using.
E.g., types.h is missing.

> +/* RPMI version encode/decode macros */
> +#define RPMI_VER_MAJOR(__ver)		(((__ver) >> 16) & 0xffff)
> +#define RPMI_VER_MINOR(__ver)		((__ver) & 0xffff)

Same comment as per previous patch.

...

> +	RPMI_ERR_NO_DATA		= -14,
> +	RPMI_ERR_RESERVED_START		= -15,
> +	RPMI_ERR_RESERVED_END		= -127,
> +	RPMI_ERR_VENDOR_START		= -128

Leave the trailing comma, as it doesn't sound like a terminator.

...

> +		return -ETIMEDOUT;
> +		return -ECOMM;
> +		return -EOPNOTSUPP;

+ errno.h

...

> +/* RPMI linux mailbox attribute IDs */
> +enum rpmi_mbox_attribute_id {
> +	RPMI_MBOX_ATTR_SPEC_VERSION = 0,

Why do you need an explicit initialiser? If it's a HW requirement, all of them
should be explicitly defined. This makes code robust against potential changes.

> +	RPMI_MBOX_ATTR_MAX_MSG_DATA_SIZE,
> +	RPMI_MBOX_ATTR_SERVICEGROUP_ID,
> +	RPMI_MBOX_ATTR_SERVICEGROUP_VERSION,
> +	RPMI_MBOX_ATTR_MAX_ID
> +};

...

> +/* RPMI linux mailbox message types */

linux --> Linux
(everywhere)

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 13/23] irqchip: Add driver for the RPMI system MSI service group

[Andy Shevchenko]

On Sun, May 25, 2025 at 02:17:00PM +0530, Anup Patel wrote:
> The RPMI specification defines a system MSI service group which
> allows application processors to receive MSIs upon system events
> such as graceful shutdown/reboot request, CPU hotplug event, memory
> hotplug event, etc.
> 
> Add an irqchip driver for the RISC-V RPMI system MSI service group
> to directly receive system MSIs in Linux kernel.

...

> +#include <linux/device.h>

Perhaps I missed something, but devm_kzalloc() is in device/devres.h. Do you
need it for something else?

> +#include <linux/dev_printk.h>
> +#include <linux/irq.h>
> +#include <linux/irqdomain.h>
> +#include <linux/mailbox_client.h>
> +#include <linux/mailbox/riscv-rpmi-message.h>
> +#include <linux/module.h>
> +#include <linux/msi.h>
> +#include <linux/of_irq.h>
> +#include <linux/platform_device.h>
> +#include <linux/types.h>

> +#include <vdso/bits.h>

Just make it linux/bits.h as vdso is for user space libvdso and related.

But

+ asm/byteorder.h

...

> +static void rpmi_sysmsi_irq_mask(struct irq_data *d)
> +{
> +	struct rpmi_sysmsi_priv *priv = irq_data_get_irq_chip_data(d);

Declare temporary variable for hwirq and do irqd_to_hwirq() only once.

> +	int ret;
> +
> +	ret = rpmi_sysmsi_set_msi_state(priv, irqd_to_hwirq(d), 0);
> +	if (ret) {
> +		dev_warn(priv->dev, "Failed to mask hwirq %lu (error %d)\n",
> +			 irqd_to_hwirq(d), ret);
> +	}
> +	irq_chip_mask_parent(d);
> +}

...

> +static void rpmi_sysmsi_irq_unmask(struct irq_data *d)

Ditto.

...

> +static void rpmi_sysmsi_set_desc(msi_alloc_info_t *arg, struct msi_desc *desc)
> +{
> +	arg->desc = desc;
> +	arg->hwirq = (u32)desc->data.icookie.value;

Hmm... Why do you need an explicit casting?

> +}

...

> +	if (WARN_ON(fwspec->param_count < 1))

+ bug.h

> +		return -EINVAL;

+ errno.h (but actually you need err.h due to PTR_ERR() et al.)

...

> +static int rpmi_sysmsi_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct rpmi_sysmsi_priv *priv;
> +	int rc;

Be consistent with variable naming for the same (semantically) stuff.

> +	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return -ENOMEM;
> +	priv->dev = dev;

> +	platform_set_drvdata(pdev, priv);

How is being used?

> +
> +	/* Setup mailbox client */
> +	priv->client.dev		= priv->dev;
> +	priv->client.rx_callback	= NULL;
> +	priv->client.tx_block		= false;
> +	priv->client.knows_txdone	= true;
> +	priv->client.tx_tout		= 0;
> +
> +	/* Request mailbox channel */
> +	priv->chan = mbox_request_channel(&priv->client, 0);
> +	if (IS_ERR(priv->chan))
> +		return PTR_ERR(priv->chan);
> +
> +	/* Get number of system MSIs */
> +	rc = rpmi_sysmsi_get_num_msi(priv);
> +	if (rc < 1) {
> +		mbox_free_channel(priv->chan);
> +		if (rc)
> +			return dev_err_probe(dev, rc, "Failed to get number of system MSIs\n");
> +		else
> +			return dev_err_probe(dev, -ENODEV, "No system MSIs found\n");
> +	}
> +	priv->nr_irqs = rc;
> +
> +	/* Set the device MSI domain if not available */
> +	if (!dev_get_msi_domain(dev)) {
> +		/*
> +		 * The device MSI domain for OF devices is only set at the
> +		 * time of populating/creating OF device. If the device MSI
> +		 * domain is discovered later after the OF device is created
> +		 * then we need to set it explicitly before using any platform
> +		 * MSI functions.
> +		 */

> +		if (is_of_node(dev_fwnode(dev)))
> +			of_msi_configure(dev, to_of_node(dev_fwnode(dev)));

		if (dev_of_node(dev))
			of_msi_configure(dev, dev_of_node(dev));

> +		if (!dev_get_msi_domain(dev)) {
> +			mbox_free_channel(priv->chan);
> +			return -EPROBE_DEFER;
> +		}
> +	}
> +
> +	if (!msi_create_device_irq_domain(dev, MSI_DEFAULT_DOMAIN,
> +					  &rpmi_sysmsi_template,
> +					  priv->nr_irqs, priv, priv)) {
> +		mbox_free_channel(priv->chan);
> +		return dev_err_probe(dev, -ENOMEM, "failed to create MSI irq domain\n");
> +	}
> +
> +	dev_info(dev, "%u system MSIs registered\n", priv->nr_irqs);
> +	return 0;
> +}

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 06/23] mailbox: Allow controller specific mapping using fwnode

[Andy Shevchenko]

On Sun, May 25, 2025 at 02:16:53PM +0530, Anup Patel wrote:
> Introduce optional fw_node() callback which allows a mailbox controller
> driver to provide controller specific mapping using fwnode.
> 
> The Linux OF framework already implements fwnode operations for the
> Linux DD framework so the fw_xlate() callback works fine with device
> tree as well.

...

>  struct mbox_chan *mbox_request_channel(struct mbox_client *cl, int index)
>  {
> +	struct fwnode_reference_args fwspec;

+ property.h (if not done yet)

> -	int ret;
> +	int i, ret;

Why is 'i' signed?

> -	if (!dev || !dev->of_node) {
> -		pr_debug("%s: No owner device node\n", __func__);
> +	if (!dev || !dev->fwnode) {

Do not dereference fwnode directly. Use dev_fwnode.

> +		pr_debug("%s: No owner %s\n", __func__, !dev ? "device" : "fwnode");

Use positive conditional.

__func__ is redundant it debug messages. With Dynamic Debug enabled it may be
switched at run-time.

>  		return ERR_PTR(-ENODEV);
>  	}
>  
> -	ret = of_parse_phandle_with_args(dev->of_node, "mboxes", "#mbox-cells",
> -					 index, &spec);
> +	ret = fwnode_property_get_reference_args(dev->fwnode, "mboxes",

	struct fwnode_handle *fwnode = dev_fwnode(dev);

> +						 "#mbox-cells", 0, index, &fwspec);
>  	if (ret) {
>  		dev_dbg(dev, "%s: can't parse \"mboxes\" property\n", __func__);
>  		return ERR_PTR(ret);
>  	}

> +	memset(&spec, 0, sizeof(spec));
> +	if (dev->of_node) {

What is this check for?

> +		spec.np = to_of_node(fwspec.fwnode);
> +		spec.args_count = fwspec.nargs;
> +		for (i = 0; i < spec.args_count; i++)
> +			spec.args[i] = fwspec.args[i];
> +	}
> +
>  	mutex_lock(&con_mutex);
>  
>  	chan = ERR_PTR(-EPROBE_DEFER);
> -	list_for_each_entry(mbox, &mbox_cons, node)
> -		if (mbox->dev->of_node == spec.np) {
> +	list_for_each_entry(mbox, &mbox_cons, node) {
> +		if (mbox->fw_xlate && mbox->dev->fwnode == fwspec.fwnode) {
> +			chan = mbox->fw_xlate(mbox, &fwspec);
> +			if (!IS_ERR(chan))
> +				break;
> +		} else if (mbox->of_xlate && mbox->dev->of_node == spec.np) {
>  			chan = mbox->of_xlate(mbox, &spec);
>  			if (!IS_ERR(chan))
>  				break;
>  		}


		if (!IS_ERR(...))
			break;

is common.


> +	}

...

> +fw_mbox_index_xlate(struct mbox_controller *mbox,
> +		    const struct fwnode_reference_args *sp)

One line?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Andy Shevchenko]

On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:
> Add a mailbox controller driver for the new SBI message proxy extension
> which is part of the SBI v3.0 specification.

...

> +#include <asm/sbi.h>

asm/* usually goes after generic linux/* ones. Why here?

> +#include <linux/cpu.h>
> +#include <linux/err.h>
> +#include <linux/init.h>
> +#include <linux/jump_label.h>

> +#include <linux/kernel.h>

What for?

> +#include <linux/mailbox_controller.h>
> +#include <linux/mailbox/riscv-rpmi-message.h>
> +#include <linux/mm.h>
> +#include <linux/module.h>
> +#include <linux/msi.h>
> +#include <linux/of_irq.h>
> +#include <linux/percpu.h>
> +#include <linux/platform_device.h>
> +#include <linux/smp.h>

Please, double check that you follow IWYU (Include What You Use) principle and
have no "proxy" headers.

(E.g., types.h is missing here, but it got included via one of the above, like
kernel.h)

...

> +
> +	get_cpu();
> +
> +	/* Get the remaining and returned fields to calculate total */
> +	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_CHANNEL_IDS,
> +			 0, 0, 0, 0, 0, 0);
> +	if (!sret.error) {

Can it be standard pattern?

	if (error)
		goto err_put_cpu;

Ditto for other similar cases.

> +		remaining = le32_to_cpu(sdata->remaining);
> +		returned = le32_to_cpu(sdata->returned);
> +		*channel_count = remaining + returned;
> +	}
> +
> +	put_cpu();
> +	return sbi_err_map_linux_errno(sret.error);
> +}

...

> +static int mpxy_get_channel_ids(u32 channel_count, u32 *channel_ids)
> +{
> +	u32 remaining, returned, sidx, start_index = 0, cidx = 0;
> +	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
> +	struct sbi_mpxy_channel_ids_data *sdata = mpxy->shmem;
> +	struct sbiret sret;
> +
> +	if (!mpxy->shmem_active)
> +		return -ENODEV;
> +	if (!channel_count || !channel_ids)
> +		return -EINVAL;
> +
> +	get_cpu();
> +
> +	do {
> +		sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_CHANNEL_IDS,
> +				 start_index, 0, 0, 0, 0, 0);
> +		if (sret.error)
> +			goto done;
> +
> +		remaining = le32_to_cpu(sdata->remaining);
> +		returned = le32_to_cpu(sdata->returned);
> +
> +		for (sidx = 0; sidx < returned && cidx < channel_count; sidx++) {
> +			channel_ids[cidx] = le32_to_cpu(sdata->channel_array[sidx]);
> +			cidx += 1;
> +		}
> +
> +		start_index = cidx;
> +
> +	} while (remaining);
> +
> +done:

It sounds to me like an 'err_put_cpu' for the name.

> +	put_cpu();
> +	return sbi_err_map_linux_errno(sret.error);
> +}

> +static int mpxy_write_attrs(u32 channel_id, u32 base_attrid, u32 attr_count,
> +			    u32 *attrs_buf)
> +{
> +	struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
> +	struct sbiret sret;
> +	u32 i;
> +
> +	if (!mpxy->shmem_active)
> +		return -ENODEV;
> +	if (!attr_count || !attrs_buf)
> +		return -EINVAL;
> +
> +	get_cpu();
> +
> +	for (i = 0; i < attr_count; i++)
> +		((__le32 *)mpxy->shmem)[i] = cpu_to_le32(attrs_buf[i]);

Don't we have helpers for this? They are suffixed with _array.
https://elixir.bootlin.com/linux/v6.15-rc6/source/include/linux/byteorder/generic.h#L168
Don't forget to have asm/byteorder.h being included.

Ditto for the similar case(s).

> +	sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_WRITE_ATTRS,
> +			 channel_id, base_attrid, attr_count, 0, 0, 0);
> +
> +	put_cpu();
> +	return sbi_err_map_linux_errno(sret.error);
> +}

...

> +	msg->error = rc;
> +	return 0;

What's the point of having int and not void function?

...

> +			       sizeof(mchan->rpmi_attrs) / sizeof(u32),
> +			       (u32 *)&mchan->rpmi_attrs);

Why casting? What about alignment?

...

> +	while (1) {

Do it as do {} while, it will give an idea that the loop will run at least once
(without looking into condition), ideally it shouldn't be infinite loop, but it
might be harder to realise.

> +		rc = mpxy_get_notifications(mchan->channel_id, notif, &data_len);
> +		if (rc || !data_len)
> +			break;
> +
> +		if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID)
> +			mpxy_mbox_peek_rpmi_data(chan, mchan, notif, data_len);
> +
> +		have_notifications = true;
> +	}

...

> +static void mpxy_mbox_cleanup_events(struct mpxy_mbox_channel *mchan)
> +{
> +	struct device *dev = mchan->mbox->dev;
> +	int rc;
> +
> +	/* Do nothing if events state not supported */
> +	if (!mchan->have_events_state)
> +		return;
> +
> +	/* Do nothing if events state already disabled */
> +	if (!mchan->attrs.events_state_ctrl)
> +		return;
> +
> +	/* Disable channel events state */
> +	mchan->attrs.events_state_ctrl = 0;
> +	rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_EVENTS_STATE_CONTROL,
> +			      1, &mchan->attrs.events_state_ctrl);

> +	if (rc) {
> +		dev_err(dev, "disable events state failed for MPXY channel 0x%x\n",
> +			mchan->channel_id);
> +	}

Redundant {}.

> +}

...

> +static int mpxy_mbox_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct mpxy_mbox_channel *mchan;
> +	struct mpxy_mbox *mbox;
> +	int i, msi_idx, rc;

Why is 'i' signed? Also make sure either they use POD (like unsigned int) or
exact data type the upper limit does or API where you pass it in.

> +	u32 *channel_ids;
> +
> +	/*
> +	 * Initialize MPXY shared memory only once. This also ensures
> +	 * that SBI MPXY mailbox is probed only once.
> +	 */
> +	if (mpxy_shmem_init_done) {
> +		dev_err(dev, "SBI MPXY mailbox already initialized\n");
> +		return -EALREADY;
> +	}
> +
> +	/* Probe for SBI MPXY extension */
> +	if (sbi_spec_version < sbi_mk_version(1, 0) ||
> +	    sbi_probe_extension(SBI_EXT_MPXY) <= 0) {
> +		dev_info(dev, "SBI MPXY extension not available\n");
> +		return -ENODEV;
> +	}
> +
> +	/* Find-out shared memory size */
> +	mpxy_shmem_size = mpxy_get_shmem_size();
> +
> +	/*
> +	 * Setup MPXY shared memory on each CPU
> +	 *
> +	 * Note: Don't cleanup MPXY shared memory upon CPU power-down
> +	 * because the RPMI System MSI irqchip driver needs it to be
> +	 * available when migrating IRQs in CPU power-down path.
> +	 */
> +	cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "riscv/sbi-mpxy-shmem",
> +			  mpxy_setup_shmem, NULL);
> +
> +	/* Mark as MPXY shared memory initialization done */
> +	mpxy_shmem_init_done = true;
> +
> +	/* Allocate mailbox instance */
> +	mbox = devm_kzalloc(dev, sizeof(*mbox), GFP_KERNEL);
> +	if (!mbox)
> +		return -ENOMEM;
> +	mbox->dev = dev;
> +	platform_set_drvdata(pdev, mbox);
> +
> +	/* Find-out of number of channels */
> +	rc = mpxy_get_channel_count(&mbox->channel_count);
> +	if (rc)
> +		return dev_err_probe(dev, rc, "failed to get number of MPXY channels\n");
> +	if (!mbox->channel_count)
> +		dev_err_probe(dev, -ENODEV, "no MPXY channels available\n");
> +
> +	/* Allocate and fetch all channel IDs */
> +	channel_ids = devm_kcalloc(dev, mbox->channel_count,
> +				   sizeof(*channel_ids), GFP_KERNEL);
> +	if (!channel_ids)
> +		return -ENOMEM;
> +	rc = mpxy_get_channel_ids(mbox->channel_count, channel_ids);
> +	if (rc)
> +		return dev_err_probe(dev, rc, "failed to MPXY channel IDs\n");
> +
> +	/* Populate all channels */
> +	mbox->channels = devm_kcalloc(dev, mbox->channel_count,
> +				      sizeof(*mbox->channels), GFP_KERNEL);
> +	if (!mbox->channels)
> +		return -ENOMEM;
> +	for (i = 0; i < mbox->channel_count; i++) {
> +		mchan = &mbox->channels[i];
> +		mchan->mbox = mbox;
> +		mchan->channel_id = channel_ids[i];
> +
> +		rc = mpxy_read_attrs(mchan->channel_id, SBI_MPXY_ATTR_MSG_PROT_ID,
> +				     sizeof(mchan->attrs) / sizeof(u32),
> +				     (u32 *)&mchan->attrs);
> +		if (rc) {
> +			return dev_err_probe(dev, rc,
> +					     "MPXY channel 0x%x read attrs failed\n",
> +					     mchan->channel_id);
> +		}
> +
> +		if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID) {
> +			rc = mpxy_mbox_read_rpmi_attrs(mchan);
> +			if (rc) {
> +				return dev_err_probe(dev, rc,
> +						     "MPXY channel 0x%x read RPMI attrs failed\n",
> +						     mchan->channel_id);
> +			}
> +		}
> +
> +		mchan->notif = devm_kzalloc(dev, mpxy_shmem_size, GFP_KERNEL);
> +		if (!mchan->notif)
> +			return -ENOMEM;
> +
> +		mchan->max_xfer_len = min(mpxy_shmem_size, mchan->attrs.msg_max_len);
> +
> +		if ((mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS) &&
> +		    (mchan->attrs.capability & SBI_MPXY_CHAN_CAP_EVENTS_STATE))
> +			mchan->have_events_state = true;
> +
> +		if ((mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS) &&
> +		    (mchan->attrs.capability & SBI_MPXY_CHAN_CAP_MSI))
> +			mchan->msi_index = mbox->msi_count++;
> +		else
> +			mchan->msi_index = U32_MAX;
> +		mchan->msi_irq = U32_MAX;
> +	}

> +	/* Free-up channel IDs */
> +	devm_kfree(dev, channel_ids);

Just no. This is most likely (99.9%) wrong use of managed resources. This is
the biggest issue with this patch so far.

> +	/* Initialize mailbox controller */
> +	mbox->controller.txdone_irq = false;
> +	mbox->controller.txdone_poll = false;
> +	mbox->controller.ops = &mpxy_mbox_ops;
> +	mbox->controller.dev = dev;
> +	mbox->controller.num_chans = mbox->channel_count;
> +	mbox->controller.fw_xlate = mpxy_mbox_fw_xlate;
> +	mbox->controller.chans = devm_kcalloc(dev, mbox->channel_count,
> +					      sizeof(*mbox->controller.chans),
> +					      GFP_KERNEL);
> +	if (!mbox->controller.chans)
> +		return -ENOMEM;
> +	for (i = 0; i < mbox->channel_count; i++)
> +		mbox->controller.chans[i].con_priv = &mbox->channels[i];
> +
> +	/* Set the MSI domain if not available */
> +	if (!dev_get_msi_domain(dev)) {
> +		/*
> +		 * The device MSI domain for OF devices is only set at the
> +		 * time of populating/creating OF device. If the device MSI
> +		 * domain is discovered later after the OF device is created
> +		 * then we need to set it explicitly before using any platform
> +		 * MSI functions.
> +		 */
> +		if (is_of_node(dev_fwnode(dev)))
> +			of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
> +	}
> +
> +	/* Setup MSIs for mailbox (if required) */
> +	if (mbox->msi_count) {
> +		mbox->msi_index_to_channel = devm_kcalloc(dev, mbox->msi_count,
> +							  sizeof(*mbox->msi_index_to_channel),
> +							  GFP_KERNEL);
> +		if (!mbox->msi_index_to_channel)
> +			return -ENOMEM;
> +
> +		for (msi_idx = 0; msi_idx < mbox->msi_count; msi_idx++) {
> +			for (i = 0; i < mbox->channel_count; i++) {
> +				mchan = &mbox->channels[i];
> +				if (mchan->msi_index == msi_idx) {
> +					mbox->msi_index_to_channel[msi_idx] = mchan;
> +					break;
> +				}
> +			}
> +		}
> +
> +		rc = platform_device_msi_init_and_alloc_irqs(dev, mbox->msi_count,
> +							     mpxy_mbox_msi_write);
> +		if (rc) {
> +			return dev_err_probe(dev, rc, "Failed to allocate %d MSIs\n",
> +					     mbox->msi_count);
> +		}
> +
> +		for (i = 0; i < mbox->channel_count; i++) {
> +			mchan = &mbox->channels[i];
> +			if (mchan->msi_index == U32_MAX)
> +				continue;
> +			mchan->msi_irq = msi_get_virq(dev, mchan->msi_index);
> +		}
> +	}
> +
> +	/* Register mailbox controller */
> +	rc = devm_mbox_controller_register(dev, &mbox->controller);
> +	if (rc) {
> +		dev_err_probe(dev, rc, "Registering SBI MPXY mailbox failed\n");
> +		if (mbox->msi_count)
> +			platform_device_msi_free_irqs_all(dev);
> +		return rc;
> +	}
> +
> +	dev_info(dev, "mailbox registered with %d channels\n",
> +		 mbox->channel_count);
> +	return 0;
> +}

...

> +	if (mbox->msi_count)

Is this check really needed?

> +		platform_device_msi_free_irqs_all(mbox->dev);
> +}

...

> +static const struct of_device_id mpxy_mbox_of_match[] = {
> +	{.compatible = "riscv,sbi-mpxy-mbox", },
> +	{},

No comma for the terminator entry.
You should apply the reviewers' comments on all of your code, not just where it
was commented.

> +};

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 17/23] ACPI: RISC-V: Create interrupt controller list in sorted order

[Andy Shevchenko]

On Sun, May 25, 2025 at 02:17:04PM +0530, Anup Patel wrote:
> 
> Currently, the interrupt controller list is created without any order.
> Create the list sorted with the GSI base of the interrupt controllers.

...

> -	list_add_tail(&ext_intc_element->list, &ext_intc_list);
> +	if (list_empty(&ext_intc_list)) {
> +		list_add(&ext_intc_element->list, &ext_intc_list);
> +		return 0;
> +	}

With the below done the above can be optimized (hopefully).

> +	list_for_each_entry(node, &ext_intc_list, list) {
> +		if (node->gsi_base < ext_intc_element->gsi_base)
> +			break;
> +	}
> +
> +	__list_add(&ext_intc_element->list, node->list.prev, &node->list);

Is this reimplementation of list_add_tail()? And why list debug is excluded here?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 08/23] dt-bindings: clock: Add RPMI clock service message proxy bindings

[Conor Dooley]

[-- Attachment #1: Type: text/plain, Size: 736 bytes --]

On Sun, May 25, 2025 at 02:16:55PM +0530, Anup Patel wrote:
> +  riscv,sbi-mpxy-channel-id:
> +    $ref: /schemas/types.yaml#/definitions/uint32
> +    description:
> +      The SBI MPXY channel id to be used for providing RPMI access to
> +      the supervisor software.

Sorry for the delay in reviewing this series, wanted to talk to some
folks at work before doing so. Overall these bindings all look pretty
good to me. I don't think the description on this property here is
particularly great, I didn't follow what it was meant to be used for
immediately. But I don't have anything meaningfully better to
suggest nor do I think that anyone actually writing the dts for a
platform using this binding will struggle to understand it.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH v4 09/23] dt-bindings: clock: Add RPMI clock service controller bindings

[Conor Dooley]

[-- Attachment #1: Type: text/plain, Size: 3694 bytes --]

On Sun, May 25, 2025 at 02:16:56PM +0530, Anup Patel wrote:
> Add device tree bindings for the RPMI clock service group based
> controller for the supervisor software.
> 
> The RPMI clock service group is defined by the RISC-V platform
> management interface (RPMI) specification.
> 
> Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> ---
>  .../bindings/clock/riscv,rpmi-clock.yaml      | 61 +++++++++++++++++++
>  1 file changed, 61 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
> 
> diff --git a/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml b/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
> new file mode 100644
> index 000000000000..9c672a38595a
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
> @@ -0,0 +1,61 @@
> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/clock/riscv,rpmi-clock.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: RISC-V RPMI clock service group based clock controller
> +
> +maintainers:
> +  - Anup Patel <anup@brainfault.org>
> +
> +description: |
> +  The RISC-V Platform Management Interface (RPMI) [1] defines a
> +  messaging protocol which is modular and extensible. The supervisor
> +  software can send/receive RPMI messages via SBI MPXY extension [2]
> +  or some dedicated supervisor-mode RPMI transport.
> +
> +  The RPMI specification [1] defines clock service group for accessing
> +  system clocks managed by a platform microcontroller. The supervisor
> +  software can access RPMI clock service group via SBI MPXY channel or
> +  some dedicated supervisor-mode RPMI transport.
> +
> +  ===========================================
> +  References
> +  ===========================================
> +
> +  [1] RISC-V Platform Management Interface (RPMI)
> +      https://github.com/riscv-non-isa/riscv-rpmi/releases
> +
> +  [2] RISC-V Supervisor Binary Interface (SBI)
> +      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
> +
> +properties:
> +  compatible:
> +    description:
> +      Intended for use by the supervisor software.
> +    const: riscv,rpmi-clock
> +
> +  mboxes:
> +    maxItems: 1
> +    description:
> +      Mailbox channel of the underlying RPMI transport or SBI message proxy channel.
> +
> +  "#clock-cells":
> +    const: 1

Could you please add some description here as to what this clock-cell
actually does? On a normal clock controller someone might cite an
include file with a huge list of defines for what numbers map to what
clock. In this case, this value is CLOCK_ID in the spec, so it's
completely platform specific as to what they mean so citing some include
isn't helpful, so just mention that it is CLOCK_ID and the meanings are
platform specific.

I suppose technically it can be something other than CLOCK_ID, if this is
used when the SBI message proxy extension is provided by an SBI
implementation that uses a non-RPMI transport, but I don't think that's a
can of worms worth bringing up in the binding. Anyone doing that can put
2+2 together I think.

Otherwise, the bindings are all:
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>

Cheers,
Conor.

> +
> +required:
> +  - compatible
> +  - mboxes
> +  - "#clock-cells"
> +
> +additionalProperties: false
> +
> +examples:
> +  - |
> +    clock-controller {
> +        compatible = "riscv,rpmi-clock";
> +        mboxes = <&mpxy_mbox 0x1000 0x0>;
> +        #clock-cells = <1>;
> +    };
> +...
> -- 
> 2.43.0
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH v4 11/23] dt-bindings: Add RPMI system MSI message proxy bindings

[Atish Patra]

On 5/25/25 1:46 AM, Anup Patel wrote:
> Add device tree bindings for the RPMI system MSI service group
> based message proxy implemented by the SBI implementation (machine
> mode firmware or hypervisor).
>
> The RPMI system MSI service group is defined by the RISC-V
> platform management interface (RPMI) specification.
>
> Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> ---
>   .../riscv,rpmi-mpxy-system-msi.yaml           | 67 +++++++++++++++++++
>   1 file changed, 67 insertions(+)
>   create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
>
> diff --git a/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
> new file mode 100644
> index 000000000000..26dd13731350
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
> @@ -0,0 +1,67 @@
> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: RISC-V RPMI system MSI service group based message proxy
> +
> +maintainers:
> +  - Anup Patel <anup@brainfault.org>
> +
> +description: |
> +  The RISC-V Platform Management Interface (RPMI) [1] defines a
> +  messaging protocol which is modular and extensible. The supervisor
> +  software can send/receive RPMI messages via SBI MPXY extension [2]
> +  or some dedicated supervisor-mode RPMI transport.
> +
> +  The RPMI specification [1] defines system MSI service group which
> +  allow application processors to receive MSIs upon system events
> +  such as P2A doorbell, graceful shutdown/reboot request, CPU hotplug
> +  event, memory hotplug event, etc from the platform microcontroller.
> +  The SBI implementation machine mode firmware or hypervisor) can
nit:
The SBI implementation in machine mode
> +  implement an SBI MPXY channel to allow RPMI system MSI service
> +  group access to the supervisor software.
> +
> +  ===========================================
> +  References
> +  ===========================================
> +
> +  [1] RISC-V Platform Management Interface (RPMI)
> +      https://github.com/riscv-non-isa/riscv-rpmi/releases
> +
> +  [2] RISC-V Supervisor Binary Interface (SBI)
> +      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
> +
nit: Refer the latest frozen version of the spec ?
> +properties:
> +  compatible:
> +    description:
> +      Intended for use by the SBI implementation.
> +    const: riscv,rpmi-mpxy-system-msi
> +
> +  mboxes:
> +    maxItems: 1
> +    description:
> +      Mailbox channel of the underlying RPMI transport.
> +
> +  riscv,sbi-mpxy-channel-id:
> +    $ref: /schemas/types.yaml#/definitions/uint32
> +    description:
> +      The SBI MPXY channel id to be used for providing RPMI access to
> +      the supervisor software.
> +
> +required:
> +  - compatible
> +  - mboxes
> +  - riscv,sbi-mpxy-channel-id
> +
> +additionalProperties: false
> +
> +examples:
> +  - |
> +    interrupt-controller {
> +        compatible = "riscv,rpmi-mpxy-system-msi";
> +        mboxes = <&rpmi_shmem_mbox 0x2>;
> +        riscv,sbi-mpxy-channel-id = <0x2000>;
> +    };
> +...

Otherwise, lgtm.

Reviewed-by: Atish Patra <atishp@rivosinc.com>

Re: [PATCH v4 12/23] dt-bindings: Add RPMI system MSI interrupt controller bindings

[Atish Patra]

On 5/25/25 1:46 AM, Anup Patel wrote:
> Add device tree bindings for the RPMI system MSI service group
> based interrupt controller for the supervisor software.
>
> The RPMI system MSI service group is defined by the RISC-V
> platform management interface (RPMI) specification.
>
> Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> ---
>   .../riscv,rpmi-system-msi.yaml                | 74 +++++++++++++++++++
>   1 file changed, 74 insertions(+)
>   create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
>
> diff --git a/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
> new file mode 100644
> index 000000000000..ac13cec0666e
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
> @@ -0,0 +1,74 @@
> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/interrupt-controller/riscv,rpmi-system-msi.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: RISC-V RPMI system MSI service group based interrupt controller
> +
> +maintainers:
> +  - Anup Patel <anup@brainfault.org>
> +
> +description: |
> +  The RISC-V Platform Management Interface (RPMI) [1] defines a
> +  messaging protocol which is modular and extensible. The supervisor
> +  software can send/receive RPMI messages via SBI MPXY extension [2]
> +  or some dedicated supervisor-mode RPMI transport.
> +
> +  The RPMI specification [1] defines system MSI service group which
> +  allow application processors to receive MSIs upon system events
> +  such as P2A doorbell, graceful shutdown/reboot request, CPU hotplug
> +  event, memory hotplug event, etc from the platform microcontroller.
> +  The supervisor software can access RPMI system MSI service group via
> +  SBI MPXY channel or some dedicated supervisor-mode RPMI transport.
> +
> +  ===========================================
> +  References
> +  ===========================================
> +
> +  [1] RISC-V Platform Management Interface (RPMI)
> +      https://github.com/riscv-non-isa/riscv-rpmi/releases
> +
> +  [2] RISC-V Supervisor Binary Interface (SBI)
> +      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
> +
nit: Same comment as previous patch.
> +allOf:
> +  - $ref: /schemas/interrupt-controller.yaml#
> +
> +properties:
> +  compatible:
> +    description:
> +      Intended for use by the supervisor software.
> +    const: riscv,rpmi-system-msi
> +
> +  mboxes:
> +    maxItems: 1
> +    description:
> +      Mailbox channel of the underlying RPMI transport or SBI message proxy channel.
> +
> +  msi-parent: true
> +
> +  interrupt-controller: true
> +
> +  "#interrupt-cells":
> +    const: 1
> +
> +required:
> +  - compatible
> +  - mboxes
> +  - msi-parent
> +  - interrupt-controller
> +  - "#interrupt-cells"
> +
> +additionalProperties: false
> +
> +examples:
> +  - |
> +    interrupt-controller {
> +        compatible = "riscv,rpmi-system-msi";
> +        mboxes = <&mpxy_mbox 0x2000 0x0>;
> +        msi-parent = <&imsic_slevel>;
> +        interrupt-controller;
> +        #interrupt-cells = <1>;
> +    };
> +...

Otherwise, LGTM.
Reviewed-by: Atish Patra <atishp@rivosinc.com>

Re: [PATCH v4 01/23] riscv: Add new error codes defined by SBI v3.0

[Atish Patra]

On 5/25/25 1:46 AM, Anup Patel wrote:
> The SBI v3.0 defines new error codes so add these new error codes
> to the asm/sbi.h for use by newer SBI extensions.

This patch can be dropped as it is part of the FWFT series with minor 
modifications in error mappings.

https://lore.kernel.org/linux-riscv/20250523101932.1594077-4-cleger@rivosinc.com/


> Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> ---
>   arch/riscv/include/asm/sbi.h | 12 ++++++++++++
>   1 file changed, 12 insertions(+)
>
> diff --git a/arch/riscv/include/asm/sbi.h b/arch/riscv/include/asm/sbi.h
> index 3d250824178b..4dd6aafb8468 100644
> --- a/arch/riscv/include/asm/sbi.h
> +++ b/arch/riscv/include/asm/sbi.h
> @@ -419,6 +419,11 @@ enum sbi_ext_nacl_feature {
>   #define SBI_ERR_ALREADY_STARTED -7
>   #define SBI_ERR_ALREADY_STOPPED -8
>   #define SBI_ERR_NO_SHMEM	-9
> +#define SBI_ERR_INVALID_STATE	-10
> +#define SBI_ERR_BAD_RANGE	-11
> +#define SBI_ERR_TIMEOUT		-12
> +#define SBI_ERR_IO		-13
> +#define SBI_ERR_DENIED_LOCKED	-14
>   
>   extern unsigned long sbi_spec_version;
>   struct sbiret {
> @@ -503,11 +508,18 @@ static inline int sbi_err_map_linux_errno(int err)
>   	case SBI_SUCCESS:
>   		return 0;
>   	case SBI_ERR_DENIED:
> +	case SBI_ERR_DENIED_LOCKED:
>   		return -EPERM;
>   	case SBI_ERR_INVALID_PARAM:
> +	case SBI_ERR_INVALID_STATE:
> +	case SBI_ERR_BAD_RANGE:
>   		return -EINVAL;
>   	case SBI_ERR_INVALID_ADDRESS:
>   		return -EFAULT;
> +	case SBI_ERR_TIMEOUT:
> +		return -ETIMEDOUT;
> +	case SBI_ERR_IO:
> +		return -EIO;
>   	case SBI_ERR_NOT_SUPPORTED:
>   	case SBI_ERR_FAILURE:
>   	default:

Re: [PATCH v4 01/23] riscv: Add new error codes defined by SBI v3.0

[Anup Patel]

On Sat, Jun 7, 2025 at 5:21 AM Atish Patra <atish.patra@linux.dev> wrote:
>
>
> On 5/25/25 1:46 AM, Anup Patel wrote:
> > The SBI v3.0 defines new error codes so add these new error codes
> > to the asm/sbi.h for use by newer SBI extensions.
>
> This patch can be dropped as it is part of the FWFT series with minor
> modifications in error mappings.
>
> https://lore.kernel.org/linux-riscv/20250523101932.1594077-4-cleger@rivosinc.com/

Yes, I am aware of Clement's patch. I was temporarily carrying
this patch over here.

Regards,
Anup

>
>
> > Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> > ---
> >   arch/riscv/include/asm/sbi.h | 12 ++++++++++++
> >   1 file changed, 12 insertions(+)
> >
> > diff --git a/arch/riscv/include/asm/sbi.h b/arch/riscv/include/asm/sbi.h
> > index 3d250824178b..4dd6aafb8468 100644
> > --- a/arch/riscv/include/asm/sbi.h
> > +++ b/arch/riscv/include/asm/sbi.h
> > @@ -419,6 +419,11 @@ enum sbi_ext_nacl_feature {
> >   #define SBI_ERR_ALREADY_STARTED -7
> >   #define SBI_ERR_ALREADY_STOPPED -8
> >   #define SBI_ERR_NO_SHMEM    -9
> > +#define SBI_ERR_INVALID_STATE        -10
> > +#define SBI_ERR_BAD_RANGE    -11
> > +#define SBI_ERR_TIMEOUT              -12
> > +#define SBI_ERR_IO           -13
> > +#define SBI_ERR_DENIED_LOCKED        -14
> >
> >   extern unsigned long sbi_spec_version;
> >   struct sbiret {
> > @@ -503,11 +508,18 @@ static inline int sbi_err_map_linux_errno(int err)
> >       case SBI_SUCCESS:
> >               return 0;
> >       case SBI_ERR_DENIED:
> > +     case SBI_ERR_DENIED_LOCKED:
> >               return -EPERM;
> >       case SBI_ERR_INVALID_PARAM:
> > +     case SBI_ERR_INVALID_STATE:
> > +     case SBI_ERR_BAD_RANGE:
> >               return -EINVAL;
> >       case SBI_ERR_INVALID_ADDRESS:
> >               return -EFAULT;
> > +     case SBI_ERR_TIMEOUT:
> > +             return -ETIMEDOUT;
> > +     case SBI_ERR_IO:
> > +             return -EIO;
> >       case SBI_ERR_NOT_SUPPORTED:
> >       case SBI_ERR_FAILURE:
> >       default:

Re: [PATCH v4 04/23] RISC-V: Add defines for the SBI message proxy extension

[Anup Patel]

On Tue, May 27, 2025 at 2:17 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Sun, May 25, 2025 at 02:16:51PM +0530, Anup Patel wrote:
> > Add defines for the new SBI message proxy extension which is part
> > of the SBI v3.0 specification.
>
> ...
>
> > +/** RPMI message protocol specific MPXY attributes */
>
> If you do a kernel-doc, do it properly.

Okay, I will update.

>
> > +enum sbi_mpxy_rpmi_attribute_id {
> > +     SBI_MPXY_RPMI_ATTR_SERVICEGROUP_ID = SBI_MPXY_ATTR_MSGPROTO_ATTR_START,
> > +     SBI_MPXY_RPMI_ATTR_SERVICEGROUP_VERSION,
>
> > +     SBI_MPXY_RPMI_ATTR_MAX_ID,
>
> Remove trailing comma for the terminator.

Okay, I will update.

>
> > +};
> > +
> > +/* Encoding of MSG_PROT_VER attribute */
> > +#define SBI_MPXY_MSG_PROT_VER_MAJOR(__ver)   (((__ver) >> 16) & 0xffff)
> > +#define SBI_MPXY_MSG_PROT_VER_MINOR(__ver)   ((__ver) & 0xffff)
>
> This can utilise GENAMSK() or even upper_16_bits()/lower_16_bits().

Okay, I will update.

Regards,
Anup

Re: [PATCH v4 05/23] mailbox: Add common header for RPMI messages sent via mailbox

[Anup Patel]

On Tue, May 27, 2025 at 4:46 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Sun, May 25, 2025 at 02:16:52PM +0530, Anup Patel wrote:
> > The RPMI based mailbox controller drivers and mailbox clients need to
> > share defines related to RPMI messages over mailbox interface so add
> > a common header for this purpose.
>
> ...
>
> > +#include <linux/mailbox_client.h>
>
> This is not even closer to the list of the headers the header is using.
> E.g., types.h is missing.

Okay, I will add types.h

>
> > +/* RPMI version encode/decode macros */
> > +#define RPMI_VER_MAJOR(__ver)                (((__ver) >> 16) & 0xffff)
> > +#define RPMI_VER_MINOR(__ver)                ((__ver) & 0xffff)
>
> Same comment as per previous patch.

Okay, I will use macros from linux/wordpart.h

>
> ...
>
> > +     RPMI_ERR_NO_DATA                = -14,
> > +     RPMI_ERR_RESERVED_START         = -15,
> > +     RPMI_ERR_RESERVED_END           = -127,
> > +     RPMI_ERR_VENDOR_START           = -128
>
> Leave the trailing comma, as it doesn't sound like a terminator.

Okay

>
> ...
>
> > +             return -ETIMEDOUT;
> > +             return -ECOMM;
> > +             return -EOPNOTSUPP;
>
> + errno.h

Okay, I will add errno.h

>
> ...
>
> > +/* RPMI linux mailbox attribute IDs */
> > +enum rpmi_mbox_attribute_id {
> > +     RPMI_MBOX_ATTR_SPEC_VERSION = 0,
>
> Why do you need an explicit initialiser? If it's a HW requirement, all of them
> should be explicitly defined. This makes code robust against potential changes.

Explicit initializers are not needed. I will drop in the next revision.

>
> > +     RPMI_MBOX_ATTR_MAX_MSG_DATA_SIZE,
> > +     RPMI_MBOX_ATTR_SERVICEGROUP_ID,
> > +     RPMI_MBOX_ATTR_SERVICEGROUP_VERSION,
> > +     RPMI_MBOX_ATTR_MAX_ID
> > +};
>
> ...
>
> > +/* RPMI linux mailbox message types */
>
> linux --> Linux
> (everywhere)

Okay, I will update.

Regards,
Anup

Re: [PATCH v4 06/23] mailbox: Allow controller specific mapping using fwnode

[Anup Patel]

On Tue, May 27, 2025 at 5:11 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Sun, May 25, 2025 at 02:16:53PM +0530, Anup Patel wrote:
> > Introduce optional fw_node() callback which allows a mailbox controller
> > driver to provide controller specific mapping using fwnode.
> >
> > The Linux OF framework already implements fwnode operations for the
> > Linux DD framework so the fw_xlate() callback works fine with device
> > tree as well.
>
> ...
>
> >  struct mbox_chan *mbox_request_channel(struct mbox_client *cl, int index)
> >  {
> > +     struct fwnode_reference_args fwspec;
>
> + property.h (if not done yet)

This needs fwnode.h

>
> > -     int ret;
> > +     int i, ret;
>
> Why is 'i' signed?

No particular reason because 'i' is not being assigned a negative
value so both signed and unsigned will work fine.

>
> > -     if (!dev || !dev->of_node) {
> > -             pr_debug("%s: No owner device node\n", __func__);
> > +     if (!dev || !dev->fwnode) {
>
> Do not dereference fwnode directly. Use dev_fwnode.
>
> > +             pr_debug("%s: No owner %s\n", __func__, !dev ? "device" : "fwnode");
>
> Use positive conditional.
>
> __func__ is redundant it debug messages. With Dynamic Debug enabled it may be
> switched at run-time.

Okay, I will drop the __func__

>
> >               return ERR_PTR(-ENODEV);
> >       }
> >
> > -     ret = of_parse_phandle_with_args(dev->of_node, "mboxes", "#mbox-cells",
> > -                                      index, &spec);
> > +     ret = fwnode_property_get_reference_args(dev->fwnode, "mboxes",
>
>         struct fwnode_handle *fwnode = dev_fwnode(dev);
>
> > +                                              "#mbox-cells", 0, index, &fwspec);
> >       if (ret) {
> >               dev_dbg(dev, "%s: can't parse \"mboxes\" property\n", __func__);
> >               return ERR_PTR(ret);
> >       }
>
> > +     memset(&spec, 0, sizeof(spec));
> > +     if (dev->of_node) {
>
> What is this check for?

It's a paranoid check on my part. I will drop the check
in the next revision.

>
> > +             spec.np = to_of_node(fwspec.fwnode);
> > +             spec.args_count = fwspec.nargs;
> > +             for (i = 0; i < spec.args_count; i++)
> > +                     spec.args[i] = fwspec.args[i];
> > +     }
> > +
> >       mutex_lock(&con_mutex);
> >
> >       chan = ERR_PTR(-EPROBE_DEFER);
> > -     list_for_each_entry(mbox, &mbox_cons, node)
> > -             if (mbox->dev->of_node == spec.np) {
> > +     list_for_each_entry(mbox, &mbox_cons, node) {
> > +             if (mbox->fw_xlate && mbox->dev->fwnode == fwspec.fwnode) {
> > +                     chan = mbox->fw_xlate(mbox, &fwspec);
> > +                     if (!IS_ERR(chan))
> > +                             break;
> > +             } else if (mbox->of_xlate && mbox->dev->of_node == spec.np) {
> >                       chan = mbox->of_xlate(mbox, &spec);
> >                       if (!IS_ERR(chan))
> >                               break;
> >               }
>
>
>                 if (!IS_ERR(...))
>                         break;
>
> is common.

Latest mailbox.c has changed so I will check and update accordingly.

>
>
> > +     }
>
> ...
>
> > +fw_mbox_index_xlate(struct mbox_controller *mbox,
> > +                 const struct fwnode_reference_args *sp)
>
> One line?

Okay, I will update.

Regards,
Anup

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Anup Patel]

On Wed, May 28, 2025 at 4:23 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:
> > Add a mailbox controller driver for the new SBI message proxy extension
> > which is part of the SBI v3.0 specification.
>
> ...
>
> > +#include <asm/sbi.h>
>
> asm/* usually goes after generic linux/* ones. Why here?

I am not aware of any such convention but I will update anyway.

>
> > +#include <linux/cpu.h>
> > +#include <linux/err.h>
> > +#include <linux/init.h>
> > +#include <linux/jump_label.h>
>
> > +#include <linux/kernel.h>
>
> What for?
>
> > +#include <linux/mailbox_controller.h>
> > +#include <linux/mailbox/riscv-rpmi-message.h>
> > +#include <linux/mm.h>
> > +#include <linux/module.h>
> > +#include <linux/msi.h>
> > +#include <linux/of_irq.h>
> > +#include <linux/percpu.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/smp.h>
>
> Please, double check that you follow IWYU (Include What You Use) principle and
> have no "proxy" headers.
>
> (E.g., types.h is missing here, but it got included via one of the above, like
> kernel.h)
>

Okay, I will update.

> ...
>
> > +
> > +     get_cpu();
> > +
> > +     /* Get the remaining and returned fields to calculate total */
> > +     sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_CHANNEL_IDS,
> > +                      0, 0, 0, 0, 0, 0);
> > +     if (!sret.error) {
>
> Can it be standard pattern?
>
>         if (error)
>                 goto err_put_cpu;
>
> Ditto for other similar cases.

Okay, I will update.

>
> > +             remaining = le32_to_cpu(sdata->remaining);
> > +             returned = le32_to_cpu(sdata->returned);
> > +             *channel_count = remaining + returned;
> > +     }
> > +
> > +     put_cpu();
> > +     return sbi_err_map_linux_errno(sret.error);
> > +}
>
> ...
>
> > +static int mpxy_get_channel_ids(u32 channel_count, u32 *channel_ids)
> > +{
> > +     u32 remaining, returned, sidx, start_index = 0, cidx = 0;
> > +     struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
> > +     struct sbi_mpxy_channel_ids_data *sdata = mpxy->shmem;
> > +     struct sbiret sret;
> > +
> > +     if (!mpxy->shmem_active)
> > +             return -ENODEV;
> > +     if (!channel_count || !channel_ids)
> > +             return -EINVAL;
> > +
> > +     get_cpu();
> > +
> > +     do {
> > +             sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_GET_CHANNEL_IDS,
> > +                              start_index, 0, 0, 0, 0, 0);
> > +             if (sret.error)
> > +                     goto done;
> > +
> > +             remaining = le32_to_cpu(sdata->remaining);
> > +             returned = le32_to_cpu(sdata->returned);
> > +
> > +             for (sidx = 0; sidx < returned && cidx < channel_count; sidx++) {
> > +                     channel_ids[cidx] = le32_to_cpu(sdata->channel_array[sidx]);
> > +                     cidx += 1;
> > +             }
> > +
> > +             start_index = cidx;
> > +
> > +     } while (remaining);
> > +
> > +done:
>
> It sounds to me like an 'err_put_cpu' for the name.

Okay, I will update.

>
> > +     put_cpu();
> > +     return sbi_err_map_linux_errno(sret.error);
> > +}
>
> > +static int mpxy_write_attrs(u32 channel_id, u32 base_attrid, u32 attr_count,
> > +                         u32 *attrs_buf)
> > +{
> > +     struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
> > +     struct sbiret sret;
> > +     u32 i;
> > +
> > +     if (!mpxy->shmem_active)
> > +             return -ENODEV;
> > +     if (!attr_count || !attrs_buf)
> > +             return -EINVAL;
> > +
> > +     get_cpu();
> > +
> > +     for (i = 0; i < attr_count; i++)
> > +             ((__le32 *)mpxy->shmem)[i] = cpu_to_le32(attrs_buf[i]);
>
> Don't we have helpers for this? They are suffixed with _array.
> https://elixir.bootlin.com/linux/v6.15-rc6/source/include/linux/byteorder/generic.h#L168
> Don't forget to have asm/byteorder.h being included.
>
> Ditto for the similar case(s).

The cpu_to_le32_array() and le32_to_cpu_array() helpers update data
in-place but over here we have separate source and destination.

>
> > +     sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_WRITE_ATTRS,
> > +                      channel_id, base_attrid, attr_count, 0, 0, 0);
> > +
> > +     put_cpu();
> > +     return sbi_err_map_linux_errno(sret.error);
> > +}
>
> ...
>
> > +     msg->error = rc;
> > +     return 0;
>
> What's the point of having int and not void function?

Okay, I will update.

>
> ...
>
> > +                            sizeof(mchan->rpmi_attrs) / sizeof(u32),
> > +                            (u32 *)&mchan->rpmi_attrs);
>
> Why casting? What about alignment?

The RPMI attributes (aka struct sbi_mpxy_rpmi_channel_attrs) are
a collection of u32 attributes hence we can also treat rpmi_attrs
as a u32 array. Further, the rpmi_attrs is XLEN aligned within the
struct mpxy_mbox_channel so no alignment issue with the casting
on both RV32 and RV64.

If we want to avoid the casting then we will have to use a temporary
u32 array plus additional memcpy().

>
> ...
>
> > +     while (1) {
>
> Do it as do {} while, it will give an idea that the loop will run at least once
> (without looking into condition), ideally it shouldn't be infinite loop, but it
> might be harder to realise.

Okay, I will update.

>
> > +             rc = mpxy_get_notifications(mchan->channel_id, notif, &data_len);
> > +             if (rc || !data_len)
> > +                     break;
> > +
> > +             if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID)
> > +                     mpxy_mbox_peek_rpmi_data(chan, mchan, notif, data_len);
> > +
> > +             have_notifications = true;
> > +     }
>
> ...
>
> > +static void mpxy_mbox_cleanup_events(struct mpxy_mbox_channel *mchan)
> > +{
> > +     struct device *dev = mchan->mbox->dev;
> > +     int rc;
> > +
> > +     /* Do nothing if events state not supported */
> > +     if (!mchan->have_events_state)
> > +             return;
> > +
> > +     /* Do nothing if events state already disabled */
> > +     if (!mchan->attrs.events_state_ctrl)
> > +             return;
> > +
> > +     /* Disable channel events state */
> > +     mchan->attrs.events_state_ctrl = 0;
> > +     rc = mpxy_write_attrs(mchan->channel_id, SBI_MPXY_ATTR_EVENTS_STATE_CONTROL,
> > +                           1, &mchan->attrs.events_state_ctrl);
>
> > +     if (rc) {
> > +             dev_err(dev, "disable events state failed for MPXY channel 0x%x\n",
> > +                     mchan->channel_id);
> > +     }
>
> Redundant {}.

Okay, I will update.

>
> > +}
>
> ...
>
> > +static int mpxy_mbox_probe(struct platform_device *pdev)
> > +{
> > +     struct device *dev = &pdev->dev;
> > +     struct mpxy_mbox_channel *mchan;
> > +     struct mpxy_mbox *mbox;
> > +     int i, msi_idx, rc;
>
> Why is 'i' signed? Also make sure either they use POD (like unsigned int) or
> exact data type the upper limit does or API where you pass it in.

'i' is used to iterate over channels where channel_count and channel ids
are u32 so better to use the u32 data type.

>
> > +     u32 *channel_ids;
> > +
> > +     /*
> > +      * Initialize MPXY shared memory only once. This also ensures
> > +      * that SBI MPXY mailbox is probed only once.
> > +      */
> > +     if (mpxy_shmem_init_done) {
> > +             dev_err(dev, "SBI MPXY mailbox already initialized\n");
> > +             return -EALREADY;
> > +     }
> > +
> > +     /* Probe for SBI MPXY extension */
> > +     if (sbi_spec_version < sbi_mk_version(1, 0) ||
> > +         sbi_probe_extension(SBI_EXT_MPXY) <= 0) {
> > +             dev_info(dev, "SBI MPXY extension not available\n");
> > +             return -ENODEV;
> > +     }
> > +
> > +     /* Find-out shared memory size */
> > +     mpxy_shmem_size = mpxy_get_shmem_size();
> > +
> > +     /*
> > +      * Setup MPXY shared memory on each CPU
> > +      *
> > +      * Note: Don't cleanup MPXY shared memory upon CPU power-down
> > +      * because the RPMI System MSI irqchip driver needs it to be
> > +      * available when migrating IRQs in CPU power-down path.
> > +      */
> > +     cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "riscv/sbi-mpxy-shmem",
> > +                       mpxy_setup_shmem, NULL);
> > +
> > +     /* Mark as MPXY shared memory initialization done */
> > +     mpxy_shmem_init_done = true;
> > +
> > +     /* Allocate mailbox instance */
> > +     mbox = devm_kzalloc(dev, sizeof(*mbox), GFP_KERNEL);
> > +     if (!mbox)
> > +             return -ENOMEM;
> > +     mbox->dev = dev;
> > +     platform_set_drvdata(pdev, mbox);
> > +
> > +     /* Find-out of number of channels */
> > +     rc = mpxy_get_channel_count(&mbox->channel_count);
> > +     if (rc)
> > +             return dev_err_probe(dev, rc, "failed to get number of MPXY channels\n");
> > +     if (!mbox->channel_count)
> > +             dev_err_probe(dev, -ENODEV, "no MPXY channels available\n");
> > +
> > +     /* Allocate and fetch all channel IDs */
> > +     channel_ids = devm_kcalloc(dev, mbox->channel_count,
> > +                                sizeof(*channel_ids), GFP_KERNEL);
> > +     if (!channel_ids)
> > +             return -ENOMEM;
> > +     rc = mpxy_get_channel_ids(mbox->channel_count, channel_ids);
> > +     if (rc)
> > +             return dev_err_probe(dev, rc, "failed to MPXY channel IDs\n");
> > +
> > +     /* Populate all channels */
> > +     mbox->channels = devm_kcalloc(dev, mbox->channel_count,
> > +                                   sizeof(*mbox->channels), GFP_KERNEL);
> > +     if (!mbox->channels)
> > +             return -ENOMEM;
> > +     for (i = 0; i < mbox->channel_count; i++) {
> > +             mchan = &mbox->channels[i];
> > +             mchan->mbox = mbox;
> > +             mchan->channel_id = channel_ids[i];
> > +
> > +             rc = mpxy_read_attrs(mchan->channel_id, SBI_MPXY_ATTR_MSG_PROT_ID,
> > +                                  sizeof(mchan->attrs) / sizeof(u32),
> > +                                  (u32 *)&mchan->attrs);
> > +             if (rc) {
> > +                     return dev_err_probe(dev, rc,
> > +                                          "MPXY channel 0x%x read attrs failed\n",
> > +                                          mchan->channel_id);
> > +             }
> > +
> > +             if (mchan->attrs.msg_proto_id == SBI_MPXY_MSGPROTO_RPMI_ID) {
> > +                     rc = mpxy_mbox_read_rpmi_attrs(mchan);
> > +                     if (rc) {
> > +                             return dev_err_probe(dev, rc,
> > +                                                  "MPXY channel 0x%x read RPMI attrs failed\n",
> > +                                                  mchan->channel_id);
> > +                     }
> > +             }
> > +
> > +             mchan->notif = devm_kzalloc(dev, mpxy_shmem_size, GFP_KERNEL);
> > +             if (!mchan->notif)
> > +                     return -ENOMEM;
> > +
> > +             mchan->max_xfer_len = min(mpxy_shmem_size, mchan->attrs.msg_max_len);
> > +
> > +             if ((mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS) &&
> > +                 (mchan->attrs.capability & SBI_MPXY_CHAN_CAP_EVENTS_STATE))
> > +                     mchan->have_events_state = true;
> > +
> > +             if ((mchan->attrs.capability & SBI_MPXY_CHAN_CAP_GET_NOTIFICATIONS) &&
> > +                 (mchan->attrs.capability & SBI_MPXY_CHAN_CAP_MSI))
> > +                     mchan->msi_index = mbox->msi_count++;
> > +             else
> > +                     mchan->msi_index = U32_MAX;
> > +             mchan->msi_irq = U32_MAX;
> > +     }
>
> > +     /* Free-up channel IDs */
> > +     devm_kfree(dev, channel_ids);
>
> Just no. This is most likely (99.9%) wrong use of managed resources. This is
> the biggest issue with this patch so far.

Okay, I will use __free(kfree) for channel_ids[].

>
> > +     /* Initialize mailbox controller */
> > +     mbox->controller.txdone_irq = false;
> > +     mbox->controller.txdone_poll = false;
> > +     mbox->controller.ops = &mpxy_mbox_ops;
> > +     mbox->controller.dev = dev;
> > +     mbox->controller.num_chans = mbox->channel_count;
> > +     mbox->controller.fw_xlate = mpxy_mbox_fw_xlate;
> > +     mbox->controller.chans = devm_kcalloc(dev, mbox->channel_count,
> > +                                           sizeof(*mbox->controller.chans),
> > +                                           GFP_KERNEL);
> > +     if (!mbox->controller.chans)
> > +             return -ENOMEM;
> > +     for (i = 0; i < mbox->channel_count; i++)
> > +             mbox->controller.chans[i].con_priv = &mbox->channels[i];
> > +
> > +     /* Set the MSI domain if not available */
> > +     if (!dev_get_msi_domain(dev)) {
> > +             /*
> > +              * The device MSI domain for OF devices is only set at the
> > +              * time of populating/creating OF device. If the device MSI
> > +              * domain is discovered later after the OF device is created
> > +              * then we need to set it explicitly before using any platform
> > +              * MSI functions.
> > +              */
> > +             if (is_of_node(dev_fwnode(dev)))
> > +                     of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
> > +     }
> > +
> > +     /* Setup MSIs for mailbox (if required) */
> > +     if (mbox->msi_count) {
> > +             mbox->msi_index_to_channel = devm_kcalloc(dev, mbox->msi_count,
> > +                                                       sizeof(*mbox->msi_index_to_channel),
> > +                                                       GFP_KERNEL);
> > +             if (!mbox->msi_index_to_channel)
> > +                     return -ENOMEM;
> > +
> > +             for (msi_idx = 0; msi_idx < mbox->msi_count; msi_idx++) {
> > +                     for (i = 0; i < mbox->channel_count; i++) {
> > +                             mchan = &mbox->channels[i];
> > +                             if (mchan->msi_index == msi_idx) {
> > +                                     mbox->msi_index_to_channel[msi_idx] = mchan;
> > +                                     break;
> > +                             }
> > +                     }
> > +             }
> > +
> > +             rc = platform_device_msi_init_and_alloc_irqs(dev, mbox->msi_count,
> > +                                                          mpxy_mbox_msi_write);
> > +             if (rc) {
> > +                     return dev_err_probe(dev, rc, "Failed to allocate %d MSIs\n",
> > +                                          mbox->msi_count);
> > +             }
> > +
> > +             for (i = 0; i < mbox->channel_count; i++) {
> > +                     mchan = &mbox->channels[i];
> > +                     if (mchan->msi_index == U32_MAX)
> > +                             continue;
> > +                     mchan->msi_irq = msi_get_virq(dev, mchan->msi_index);
> > +             }
> > +     }
> > +
> > +     /* Register mailbox controller */
> > +     rc = devm_mbox_controller_register(dev, &mbox->controller);
> > +     if (rc) {
> > +             dev_err_probe(dev, rc, "Registering SBI MPXY mailbox failed\n");
> > +             if (mbox->msi_count)
> > +                     platform_device_msi_free_irqs_all(dev);
> > +             return rc;
> > +     }
> > +
> > +     dev_info(dev, "mailbox registered with %d channels\n",
> > +              mbox->channel_count);
> > +     return 0;
> > +}
>
> ...
>
> > +     if (mbox->msi_count)
>
> Is this check really needed?

MSIs are optional for the SBI MPXY mailbox so we should only use
platform_device_msi_xyz() APIs only when MSIs are available.

>
> > +             platform_device_msi_free_irqs_all(mbox->dev);
> > +}
>
> ...
>
> > +static const struct of_device_id mpxy_mbox_of_match[] = {
> > +     {.compatible = "riscv,sbi-mpxy-mbox", },
> > +     {},
>
> No comma for the terminator entry.
> You should apply the reviewers' comments on all of your code, not just where it
> was commented.

Okay, I will update.

Regards,
Anup

Re: [PATCH v4 06/23] mailbox: Allow controller specific mapping using fwnode

[Andy Shevchenko]

On Mon, Jun 09, 2025 at 02:40:32PM +0530, Anup Patel wrote:
> On Tue, May 27, 2025 at 5:11 PM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Sun, May 25, 2025 at 02:16:53PM +0530, Anup Patel wrote:

...

> > >  struct mbox_chan *mbox_request_channel(struct mbox_client *cl, int index)
> > >  {
> > > +     struct fwnode_reference_args fwspec;
> >
> > + property.h (if not done yet)
> 
> This needs fwnode.h

No, it needs property.h. Please, read the note at the top of fwnode.h why.

...

> > > -     int ret;
> > > +     int i, ret;
> >
> > Why is 'i' signed?
> 
> No particular reason because 'i' is not being assigned a negative
>>  value so both signed and unsigned will work fine.

Besides, splitting will make it easier to check and work on as being less
intrusive change.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Andy Shevchenko]

On Mon, Jun 09, 2025 at 05:59:40PM +0530, Anup Patel wrote:
> On Wed, May 28, 2025 at 4:23 PM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:

...

> > > +#include <asm/sbi.h>
> >
> > asm/* usually goes after generic linux/* ones. Why here?
> 
> I am not aware of any such convention but I will update anyway.

It's just a common sense. We include most generic first and most custom at
last.

...

> > > +static int mpxy_write_attrs(u32 channel_id, u32 base_attrid, u32 attr_count,
> > > +                         u32 *attrs_buf)
> > > +{
> > > +     struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
> > > +     struct sbiret sret;
> > > +     u32 i;
> > > +
> > > +     if (!mpxy->shmem_active)
> > > +             return -ENODEV;
> > > +     if (!attr_count || !attrs_buf)
> > > +             return -EINVAL;
> > > +
> > > +     get_cpu();
> > > +
> > > +     for (i = 0; i < attr_count; i++)
> > > +             ((__le32 *)mpxy->shmem)[i] = cpu_to_le32(attrs_buf[i]);
> >
> > Don't we have helpers for this? They are suffixed with _array.
> > https://elixir.bootlin.com/linux/v6.15-rc6/source/include/linux/byteorder/generic.h#L168
> > Don't forget to have asm/byteorder.h being included.
> >
> > Ditto for the similar case(s).
> 
> The cpu_to_le32_array() and le32_to_cpu_array() helpers update data
> in-place but over here we have separate source and destination.

Fair enough. Perhaps add something like memcpy_to_le32() / memcpy_from_le32()
or alike for your case?

> > > +     sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_WRITE_ATTRS,
> > > +                      channel_id, base_attrid, attr_count, 0, 0, 0);
> > > +
> > > +     put_cpu();
> > > +     return sbi_err_map_linux_errno(sret.error);
> > > +}

...

> > > +                            sizeof(mchan->rpmi_attrs) / sizeof(u32),
> > > +                            (u32 *)&mchan->rpmi_attrs);
> >
> > Why casting? What about alignment?
> 
> The RPMI attributes (aka struct sbi_mpxy_rpmi_channel_attrs) are
> a collection of u32 attributes hence we can also treat rpmi_attrs
> as a u32 array. Further, the rpmi_attrs is XLEN aligned within the
> struct mpxy_mbox_channel so no alignment issue with the casting
> on both RV32 and RV64.
> 
> If we want to avoid the casting then we will have to use a temporary
> u32 array plus additional memcpy().

OK.

...

> > > +     if (mbox->msi_count)
> >
> > Is this check really needed?
> 
> MSIs are optional for the SBI MPXY mailbox so we should only use
> platform_device_msi_xyz() APIs only when MSIs are available.

> > > +             platform_device_msi_free_irqs_all(mbox->dev);

Hmm... I am not sure why. Do you have any Oops or warnings if the check
is not there and no MSI provided?

> > > +}

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Anup Patel]

On Tue, Jun 10, 2025 at 1:34 AM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Mon, Jun 09, 2025 at 05:59:40PM +0530, Anup Patel wrote:
> > On Wed, May 28, 2025 at 4:23 PM Andy Shevchenko
> > <andriy.shevchenko@linux.intel.com> wrote:
> > > On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:
>
> ...
>
> > > > +#include <asm/sbi.h>
> > >
> > > asm/* usually goes after generic linux/* ones. Why here?
> >
> > I am not aware of any such convention but I will update anyway.
>
> It's just a common sense. We include most generic first and most custom at
> last.
>
> ...
>
> > > > +static int mpxy_write_attrs(u32 channel_id, u32 base_attrid, u32 attr_count,
> > > > +                         u32 *attrs_buf)
> > > > +{
> > > > +     struct mpxy_local *mpxy = this_cpu_ptr(&mpxy_local);
> > > > +     struct sbiret sret;
> > > > +     u32 i;
> > > > +
> > > > +     if (!mpxy->shmem_active)
> > > > +             return -ENODEV;
> > > > +     if (!attr_count || !attrs_buf)
> > > > +             return -EINVAL;
> > > > +
> > > > +     get_cpu();
> > > > +
> > > > +     for (i = 0; i < attr_count; i++)
> > > > +             ((__le32 *)mpxy->shmem)[i] = cpu_to_le32(attrs_buf[i]);
> > >
> > > Don't we have helpers for this? They are suffixed with _array.
> > > https://elixir.bootlin.com/linux/v6.15-rc6/source/include/linux/byteorder/generic.h#L168
> > > Don't forget to have asm/byteorder.h being included.
> > >
> > > Ditto for the similar case(s).
> >
> > The cpu_to_le32_array() and le32_to_cpu_array() helpers update data
> > in-place but over here we have separate source and destination.
>
> Fair enough. Perhaps add something like memcpy_to_le32() / memcpy_from_le32()
> or alike for your case?

Okay, I will add memcpy_to_le32() / memcpy_from_le32()
in include/linux/byteorder/generic.h and use it over here.

>
> > > > +     sret = sbi_ecall(SBI_EXT_MPXY, SBI_EXT_MPXY_WRITE_ATTRS,
> > > > +                      channel_id, base_attrid, attr_count, 0, 0, 0);
> > > > +
> > > > +     put_cpu();
> > > > +     return sbi_err_map_linux_errno(sret.error);
> > > > +}
>
> ...
>
> > > > +                            sizeof(mchan->rpmi_attrs) / sizeof(u32),
> > > > +                            (u32 *)&mchan->rpmi_attrs);
> > >
> > > Why casting? What about alignment?
> >
> > The RPMI attributes (aka struct sbi_mpxy_rpmi_channel_attrs) are
> > a collection of u32 attributes hence we can also treat rpmi_attrs
> > as a u32 array. Further, the rpmi_attrs is XLEN aligned within the
> > struct mpxy_mbox_channel so no alignment issue with the casting
> > on both RV32 and RV64.
> >
> > If we want to avoid the casting then we will have to use a temporary
> > u32 array plus additional memcpy().
>
> OK.
>
> ...
>
> > > > +     if (mbox->msi_count)
> > >
> > > Is this check really needed?
> >
> > MSIs are optional for the SBI MPXY mailbox so we should only use
> > platform_device_msi_xyz() APIs only when MSIs are available.
>
> > > > +             platform_device_msi_free_irqs_all(mbox->dev);
>
> Hmm... I am not sure why. Do you have any Oops or warnings if the check
> is not there and no MSI provided?

We don't see any oops or warnings. This check is to avoid unnecessary
work (such as acquiring lock, checking default domain, etc) in the
msi_domain_free_irqs_all() called by platform_device_msi_free_irqs_all().

I don't mind dropping the check so I will update in the next revision.

Regards,
Anup

Re: [PATCH v4 17/23] ACPI: RISC-V: Create interrupt controller list in sorted order

[Sunil V L]

On Wed, May 28, 2025 at 02:05:36PM +0300, Andy Shevchenko wrote:
> On Sun, May 25, 2025 at 02:17:04PM +0530, Anup Patel wrote:
> > 
> > Currently, the interrupt controller list is created without any order.
> > Create the list sorted with the GSI base of the interrupt controllers.
> 
> ...
> 
> > -	list_add_tail(&ext_intc_element->list, &ext_intc_list);
> > +	if (list_empty(&ext_intc_list)) {
> > +		list_add(&ext_intc_element->list, &ext_intc_list);
> > +		return 0;
> > +	}
> 
> With the below done the above can be optimized (hopefully).
> 
> > +	list_for_each_entry(node, &ext_intc_list, list) {
> > +		if (node->gsi_base < ext_intc_element->gsi_base)
> > +			break;
> > +	}
> > +
> > +	__list_add(&ext_intc_element->list, node->list.prev, &node->list);
> 
> Is this reimplementation of list_add_tail()? And why list debug is excluded here?
> 
Sure. Let me use list_add_tail() itself in the next revision. However, I
didn't understand the list debug question. IIUC, list_add_tail() is a
wrapper around __list_add() and doesn't exclude the list debug, right?

Thanks!
Sunil

Re: [PATCH v4 08/23] dt-bindings: clock: Add RPMI clock service message proxy bindings

[Anup Patel]

On Fri, May 30, 2025 at 9:58 PM Conor Dooley <conor@kernel.org> wrote:
>
> On Sun, May 25, 2025 at 02:16:55PM +0530, Anup Patel wrote:
> > +  riscv,sbi-mpxy-channel-id:
> > +    $ref: /schemas/types.yaml#/definitions/uint32
> > +    description:
> > +      The SBI MPXY channel id to be used for providing RPMI access to
> > +      the supervisor software.
>
> Sorry for the delay in reviewing this series, wanted to talk to some
> folks at work before doing so. Overall these bindings all look pretty
> good to me. I don't think the description on this property here is
> particularly great, I didn't follow what it was meant to be used for
> immediately. But I don't have anything meaningfully better to
> suggest nor do I think that anyone actually writing the dts for a
> platform using this binding will struggle to understand it.

Okay, I will keep this as-is until we have a better suggestion.

Regards,
Anup

Re: [PATCH v4 09/23] dt-bindings: clock: Add RPMI clock service controller bindings

[Anup Patel]

On Fri, May 30, 2025 at 10:11 PM Conor Dooley <conor@kernel.org> wrote:
>
> On Sun, May 25, 2025 at 02:16:56PM +0530, Anup Patel wrote:
> > Add device tree bindings for the RPMI clock service group based
> > controller for the supervisor software.
> >
> > The RPMI clock service group is defined by the RISC-V platform
> > management interface (RPMI) specification.
> >
> > Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> > ---
> >  .../bindings/clock/riscv,rpmi-clock.yaml      | 61 +++++++++++++++++++
> >  1 file changed, 61 insertions(+)
> >  create mode 100644 Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml b/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
> > new file mode 100644
> > index 000000000000..9c672a38595a
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/clock/riscv,rpmi-clock.yaml
> > @@ -0,0 +1,61 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/clock/riscv,rpmi-clock.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: RISC-V RPMI clock service group based clock controller
> > +
> > +maintainers:
> > +  - Anup Patel <anup@brainfault.org>
> > +
> > +description: |
> > +  The RISC-V Platform Management Interface (RPMI) [1] defines a
> > +  messaging protocol which is modular and extensible. The supervisor
> > +  software can send/receive RPMI messages via SBI MPXY extension [2]
> > +  or some dedicated supervisor-mode RPMI transport.
> > +
> > +  The RPMI specification [1] defines clock service group for accessing
> > +  system clocks managed by a platform microcontroller. The supervisor
> > +  software can access RPMI clock service group via SBI MPXY channel or
> > +  some dedicated supervisor-mode RPMI transport.
> > +
> > +  ===========================================
> > +  References
> > +  ===========================================
> > +
> > +  [1] RISC-V Platform Management Interface (RPMI)
> > +      https://github.com/riscv-non-isa/riscv-rpmi/releases
> > +
> > +  [2] RISC-V Supervisor Binary Interface (SBI)
> > +      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
> > +
> > +properties:
> > +  compatible:
> > +    description:
> > +      Intended for use by the supervisor software.
> > +    const: riscv,rpmi-clock
> > +
> > +  mboxes:
> > +    maxItems: 1
> > +    description:
> > +      Mailbox channel of the underlying RPMI transport or SBI message proxy channel.
> > +
> > +  "#clock-cells":
> > +    const: 1
>
> Could you please add some description here as to what this clock-cell
> actually does? On a normal clock controller someone might cite an
> include file with a huge list of defines for what numbers map to what
> clock. In this case, this value is CLOCK_ID in the spec, so it's
> completely platform specific as to what they mean so citing some include
> isn't helpful, so just mention that it is CLOCK_ID and the meanings are
> platform specific.
>
> I suppose technically it can be something other than CLOCK_ID, if this is
> used when the SBI message proxy extension is provided by an SBI
> implementation that uses a non-RPMI transport, but I don't think that's a
> can of worms worth bringing up in the binding. Anyone doing that can put
> 2+2 together I think.

The #clock-cell value must be the platform specific CLOCK_ID as defined by
the RISC-V Platform Management Interface (RPMI) specification irrespective
of the underlying mechanism (e.g. SBI MPXY or RPMI transport) used to
access the RPMI Clock service group.

Each platform must publish their CLOCK_ID values as part of their SoC docs
or as a header used by their SoC DTS files.

I will add "description:" for clock-cells along these lines.

Regards,
Anup

Re: [PATCH v4 11/23] dt-bindings: Add RPMI system MSI message proxy bindings

[Anup Patel]

On Sat, Jun 7, 2025 at 4:29 AM Atish Patra <atish.patra@linux.dev> wrote:
>
>
> On 5/25/25 1:46 AM, Anup Patel wrote:
> > Add device tree bindings for the RPMI system MSI service group
> > based message proxy implemented by the SBI implementation (machine
> > mode firmware or hypervisor).
> >
> > The RPMI system MSI service group is defined by the RISC-V
> > platform management interface (RPMI) specification.
> >
> > Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> > ---
> >   .../riscv,rpmi-mpxy-system-msi.yaml           | 67 +++++++++++++++++++
> >   1 file changed, 67 insertions(+)
> >   create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
> > new file mode 100644
> > index 000000000000..26dd13731350
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml
> > @@ -0,0 +1,67 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/interrupt-controller/riscv,rpmi-mpxy-system-msi.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: RISC-V RPMI system MSI service group based message proxy
> > +
> > +maintainers:
> > +  - Anup Patel <anup@brainfault.org>
> > +
> > +description: |
> > +  The RISC-V Platform Management Interface (RPMI) [1] defines a
> > +  messaging protocol which is modular and extensible. The supervisor
> > +  software can send/receive RPMI messages via SBI MPXY extension [2]
> > +  or some dedicated supervisor-mode RPMI transport.
> > +
> > +  The RPMI specification [1] defines system MSI service group which
> > +  allow application processors to receive MSIs upon system events
> > +  such as P2A doorbell, graceful shutdown/reboot request, CPU hotplug
> > +  event, memory hotplug event, etc from the platform microcontroller.
> > +  The SBI implementation machine mode firmware or hypervisor) can
> nit:
> The SBI implementation in machine mode

Actually there should be a "(" just before "machine mode ...". I will
update in the next revision.

> > +  implement an SBI MPXY channel to allow RPMI system MSI service
> > +  group access to the supervisor software.
> > +
> > +  ===========================================
> > +  References
> > +  ===========================================
> > +
> > +  [1] RISC-V Platform Management Interface (RPMI)
> > +      https://github.com/riscv-non-isa/riscv-rpmi/releases
> > +
> > +  [2] RISC-V Supervisor Binary Interface (SBI)
> > +      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
> > +
> nit: Refer the latest frozen version of the spec ?

Actually, we should specify the spec version number in the text
and avoid a specific link because the link of the final ratified version
will be different.

> > +properties:
> > +  compatible:
> > +    description:
> > +      Intended for use by the SBI implementation.
> > +    const: riscv,rpmi-mpxy-system-msi
> > +
> > +  mboxes:
> > +    maxItems: 1
> > +    description:
> > +      Mailbox channel of the underlying RPMI transport.
> > +
> > +  riscv,sbi-mpxy-channel-id:
> > +    $ref: /schemas/types.yaml#/definitions/uint32
> > +    description:
> > +      The SBI MPXY channel id to be used for providing RPMI access to
> > +      the supervisor software.
> > +
> > +required:
> > +  - compatible
> > +  - mboxes
> > +  - riscv,sbi-mpxy-channel-id
> > +
> > +additionalProperties: false
> > +
> > +examples:
> > +  - |
> > +    interrupt-controller {
> > +        compatible = "riscv,rpmi-mpxy-system-msi";
> > +        mboxes = <&rpmi_shmem_mbox 0x2>;
> > +        riscv,sbi-mpxy-channel-id = <0x2000>;
> > +    };
> > +...
>
> Otherwise, lgtm.
>
> Reviewed-by: Atish Patra <atishp@rivosinc.com>
>

Regards,
Anup

Re: [PATCH v4 12/23] dt-bindings: Add RPMI system MSI interrupt controller bindings

[Anup Patel]

On Sat, Jun 7, 2025 at 4:33 AM Atish Patra <atish.patra@linux.dev> wrote:
>
>
> On 5/25/25 1:46 AM, Anup Patel wrote:
> > Add device tree bindings for the RPMI system MSI service group
> > based interrupt controller for the supervisor software.
> >
> > The RPMI system MSI service group is defined by the RISC-V
> > platform management interface (RPMI) specification.
> >
> > Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> > ---
> >   .../riscv,rpmi-system-msi.yaml                | 74 +++++++++++++++++++
> >   1 file changed, 74 insertions(+)
> >   create mode 100644 Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
> > new file mode 100644
> > index 000000000000..ac13cec0666e
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/interrupt-controller/riscv,rpmi-system-msi.yaml
> > @@ -0,0 +1,74 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/interrupt-controller/riscv,rpmi-system-msi.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: RISC-V RPMI system MSI service group based interrupt controller
> > +
> > +maintainers:
> > +  - Anup Patel <anup@brainfault.org>
> > +
> > +description: |
> > +  The RISC-V Platform Management Interface (RPMI) [1] defines a
> > +  messaging protocol which is modular and extensible. The supervisor
> > +  software can send/receive RPMI messages via SBI MPXY extension [2]
> > +  or some dedicated supervisor-mode RPMI transport.
> > +
> > +  The RPMI specification [1] defines system MSI service group which
> > +  allow application processors to receive MSIs upon system events
> > +  such as P2A doorbell, graceful shutdown/reboot request, CPU hotplug
> > +  event, memory hotplug event, etc from the platform microcontroller.
> > +  The supervisor software can access RPMI system MSI service group via
> > +  SBI MPXY channel or some dedicated supervisor-mode RPMI transport.
> > +
> > +  ===========================================
> > +  References
> > +  ===========================================
> > +
> > +  [1] RISC-V Platform Management Interface (RPMI)
> > +      https://github.com/riscv-non-isa/riscv-rpmi/releases
> > +
> > +  [2] RISC-V Supervisor Binary Interface (SBI)
> > +      https://github.com/riscv-non-isa/riscv-sbi-doc/releases
> > +
> nit: Same comment as previous patch.

Like mentioned in the previous patch, I will add the spec version to the text.

> > +allOf:
> > +  - $ref: /schemas/interrupt-controller.yaml#
> > +
> > +properties:
> > +  compatible:
> > +    description:
> > +      Intended for use by the supervisor software.
> > +    const: riscv,rpmi-system-msi
> > +
> > +  mboxes:
> > +    maxItems: 1
> > +    description:
> > +      Mailbox channel of the underlying RPMI transport or SBI message proxy channel.
> > +
> > +  msi-parent: true
> > +
> > +  interrupt-controller: true
> > +
> > +  "#interrupt-cells":
> > +    const: 1
> > +
> > +required:
> > +  - compatible
> > +  - mboxes
> > +  - msi-parent
> > +  - interrupt-controller
> > +  - "#interrupt-cells"
> > +
> > +additionalProperties: false
> > +
> > +examples:
> > +  - |
> > +    interrupt-controller {
> > +        compatible = "riscv,rpmi-system-msi";
> > +        mboxes = <&mpxy_mbox 0x2000 0x0>;
> > +        msi-parent = <&imsic_slevel>;
> > +        interrupt-controller;
> > +        #interrupt-cells = <1>;
> > +    };
> > +...
>
> Otherwise, LGTM.
> Reviewed-by: Atish Patra <atishp@rivosinc.com>
>
>
>

Regards,
Anup

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Andy Shevchenko]

On Tue, Jun 10, 2025 at 10:05:27AM +0530, Anup Patel wrote:
> On Tue, Jun 10, 2025 at 1:34 AM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Mon, Jun 09, 2025 at 05:59:40PM +0530, Anup Patel wrote:
> > > On Wed, May 28, 2025 at 4:23 PM Andy Shevchenko
> > > <andriy.shevchenko@linux.intel.com> wrote:
> > > > On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:

...

> > > > > +     if (mbox->msi_count)
> > > >
> > > > Is this check really needed?
> > >
> > > MSIs are optional for the SBI MPXY mailbox so we should only use
> > > platform_device_msi_xyz() APIs only when MSIs are available.
> >
> > > > > +             platform_device_msi_free_irqs_all(mbox->dev);
> >
> > Hmm... I am not sure why. Do you have any Oops or warnings if the check
> > is not there and no MSI provided?
> 
> We don't see any oops or warnings. This check is to avoid unnecessary
> work (such as acquiring lock, checking default domain, etc) in the
> msi_domain_free_irqs_all() called by platform_device_msi_free_irqs_all().
> 
> I don't mind dropping the check so I will update in the next revision.

Perhaps you can rather add this check into the callee? Seems to me that
you have a justification for it. Usual pattern in the kernel that freeing
resources should be aware of the NULL pointers or optional resources
so we may call it unconditionally from the user(s).

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 13/23] irqchip: Add driver for the RPMI system MSI service group

[Anup Patel]

On Tue, May 27, 2025 at 5:03 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Sun, May 25, 2025 at 02:17:00PM +0530, Anup Patel wrote:
> > The RPMI specification defines a system MSI service group which
> > allows application processors to receive MSIs upon system events
> > such as graceful shutdown/reboot request, CPU hotplug event, memory
> > hotplug event, etc.
> >
> > Add an irqchip driver for the RISC-V RPMI system MSI service group
> > to directly receive system MSIs in Linux kernel.
>
> ...
>
> > +#include <linux/device.h>
>
> Perhaps I missed something, but devm_kzalloc() is in device/devres.h. Do you
> need it for something else?

dev_get_msi_domain() is part of linux/device.h hence it should
be included.

I will add linux/device/devres.h as well.

>
> > +#include <linux/dev_printk.h>
> > +#include <linux/irq.h>
> > +#include <linux/irqdomain.h>
> > +#include <linux/mailbox_client.h>
> > +#include <linux/mailbox/riscv-rpmi-message.h>
> > +#include <linux/module.h>
> > +#include <linux/msi.h>
> > +#include <linux/of_irq.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/types.h>
>
> > +#include <vdso/bits.h>
>
> Just make it linux/bits.h as vdso is for user space libvdso and related.
>
> But
>
> + asm/byteorder.h
>
> ...
>
> > +static void rpmi_sysmsi_irq_mask(struct irq_data *d)
> > +{
> > +     struct rpmi_sysmsi_priv *priv = irq_data_get_irq_chip_data(d);
>
> Declare temporary variable for hwirq and do irqd_to_hwirq() only once.

Okay, I will update.

>
> > +     int ret;
> > +
> > +     ret = rpmi_sysmsi_set_msi_state(priv, irqd_to_hwirq(d), 0);
> > +     if (ret) {
> > +             dev_warn(priv->dev, "Failed to mask hwirq %lu (error %d)\n",
> > +                      irqd_to_hwirq(d), ret);
> > +     }
> > +     irq_chip_mask_parent(d);
> > +}
>
> ...
>
> > +static void rpmi_sysmsi_irq_unmask(struct irq_data *d)
>
> Ditto.
>
> ...
>
> > +static void rpmi_sysmsi_set_desc(msi_alloc_info_t *arg, struct msi_desc *desc)
> > +{
> > +     arg->desc = desc;
> > +     arg->hwirq = (u32)desc->data.icookie.value;
>
> Hmm... Why do you need an explicit casting?

Not needed, I will drop the casting in the next revision.

>
> > +}
>
> ...
>
> > +     if (WARN_ON(fwspec->param_count < 1))
>
> + bug.h

Okay, I will update.

>
> > +             return -EINVAL;
>
> + errno.h (but actually you need err.h due to PTR_ERR() et al.)
>
> ...
>
> > +static int rpmi_sysmsi_probe(struct platform_device *pdev)
> > +{
> > +     struct device *dev = &pdev->dev;
> > +     struct rpmi_sysmsi_priv *priv;
> > +     int rc;
>
> Be consistent with variable naming for the same (semantically) stuff.
>
> > +     priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> > +     if (!priv)
> > +             return -ENOMEM;
> > +     priv->dev = dev;
>
> > +     platform_set_drvdata(pdev, priv);
>
> How is being used?

Typically, it's for driver_remove() but since we don't have
driver_remove() we can drop it.

>
> > +
> > +     /* Setup mailbox client */
> > +     priv->client.dev                = priv->dev;
> > +     priv->client.rx_callback        = NULL;
> > +     priv->client.tx_block           = false;
> > +     priv->client.knows_txdone       = true;
> > +     priv->client.tx_tout            = 0;
> > +
> > +     /* Request mailbox channel */
> > +     priv->chan = mbox_request_channel(&priv->client, 0);
> > +     if (IS_ERR(priv->chan))
> > +             return PTR_ERR(priv->chan);
> > +
> > +     /* Get number of system MSIs */
> > +     rc = rpmi_sysmsi_get_num_msi(priv);
> > +     if (rc < 1) {
> > +             mbox_free_channel(priv->chan);
> > +             if (rc)
> > +                     return dev_err_probe(dev, rc, "Failed to get number of system MSIs\n");
> > +             else
> > +                     return dev_err_probe(dev, -ENODEV, "No system MSIs found\n");
> > +     }
> > +     priv->nr_irqs = rc;
> > +
> > +     /* Set the device MSI domain if not available */
> > +     if (!dev_get_msi_domain(dev)) {
> > +             /*
> > +              * The device MSI domain for OF devices is only set at the
> > +              * time of populating/creating OF device. If the device MSI
> > +              * domain is discovered later after the OF device is created
> > +              * then we need to set it explicitly before using any platform
> > +              * MSI functions.
> > +              */
>
> > +             if (is_of_node(dev_fwnode(dev)))
> > +                     of_msi_configure(dev, to_of_node(dev_fwnode(dev)));
>
>                 if (dev_of_node(dev))
>                         of_msi_configure(dev, dev_of_node(dev));

Okay, I will update.

>
> > +             if (!dev_get_msi_domain(dev)) {
> > +                     mbox_free_channel(priv->chan);
> > +                     return -EPROBE_DEFER;
> > +             }
> > +     }
> > +
> > +     if (!msi_create_device_irq_domain(dev, MSI_DEFAULT_DOMAIN,
> > +                                       &rpmi_sysmsi_template,
> > +                                       priv->nr_irqs, priv, priv)) {
> > +             mbox_free_channel(priv->chan);
> > +             return dev_err_probe(dev, -ENOMEM, "failed to create MSI irq domain\n");
> > +     }
> > +
> > +     dev_info(dev, "%u system MSIs registered\n", priv->nr_irqs);
> > +     return 0;
> > +}
>
> --
> With Best Regards,
> Andy Shevchenko
>
>

Regards,
Anup

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Anup Patel]

On Tue, Jun 10, 2025 at 3:25 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Tue, Jun 10, 2025 at 10:05:27AM +0530, Anup Patel wrote:
> > On Tue, Jun 10, 2025 at 1:34 AM Andy Shevchenko
> > <andriy.shevchenko@linux.intel.com> wrote:
> > > On Mon, Jun 09, 2025 at 05:59:40PM +0530, Anup Patel wrote:
> > > > On Wed, May 28, 2025 at 4:23 PM Andy Shevchenko
> > > > <andriy.shevchenko@linux.intel.com> wrote:
> > > > > On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:
>
> ...
>
> > > > > > +     if (mbox->msi_count)
> > > > >
> > > > > Is this check really needed?
> > > >
> > > > MSIs are optional for the SBI MPXY mailbox so we should only use
> > > > platform_device_msi_xyz() APIs only when MSIs are available.
> > >
> > > > > > +             platform_device_msi_free_irqs_all(mbox->dev);
> > >
> > > Hmm... I am not sure why. Do you have any Oops or warnings if the check
> > > is not there and no MSI provided?
> >
> > We don't see any oops or warnings. This check is to avoid unnecessary
> > work (such as acquiring lock, checking default domain, etc) in the
> > msi_domain_free_irqs_all() called by platform_device_msi_free_irqs_all().
> >
> > I don't mind dropping the check so I will update in the next revision.
>
> Perhaps you can rather add this check into the callee? Seems to me that
> you have a justification for it. Usual pattern in the kernel that freeing
> resources should be aware of the NULL pointers or optional resources
> so we may call it unconditionally from the user(s).
>

Unconditionally calling platform_device_msi_free_irqs_all() when there
were no MSIs allocated causes the below crash because "dev->msi.data"
is non-NULL only when:

[    1.355735] Unable to handle kernel NULL pointer dereference at
virtual address 0000000000000008
[    1.358212] Current swapper/0 pgtable: 4K pagesize, 57-bit VAs,
pgdp=0x0000000081a2b000
[    1.360632] [0000000000000008] pgd=0000000000000000
[    1.363132] Oops [#1]
[    1.363748] Modules linked in:
[    1.364768] CPU: 3 UID: 0 PID: 1 Comm: swapper/0 Not tainted
6.16.0-rc1-00037-gab55e1c1d97a-dirty #7 NONE
[    1.368325] epc : mutex_lock+0x0/0x28
[    1.369796]  ra : __msi_lock_descs+0x32/0x3c
[    1.370234] epc : ffffffff80af96e8 ra : ffffffff800038e6 sp :
ff2000000004ba90
[    1.372412]  gp : ffffffff81819c00 tp : ff60000001dc0000 t0 :
6900000000000000
[    1.373527]  t1 : 0000000000000072 t2 : 6962732d76637369 s0 :
ff2000000004bab0
[    1.376628]  s1 : ff6000000241c410 a0 : 0000000000000008 a1 :
ffffffff8168ca58
[    1.379110]  a2 : 0000000000000010 a3 : 00000000000000a3 a4 :
0000000000000000
[    1.380410]  a5 : 0000000000000000 a6 : 0000000000000000 a7 :
000000004442434e
[    1.381019]  s2 : 0000000000000000 s3 : ff6000003fff30a0 s4 :
ff6000000241c410
[    1.381579]  s5 : ff600000039f9320 s6 : ff6000000241c400 s7 :
0000000000000002
[    1.382242]  s8 : ffffffff81821fa0 s9 : 0000000000000000 s10:
0000000000000000
[    1.384018]  s11: 0000000000000000 t3 : ffffffff81830a37 t4 :
ffffffff81830a37
[    1.385958]  t5 : ffffffff81830a38 t6 : ff2000000004b7c8
[    1.387306] status: 0000000200000120 badaddr: 0000000000000008
cause: 000000000000000d
[    1.388407] [<ffffffff80af96e8>] mutex_lock+0x0/0x28
[    1.389333] [<ffffffff80003dba>] msi_domain_free_irqs_all+0x2a/0x48
[    1.390275] [<ffffffff80714e86>] platform_device_msi_free_irqs_all+0x16/0x2c
[    1.391715] [<ffffffff808d8114>] mpxy_mbox_probe+0x6dc/0x750
[    1.392522] [<ffffffff806f1706>] platform_probe+0x4e/0xb4
[    1.393169] [<ffffffff806eef58>] really_probe+0x84/0x230
[    1.393789] [<ffffffff806ef160>] __driver_probe_device+0x5c/0xdc
[    1.394282] [<ffffffff806ef2a4>] driver_probe_device+0x2c/0xf8
[    1.396577] [<ffffffff806ef4ac>] __driver_attach+0x6c/0x15c
[    1.397634] [<ffffffff806ed146>] bus_for_each_dev+0x62/0xb0
[    1.399060] [<ffffffff806eea9a>] driver_attach+0x1a/0x24
[    1.399792] [<ffffffff806ee31e>] bus_add_driver+0xce/0x1d8
[    1.400363] [<ffffffff806f020c>] driver_register+0x40/0xdc
[    1.400832] [<ffffffff806f1414>] __platform_driver_register+0x1c/0x24
[    1.401551] [<ffffffff80c3df7e>] mpxy_mbox_driver_init+0x1a/0x24
[    1.402328] [<ffffffff800108b2>] do_one_initcall+0x56/0x1d8
[    1.403674] [<ffffffff80c01236>] kernel_init_freeable+0x266/0x2d0
[    1.404956] [<ffffffff80af549a>] kernel_init+0x1e/0x13c
[    1.405422] [<ffffffff80012266>] ret_from_fork_kernel+0xe/0xcc
[    1.405870] [<ffffffff80aff042>] ret_from_fork_kernel_asm+0x16/0x18

It is better to have the check on "mbox->msi_count" before calling
platform_device_msi_free_irqs_all().

Regards,
Anup

Re: [PATCH v4 17/23] ACPI: RISC-V: Create interrupt controller list in sorted order

[Andy Shevchenko]

On Tue, Jun 10, 2025 at 10:12:03AM +0530, Sunil V L wrote:
> On Wed, May 28, 2025 at 02:05:36PM +0300, Andy Shevchenko wrote:
> > On Sun, May 25, 2025 at 02:17:04PM +0530, Anup Patel wrote:

...

> > > +	__list_add(&ext_intc_element->list, node->list.prev, &node->list);
> > 
> > Is this reimplementation of list_add_tail()? And why list debug is excluded here?
> > 
> Sure. Let me use list_add_tail() itself in the next revision.

Thanks,

> However, I didn't understand the list debug question. IIUC, list_add_tail()
> is a wrapper around __list_add() and doesn't exclude the list debug, right?

You are right, it's not about debug, it's about using internal function as per
__list_add() documentation:

  * This is only for internal list manipulation where we know
  * the prev/next entries already!


-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 13/23] irqchip: Add driver for the RPMI system MSI service group

[Andy Shevchenko]

On Tue, Jun 10, 2025 at 04:33:46PM +0530, Anup Patel wrote:
> On Tue, May 27, 2025 at 5:03 PM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Sun, May 25, 2025 at 02:17:00PM +0530, Anup Patel wrote:

...

> > > +#include <linux/device.h>
> >
> > Perhaps I missed something, but devm_kzalloc() is in device/devres.h. Do you
> > need it for something else?
> 
> dev_get_msi_domain() is part of linux/device.h hence it should
> be included.

Ah, that's what I missed. Thanks for pointing out.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 07/23] mailbox: Add RISC-V SBI message proxy (MPXY) based mailbox driver

[Andy Shevchenko]

On Wed, Jun 11, 2025 at 10:51:15AM +0530, Anup Patel wrote:
> On Tue, Jun 10, 2025 at 3:25 PM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Tue, Jun 10, 2025 at 10:05:27AM +0530, Anup Patel wrote:
> > > On Tue, Jun 10, 2025 at 1:34 AM Andy Shevchenko
> > > <andriy.shevchenko@linux.intel.com> wrote:
> > > > On Mon, Jun 09, 2025 at 05:59:40PM +0530, Anup Patel wrote:
> > > > > On Wed, May 28, 2025 at 4:23 PM Andy Shevchenko
> > > > > <andriy.shevchenko@linux.intel.com> wrote:
> > > > > > On Sun, May 25, 2025 at 02:16:54PM +0530, Anup Patel wrote:

...

> > > > > > > +     if (mbox->msi_count)
> > > > > >
> > > > > > Is this check really needed?
> > > > >
> > > > > MSIs are optional for the SBI MPXY mailbox so we should only use
> > > > > platform_device_msi_xyz() APIs only when MSIs are available.
> > > >
> > > > > > > +             platform_device_msi_free_irqs_all(mbox->dev);
> > > >
> > > > Hmm... I am not sure why. Do you have any Oops or warnings if the check
> > > > is not there and no MSI provided?
> > >
> > > We don't see any oops or warnings. This check is to avoid unnecessary
> > > work (such as acquiring lock, checking default domain, etc) in the
> > > msi_domain_free_irqs_all() called by platform_device_msi_free_irqs_all().
> > >
> > > I don't mind dropping the check so I will update in the next revision.
> >
> > Perhaps you can rather add this check into the callee? Seems to me that
> > you have a justification for it. Usual pattern in the kernel that freeing
> > resources should be aware of the NULL pointers or optional resources
> > so we may call it unconditionally from the user(s).
> >
> 
> Unconditionally calling platform_device_msi_free_irqs_all() when there
> were no MSIs allocated causes the below crash because "dev->msi.data"
> is non-NULL only when:
> 
> [    1.355735] Unable to handle kernel NULL pointer dereference at
> virtual address 0000000000000008
> [    1.358212] Current swapper/0 pgtable: 4K pagesize, 57-bit VAs,
> pgdp=0x0000000081a2b000
> [    1.360632] [0000000000000008] pgd=0000000000000000
> [    1.363132] Oops [#1]
> [    1.363748] Modules linked in:
> [    1.364768] CPU: 3 UID: 0 PID: 1 Comm: swapper/0 Not tainted
> 6.16.0-rc1-00037-gab55e1c1d97a-dirty #7 NONE
> [    1.368325] epc : mutex_lock+0x0/0x28
> [    1.369796]  ra : __msi_lock_descs+0x32/0x3c
> [    1.370234] epc : ffffffff80af96e8 ra : ffffffff800038e6 sp :
> ff2000000004ba90
> [    1.372412]  gp : ffffffff81819c00 tp : ff60000001dc0000 t0 :
> 6900000000000000
> [    1.373527]  t1 : 0000000000000072 t2 : 6962732d76637369 s0 :
> ff2000000004bab0
> [    1.376628]  s1 : ff6000000241c410 a0 : 0000000000000008 a1 :
> ffffffff8168ca58
> [    1.379110]  a2 : 0000000000000010 a3 : 00000000000000a3 a4 :
> 0000000000000000
> [    1.380410]  a5 : 0000000000000000 a6 : 0000000000000000 a7 :
> 000000004442434e
> [    1.381019]  s2 : 0000000000000000 s3 : ff6000003fff30a0 s4 :
> ff6000000241c410
> [    1.381579]  s5 : ff600000039f9320 s6 : ff6000000241c400 s7 :
> 0000000000000002
> [    1.382242]  s8 : ffffffff81821fa0 s9 : 0000000000000000 s10:
> 0000000000000000
> [    1.384018]  s11: 0000000000000000 t3 : ffffffff81830a37 t4 :
> ffffffff81830a37
> [    1.385958]  t5 : ffffffff81830a38 t6 : ff2000000004b7c8
> [    1.387306] status: 0000000200000120 badaddr: 0000000000000008
> cause: 000000000000000d
> [    1.388407] [<ffffffff80af96e8>] mutex_lock+0x0/0x28
> [    1.389333] [<ffffffff80003dba>] msi_domain_free_irqs_all+0x2a/0x48
> [    1.390275] [<ffffffff80714e86>] platform_device_msi_free_irqs_all+0x16/0x2c
> [    1.391715] [<ffffffff808d8114>] mpxy_mbox_probe+0x6dc/0x750
> [    1.392522] [<ffffffff806f1706>] platform_probe+0x4e/0xb4
> [    1.393169] [<ffffffff806eef58>] really_probe+0x84/0x230
> [    1.393789] [<ffffffff806ef160>] __driver_probe_device+0x5c/0xdc
> [    1.394282] [<ffffffff806ef2a4>] driver_probe_device+0x2c/0xf8
> [    1.396577] [<ffffffff806ef4ac>] __driver_attach+0x6c/0x15c
> [    1.397634] [<ffffffff806ed146>] bus_for_each_dev+0x62/0xb0
> [    1.399060] [<ffffffff806eea9a>] driver_attach+0x1a/0x24
> [    1.399792] [<ffffffff806ee31e>] bus_add_driver+0xce/0x1d8
> [    1.400363] [<ffffffff806f020c>] driver_register+0x40/0xdc
> [    1.400832] [<ffffffff806f1414>] __platform_driver_register+0x1c/0x24
> [    1.401551] [<ffffffff80c3df7e>] mpxy_mbox_driver_init+0x1a/0x24
> [    1.402328] [<ffffffff800108b2>] do_one_initcall+0x56/0x1d8
> [    1.403674] [<ffffffff80c01236>] kernel_init_freeable+0x266/0x2d0
> [    1.404956] [<ffffffff80af549a>] kernel_init+0x1e/0x13c
> [    1.405422] [<ffffffff80012266>] ret_from_fork_kernel+0xe/0xcc
> [    1.405870] [<ffffffff80aff042>] ret_from_fork_kernel_asm+0x16/0x18
> 
> It is better to have the check on "mbox->msi_count" before calling
> platform_device_msi_free_irqs_all().

Right, thanks for confirming. But my point that this check should be made
inside the callee and not the caller. Can it be done once for all?

-- 
With Best Regards,
Andy Shevchenko