RISC-V CPU Idle Support

From: Anup Patel <[email protected]>

We force select CPU_PM and provide asm/cpuidle.h so that we can use CPU IDLE drivers for Linux RISC-V kernel.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

Hi!

Quoted 2 lines

From: Anup Patel <[email protected]>

Quoted 5 lines

We force select CPU_PM and provide asm/cpuidle.h so that we can use CPU IDLE drivers for Linux RISC-V kernel.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

This is quite... interesting. Normally we have one signoff per person...

Best regards,
Pavel
--
http://www.livejournal.com/~pavelmachek

Quoted 8 lines · Pavel Machek

Hi!

Quoted 5 lines

From: Anup Patel <[email protected]>

We force select CPU_PM and provide asm/cpuidle.h so that we can use CPU IDLE drivers for Linux RISC-V kernel.

Quoted 5 lines

Quoted 2 lines

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

This is quite... interesting. Normally we have one signoff per person...

I was working for Western Digital (WDC) when I first submitted this series and recently I joined Ventana Micro Systems.

Regards,
Anup

Quoted 5 lines

Best regards,
Pavel
--
http://www.livejournal.com/~pavelmachek

From: Anup Patel <[email protected]>

We force select CPU_PM and provide asm/cpuidle.h so that we can use CPU IDLE drivers for Linux RISC-V kernel.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

Reviewed-by Guo Ren <[email protected]>

---
 arch/riscv/Kconfig                |  7 +++++++
 arch/riscv/configs/defconfig      |  1 +
 arch/riscv/configs/rv32_defconfig |  1 +
 arch/riscv/include/asm/cpuidle.h  | 24 ++++++++++++++++++++++++
 arch/riscv/kernel/process.c       |  3 ++-
 5 files changed, 35 insertions(+), 1 deletion(-)
 create mode 100644 arch/riscv/include/asm/cpuidle.h

arch/riscv/Kconfig +7 -0

index 5adcbd9b5e88..76976d12b463 100644
@@ -46,6 +46,7 @@ config RISCV
        select CLONE_BACKWARDS
        select CLINT_TIMER if !MMU
        select COMMON_CLK
+       select CPU_PM if CPU_IDLE
        select EDAC_SUPPORT
        select GENERIC_ARCH_TOPOLOGY if SMP
        select GENERIC_ATOMIC64 if !64BIT

@@ -547,4 +548,10 @@ source "kernel/power/Kconfig"
 
 endmenu
 
+menu "CPU Power Management"
+
+source "drivers/cpuidle/Kconfig"
+
+endmenu
+
 source "arch/riscv/kvm/Kconfig"

arch/riscv/configs/defconfig +1 -0

index f120fcc43d0a..a5e0482a4969 100644
@@ -20,6 +20,7 @@ CONFIG_SOC_SIFIVE=y
 CONFIG_SOC_VIRT=y
 CONFIG_SMP=y
 CONFIG_HOTPLUG_CPU=y
+CONFIG_CPU_IDLE=y
 CONFIG_VIRTUALIZATION=y
 CONFIG_KVM=m
 CONFIG_JUMP_LABEL=y

arch/riscv/configs/rv32_defconfig +1 -0

index 8b56a7f1eb06..d1b87db54d68 100644
@@ -20,6 +20,7 @@ CONFIG_SOC_VIRT=y
 CONFIG_ARCH_RV32I=y
 CONFIG_SMP=y
 CONFIG_HOTPLUG_CPU=y
+CONFIG_CPU_IDLE=y
 CONFIG_VIRTUALIZATION=y
 CONFIG_KVM=m
 CONFIG_JUMP_LABEL=y

arch/riscv/include/asm/cpuidle.h new file +24 -0

new file mode 100644
index 000000000000..71fdc607d4bc
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2021 Allwinner Ltd
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ */
+
+#ifndef _ASM_RISCV_CPUIDLE_H
+#define _ASM_RISCV_CPUIDLE_H
+
+#include <asm/barrier.h>
+#include <asm/processor.h>
+
+static inline void cpu_do_idle(void)
+{
+       /*
+        * Add mb() here to ensure that all
+        * IO/MEM accesses are completed prior
+        * to entering WFI.
+        */
+       mb();
+       wait_for_interrupt();
+}
+
+#endif

arch/riscv/kernel/process.c +2 -2

index 03ac3aa611f5..504b496787aa 100644
@@ -23,6 +23,7 @@
 #include <asm/string.h>
 #include <asm/switch_to.h>
 #include <asm/thread_info.h>
+#include <asm/cpuidle.h>
 
 register unsigned long gp_in_global __asm__("gp");
 

@@ -37,7 +38,7 @@ extern asmlinkage void ret_from_kernel_thread(void);
 
 void arch_cpu_idle(void)
 {
-       wait_for_interrupt();
+       cpu_do_idle();
        raw_local_irq_enable();
 }
 
--

2.25.1

From: Anup Patel <[email protected]>

The low-level relocate() function enables mmu and relocates execution to link-time addresses. We rename relocate() function to relocate_enable_mmu() function which is more informative.

Also, the relocate_enable_mmu() function will be used in the resume path when a CPU wakes-up from a non-retentive suspend so we make it global symbol.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

Reviewed-by Guo Ren <[email protected]>

---
 arch/riscv/kernel/head.S | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

arch/riscv/kernel/head.S +4 -4

index 2363b43312fc..5f4c6b6c4974 100644
@@ -90,7 +90,8 @@ pe_head_start:
 
 .align 2
 #ifdef CONFIG_MMU
-relocate:
+       .global relocate_enable_mmu
+relocate_enable_mmu:
        /* Relocate return address */
        la a1, kernel_map
        XIP_FIXUP_OFFSET a1

@@ -185,7 +186,7 @@ secondary_start_sbi:
        /* Enable virtual memory and relocate to virtual address */
        la a0, swapper_pg_dir
        XIP_FIXUP_OFFSET a0
-       call relocate
+       call relocate_enable_mmu
 #endif
        call setup_trap_vector
        tail smp_callin

@@ -329,7 +330,7 @@ clear_bss_done:
 #ifdef CONFIG_MMU
        la a0, early_pg_dir
        XIP_FIXUP_OFFSET a0
-       call relocate
+       call relocate_enable_mmu
 #endif /* CONFIG_MMU */
 
        call setup_trap_vector
--

2.25.1

From: Anup Patel <[email protected]>

We add defines related to SBI HSM suspend call and also update HSM states naming as-per latest SBI specification.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

Reviewed-by Guo Ren <[email protected]>

---
 arch/riscv/include/asm/sbi.h    | 27 ++++++++++++++++++++++-----
 arch/riscv/kernel/cpu_ops_sbi.c |  2 +-
 arch/riscv/kvm/vcpu_sbi_hsm.c   |  4 ++--
 3 files changed, 25 insertions(+), 8 deletions(-)

arch/riscv/include/asm/sbi.h +22 -5

index d1c37479d828..06133b4f8e20 100644
@@ -71,15 +71,32 @@ enum sbi_ext_hsm_fid {
        SBI_EXT_HSM_HART_START = 0,
        SBI_EXT_HSM_HART_STOP,
        SBI_EXT_HSM_HART_STATUS,
+       SBI_EXT_HSM_HART_SUSPEND,
 };
 
-enum sbi_hsm_hart_status {
-       SBI_HSM_HART_STATUS_STARTED = 0,
-       SBI_HSM_HART_STATUS_STOPPED,
-       SBI_HSM_HART_STATUS_START_PENDING,
-       SBI_HSM_HART_STATUS_STOP_PENDING,
+enum sbi_hsm_hart_state {
+       SBI_HSM_STATE_STARTED = 0,
+       SBI_HSM_STATE_STOPPED,
+       SBI_HSM_STATE_START_PENDING,
+       SBI_HSM_STATE_STOP_PENDING,
+       SBI_HSM_STATE_SUSPENDED,
+       SBI_HSM_STATE_SUSPEND_PENDING,
+       SBI_HSM_STATE_RESUME_PENDING,
 };
 
+#define SBI_HSM_SUSP_BASE_MASK                 0x7fffffff
+#define SBI_HSM_SUSP_NON_RET_BIT               0x80000000
+#define SBI_HSM_SUSP_PLAT_BASE                 0x10000000
+
+#define SBI_HSM_SUSPEND_RET_DEFAULT            0x00000000
+#define SBI_HSM_SUSPEND_RET_PLATFORM           SBI_HSM_SUSP_PLAT_BASE
+#define SBI_HSM_SUSPEND_RET_LAST               SBI_HSM_SUSP_BASE_MASK
+#define SBI_HSM_SUSPEND_NON_RET_DEFAULT                SBI_HSM_SUSP_NON_RET_BIT
+#define SBI_HSM_SUSPEND_NON_RET_PLATFORM       (SBI_HSM_SUSP_NON_RET_BIT | \
+                                                SBI_HSM_SUSP_PLAT_BASE)
+#define SBI_HSM_SUSPEND_NON_RET_LAST           (SBI_HSM_SUSP_NON_RET_BIT | \
+                                                SBI_HSM_SUSP_BASE_MASK)
+
 enum sbi_ext_srst_fid {
        SBI_EXT_SRST_RESET = 0,
 };

arch/riscv/kernel/cpu_ops_sbi.c +1 -1

index dae29cbfe550..2e16f6732cdf 100644
@@ -111,7 +111,7 @@ static int sbi_cpu_is_stopped(unsigned int cpuid)
 
        rc = sbi_hsm_hart_get_status(hartid);
 
-       if (rc == SBI_HSM_HART_STATUS_STOPPED)
+       if (rc == SBI_HSM_STATE_STOPPED)
                return 0;
        return rc;
 }

arch/riscv/kvm/vcpu_sbi_hsm.c +2 -3

index 2e383687fa48..1ac4b2e8e4ec 100644
@@ -60,9 +60,9 @@ static int kvm_sbi_hsm_vcpu_get_status(struct kvm_vcpu *vcpu)
        if (!target_vcpu)
                return -EINVAL;
        if (!target_vcpu->arch.power_off)
-               return SBI_HSM_HART_STATUS_STARTED;
+               return SBI_HSM_STATE_STARTED;
        else
-               return SBI_HSM_HART_STATUS_STOPPED;
+               return SBI_HSM_STATE_STOPPED;
 }
 
 static int kvm_sbi_ext_hsm_handler(struct kvm_vcpu *vcpu, struct kvm_run *run,
--

2.25.1

From: Anup Patel <[email protected]>

We add defines related to SBI HSM suspend call and also update HSM states naming as-per latest SBI specification.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

Reviewed-by Guo Ren <[email protected]>

---
 arch/riscv/include/asm/sbi.h    | 27 ++++++++++++++++++++++-----
 arch/riscv/kernel/cpu_ops_sbi.c |  2 +-
 arch/riscv/kvm/vcpu_sbi_hsm.c   |  4 ++--
 3 files changed, 25 insertions(+), 8 deletions(-)

arch/riscv/include/asm/sbi.h +22 -5

index d1c37479d828..06133b4f8e20 100644
@@ -71,15 +71,32 @@ enum sbi_ext_hsm_fid {
        SBI_EXT_HSM_HART_START = 0,
        SBI_EXT_HSM_HART_STOP,
        SBI_EXT_HSM_HART_STATUS,
+       SBI_EXT_HSM_HART_SUSPEND,
 };
 
-enum sbi_hsm_hart_status {
-       SBI_HSM_HART_STATUS_STARTED = 0,
-       SBI_HSM_HART_STATUS_STOPPED,
-       SBI_HSM_HART_STATUS_START_PENDING,
-       SBI_HSM_HART_STATUS_STOP_PENDING,
+enum sbi_hsm_hart_state {
+       SBI_HSM_STATE_STARTED = 0,
+       SBI_HSM_STATE_STOPPED,
+       SBI_HSM_STATE_START_PENDING,
+       SBI_HSM_STATE_STOP_PENDING,
+       SBI_HSM_STATE_SUSPENDED,
+       SBI_HSM_STATE_SUSPEND_PENDING,
+       SBI_HSM_STATE_RESUME_PENDING,
 };
 
+#define SBI_HSM_SUSP_BASE_MASK                 0x7fffffff
+#define SBI_HSM_SUSP_NON_RET_BIT               0x80000000
+#define SBI_HSM_SUSP_PLAT_BASE                 0x10000000
+
+#define SBI_HSM_SUSPEND_RET_DEFAULT            0x00000000
+#define SBI_HSM_SUSPEND_RET_PLATFORM           SBI_HSM_SUSP_PLAT_BASE
+#define SBI_HSM_SUSPEND_RET_LAST               SBI_HSM_SUSP_BASE_MASK
+#define SBI_HSM_SUSPEND_NON_RET_DEFAULT                SBI_HSM_SUSP_NON_RET_BIT
+#define SBI_HSM_SUSPEND_NON_RET_PLATFORM       (SBI_HSM_SUSP_NON_RET_BIT | \
+                                                SBI_HSM_SUSP_PLAT_BASE)
+#define SBI_HSM_SUSPEND_NON_RET_LAST           (SBI_HSM_SUSP_NON_RET_BIT | \
+                                                SBI_HSM_SUSP_BASE_MASK)
+
 enum sbi_ext_srst_fid {
        SBI_EXT_SRST_RESET = 0,
 };

arch/riscv/kernel/cpu_ops_sbi.c +1 -1

index dae29cbfe550..2e16f6732cdf 100644
@@ -111,7 +111,7 @@ static int sbi_cpu_is_stopped(unsigned int cpuid)
 
        rc = sbi_hsm_hart_get_status(hartid);
 
-       if (rc == SBI_HSM_HART_STATUS_STOPPED)
+       if (rc == SBI_HSM_STATE_STOPPED)
                return 0;
        return rc;
 }

arch/riscv/kvm/vcpu_sbi_hsm.c +2 -3

index 2e383687fa48..1ac4b2e8e4ec 100644
@@ -60,9 +60,9 @@ static int kvm_sbi_hsm_vcpu_get_status(struct kvm_vcpu *vcpu)
        if (!target_vcpu)
                return -EINVAL;
        if (!target_vcpu->arch.power_off)
-               return SBI_HSM_HART_STATUS_STARTED;
+               return SBI_HSM_STATE_STARTED;
        else
-               return SBI_HSM_HART_STATUS_STOPPED;
+               return SBI_HSM_STATE_STOPPED;
 }
 
 static int kvm_sbi_ext_hsm_handler(struct kvm_vcpu *vcpu, struct kvm_run *run,
--

2.25.1

Hi Palmer

Quoted 9 lines · Anup Patel

From: Anup Patel <[email protected]>

We add defines related to SBI HSM suspend call and also update HSM states naming as-per latest SBI specification.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

Reviewed-by Guo Ren <[email protected]>

This patch is shared with "KVM RISC-V SBI v0.3 support".
(https://lore.kernel.org/all/20220201082227.361967-2-apatel at ventanamicro.com/T/)

How do you want to handle this ?

One option is that I take this patch through the KVM RISC-V tree and you can send this series (minus this patch) for 5.18 after the KVM RISC-V changes have been merged.

Regards,
Anup

Quoted 80 lines

---
 arch/riscv/include/asm/sbi.h    | 27 ++++++++++++++++++++++-----
 arch/riscv/kernel/cpu_ops_sbi.c |  2 +-
 arch/riscv/kvm/vcpu_sbi_hsm.c   |  4 ++--
 3 files changed, 25 insertions(+), 8 deletions(-)

arch/riscv/include/asm/sbi.h +22 -5

index d1c37479d828..06133b4f8e20 100644
@@ -71,15 +71,32 @@ enum sbi_ext_hsm_fid {
        SBI_EXT_HSM_HART_START = 0,
        SBI_EXT_HSM_HART_STOP,
        SBI_EXT_HSM_HART_STATUS,
+       SBI_EXT_HSM_HART_SUSPEND,
 };
 
-enum sbi_hsm_hart_status {
-       SBI_HSM_HART_STATUS_STARTED = 0,
-       SBI_HSM_HART_STATUS_STOPPED,
-       SBI_HSM_HART_STATUS_START_PENDING,
-       SBI_HSM_HART_STATUS_STOP_PENDING,
+enum sbi_hsm_hart_state {
+       SBI_HSM_STATE_STARTED = 0,
+       SBI_HSM_STATE_STOPPED,
+       SBI_HSM_STATE_START_PENDING,
+       SBI_HSM_STATE_STOP_PENDING,
+       SBI_HSM_STATE_SUSPENDED,
+       SBI_HSM_STATE_SUSPEND_PENDING,
+       SBI_HSM_STATE_RESUME_PENDING,
 };
 
+#define SBI_HSM_SUSP_BASE_MASK                 0x7fffffff
+#define SBI_HSM_SUSP_NON_RET_BIT               0x80000000
+#define SBI_HSM_SUSP_PLAT_BASE                 0x10000000
+
+#define SBI_HSM_SUSPEND_RET_DEFAULT            0x00000000
+#define SBI_HSM_SUSPEND_RET_PLATFORM           SBI_HSM_SUSP_PLAT_BASE
+#define SBI_HSM_SUSPEND_RET_LAST               SBI_HSM_SUSP_BASE_MASK
+#define SBI_HSM_SUSPEND_NON_RET_DEFAULT                SBI_HSM_SUSP_NON_RET_BIT
+#define SBI_HSM_SUSPEND_NON_RET_PLATFORM       (SBI_HSM_SUSP_NON_RET_BIT | \
+                                                SBI_HSM_SUSP_PLAT_BASE)
+#define SBI_HSM_SUSPEND_NON_RET_LAST           (SBI_HSM_SUSP_NON_RET_BIT | \
+                                                SBI_HSM_SUSP_BASE_MASK)
+
 enum sbi_ext_srst_fid {
        SBI_EXT_SRST_RESET = 0,
 };

arch/riscv/kernel/cpu_ops_sbi.c +1 -1

index dae29cbfe550..2e16f6732cdf 100644
@@ -111,7 +111,7 @@ static int sbi_cpu_is_stopped(unsigned int cpuid)
 
        rc = sbi_hsm_hart_get_status(hartid);
 
-       if (rc == SBI_HSM_HART_STATUS_STOPPED)
+       if (rc == SBI_HSM_STATE_STOPPED)
                return 0;
        return rc;
 }

arch/riscv/kvm/vcpu_sbi_hsm.c +2 -3

index 2e383687fa48..1ac4b2e8e4ec 100644
@@ -60,9 +60,9 @@ static int kvm_sbi_hsm_vcpu_get_status(struct kvm_vcpu *vcpu)
        if (!target_vcpu)
                return -EINVAL;
        if (!target_vcpu->arch.power_off)
-               return SBI_HSM_HART_STATUS_STARTED;
+               return SBI_HSM_STATE_STARTED;
        else
-               return SBI_HSM_HART_STATUS_STOPPED;
+               return SBI_HSM_STATE_STOPPED;
 }
 
 static int kvm_sbi_ext_hsm_handler(struct kvm_vcpu *vcpu, struct kvm_run *run,
--

2.25.1

From: Anup Patel <[email protected]>

The RISC-V SBI HSM extension provides HSM suspend call which can be used by Linux RISC-V to enter platform specific low-power state.

This patch adds a CPU idle driver based on RISC-V SBI calls which will populate idle states from device tree and use SBI calls to entry these idle states.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

---
 MAINTAINERS                         |   7 +
 drivers/cpuidle/Kconfig             |   5 +
 drivers/cpuidle/Kconfig.riscv       |  15 +
 drivers/cpuidle/Makefile            |   4 +
 drivers/cpuidle/cpuidle-riscv-sbi.c | 627 ++++++++++++++++++++++++++++
 5 files changed, 658 insertions(+)
 create mode 100644 drivers/cpuidle/Kconfig.riscv
 create mode 100644 drivers/cpuidle/cpuidle-riscv-sbi.c

MAINTAINERS +7 -0

index 39ece23e8d93..2ff0055a26a7 100644
@@ -5058,6 +5058,13 @@ S:       Supported
 F:     drivers/cpuidle/dt_idle_genpd.c
 F:     drivers/cpuidle/dt_idle_genpd.h
 
+CPUIDLE DRIVER - RISC-V SBI
+M:     Anup Patel <[email protected]>
+L:     linux-pm at vger.kernel.org
+L:     linux-riscv at lists.infradead.org
+S:     Maintained
+F:     drivers/cpuidle/cpuidle-riscv-sbi.c
+
 CRAMFS FILESYSTEM
 M:     Nicolas Pitre <[email protected]>
 S:     Maintained

drivers/cpuidle/Kconfig +5 -0

index f1afe7ab6b54..ff71dd662880 100644
@@ -66,6 +66,11 @@ depends on PPC
 source "drivers/cpuidle/Kconfig.powerpc"
 endmenu
 
+menu "RISC-V CPU Idle Drivers"
+depends on RISCV
+source "drivers/cpuidle/Kconfig.riscv"
+endmenu
+
 config HALTPOLL_CPUIDLE
        tristate "Halt poll cpuidle driver"
        depends on X86 && KVM_GUEST

drivers/cpuidle/Kconfig.riscv new file +15 -0

new file mode 100644
index 000000000000..78518c26af74
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# RISC-V CPU Idle drivers
+#
+
+config RISCV_SBI_CPUIDLE
+       bool "RISC-V SBI CPU idle Driver"
+       depends on RISCV_SBI
+       select DT_IDLE_STATES
+       select CPU_IDLE_MULTIPLE_DRIVERS
+       select DT_IDLE_GENPD if PM_GENERIC_DOMAINS_OF
+       help
+         Select this option to enable RISC-V SBI firmware based CPU idle
+         driver for RISC-V systems. This drivers also supports hierarchical
+         DT based layout of the idle state.

drivers/cpuidle/Makefile +4 -0

index 11a26cef279f..d103342b7cfc 100644
@@ -35,3 +35,7 @@ obj-$(CONFIG_MIPS_CPS_CPUIDLE)                += cpuidle-cps.o
 # POWERPC drivers
 obj-$(CONFIG_PSERIES_CPUIDLE)          += cpuidle-pseries.o
 obj-$(CONFIG_POWERNV_CPUIDLE)          += cpuidle-powernv.o
+
+###############################################################################
+# RISC-V drivers
+obj-$(CONFIG_RISCV_SBI_CPUIDLE)                += cpuidle-riscv-sbi.o

drivers/cpuidle/cpuidle-riscv-sbi.c new file +627 -1

new file mode 100644
index 000000000000..b459eda2cd37
@@ -0,0 +1,627 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * RISC-V SBI CPU idle driver.
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (c) 2022 Ventana Micro Systems Inc.
+ */
+
+#define pr_fmt(fmt) "cpuidle-riscv-sbi: " fmt
+
+#include <linux/cpuidle.h>
+#include <linux/cpumask.h>
+#include <linux/cpu_pm.h>
+#include <linux/cpu_cooling.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <linux/pm_runtime.h>
+#include <asm/cpuidle.h>
+#include <asm/sbi.h>
+#include <asm/suspend.h>
+
+#include "dt_idle_states.h"
+#include "dt_idle_genpd.h"
+
+struct sbi_cpuidle_data {
+       u32 *states;
+       struct device *dev;
+};
+
+struct sbi_domain_state {
+       bool available;
+       u32 state;
+};
+
+static DEFINE_PER_CPU_READ_MOSTLY(struct sbi_cpuidle_data, sbi_cpuidle_data);
+static DEFINE_PER_CPU(struct sbi_domain_state, domain_state);
+static bool sbi_cpuidle_use_osi;
+static bool sbi_cpuidle_use_cpuhp;
+static bool sbi_cpuidle_pd_allow_domain_state;
+
+static inline void sbi_set_domain_state(u32 state)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       data->available = true;
+       data->state = state;
+}
+
+static inline u32 sbi_get_domain_state(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       return data->state;
+}
+
+static inline void sbi_clear_domain_state(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       data->available = false;
+}
+
+static inline bool sbi_is_domain_state_available(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       return data->available;
+}
+
+static int sbi_suspend_finisher(unsigned long suspend_type,
+                               unsigned long resume_addr,
+                               unsigned long opaque)
+{
+       struct sbiret ret;
+
+       ret = sbi_ecall(SBI_EXT_HSM, SBI_EXT_HSM_HART_SUSPEND,
+                       suspend_type, resume_addr, opaque, 0, 0, 0);
+
+       return (ret.error) ? sbi_err_map_linux_errno(ret.error) : 0;
+}
+
+static int sbi_suspend(u32 state)
+{
+       if (state & SBI_HSM_SUSP_NON_RET_BIT)
+               return cpu_suspend(state, sbi_suspend_finisher);
+       else
+               return sbi_suspend_finisher(state, 0, 0);
+}
+
+static int sbi_cpuidle_enter_state(struct cpuidle_device *dev,
+                                  struct cpuidle_driver *drv, int idx)
+{
+       u32 *states = __this_cpu_read(sbi_cpuidle_data.states);
+
+       return CPU_PM_CPU_IDLE_ENTER_PARAM(sbi_suspend, idx, states[idx]);
+}
+
+static int __sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+                                         struct cpuidle_driver *drv, int idx,
+                                         bool s2idle)
+{
+       struct sbi_cpuidle_data *data = this_cpu_ptr(&sbi_cpuidle_data);
+       u32 *states = data->states;
+       struct device *pd_dev = data->dev;
+       u32 state;
+       int ret;
+
+       ret = cpu_pm_enter();
+       if (ret)
+               return -1;
+
+       /* Do runtime PM to manage a hierarchical CPU toplogy. */
+       rcu_irq_enter_irqson();
+       if (s2idle)
+               dev_pm_genpd_suspend(pd_dev);
+       else
+               pm_runtime_put_sync_suspend(pd_dev);
+       rcu_irq_exit_irqson();
+
+       if (sbi_is_domain_state_available())
+               state = sbi_get_domain_state();
+       else
+               state = states[idx];
+
+       ret = sbi_suspend(state) ? -1 : idx;
+
+       rcu_irq_enter_irqson();
+       if (s2idle)
+               dev_pm_genpd_resume(pd_dev);
+       else
+               pm_runtime_get_sync(pd_dev);
+       rcu_irq_exit_irqson();
+
+       cpu_pm_exit();
+
+       /* Clear the domain state to start fresh when back from idle. */
+       sbi_clear_domain_state();
+       return ret;
+}
+
+static int sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+                                      struct cpuidle_driver *drv, int idx)
+{
+       return __sbi_enter_domain_idle_state(dev, drv, idx, false);
+}
+
+static int sbi_enter_s2idle_domain_idle_state(struct cpuidle_device *dev,
+                                             struct cpuidle_driver *drv,
+                                             int idx)
+{
+       return __sbi_enter_domain_idle_state(dev, drv, idx, true);
+}
+
+static int sbi_cpuidle_cpuhp_up(unsigned int cpu)
+{
+       struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+       if (pd_dev)
+               pm_runtime_get_sync(pd_dev);
+
+       return 0;
+}
+
+static int sbi_cpuidle_cpuhp_down(unsigned int cpu)
+{
+       struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+       if (pd_dev) {
+               pm_runtime_put_sync(pd_dev);
+               /* Clear domain state to start fresh at next online. */
+               sbi_clear_domain_state();
+       }
+
+       return 0;
+}
+
+static void sbi_idle_init_cpuhp(void)
+{
+       int err;
+
+       if (!sbi_cpuidle_use_cpuhp)
+               return;
+
+       err = cpuhp_setup_state_nocalls(CPUHP_AP_CPU_PM_STARTING,
+                                       "cpuidle/sbi:online",
+                                       sbi_cpuidle_cpuhp_up,
+                                       sbi_cpuidle_cpuhp_down);
+       if (err)
+               pr_warn("Failed %d while setup cpuhp state\n", err);
+}
+
+static const struct of_device_id sbi_cpuidle_state_match[] = {
+       { .compatible = "riscv,idle-state",
+         .data = sbi_cpuidle_enter_state },
+       { },
+};
+
+static bool sbi_suspend_state_is_valid(u32 state)
+{
+       if (state > SBI_HSM_SUSPEND_RET_DEFAULT &&
+           state < SBI_HSM_SUSPEND_RET_PLATFORM)
+               return false;
+       if (state > SBI_HSM_SUSPEND_NON_RET_DEFAULT &&
+           state < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
+               return false;
+       return true;
+}
+
+static int sbi_dt_parse_state_node(struct device_node *np, u32 *state)
+{
+       int err = of_property_read_u32(np, "riscv,sbi-suspend-param", state);
+
+       if (err) {
+               pr_warn("%pOF missing riscv,sbi-suspend-param property\n", np);
+               return err;
+       }
+
+       if (!sbi_suspend_state_is_valid(*state)) {
+               pr_warn("Invalid SBI suspend state %#x\n", *state);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int sbi_dt_cpu_init_topology(struct cpuidle_driver *drv,
+                                    struct sbi_cpuidle_data *data,
+                                    unsigned int state_count, int cpu)
+{
+       /* Currently limit the hierarchical topology to be used in OSI mode. */
+       if (!sbi_cpuidle_use_osi)
+               return 0;
+
+       data->dev = dt_idle_attach_cpu(cpu, "sbi");
+       if (IS_ERR_OR_NULL(data->dev))
+               return PTR_ERR_OR_ZERO(data->dev);
+
+       /*
+        * Using the deepest state for the CPU to trigger a potential selection
+        * of a shared state for the domain, assumes the domain states are all
+        * deeper states.
+        */
+       drv->states[state_count - 1].enter = sbi_enter_domain_idle_state;
+       drv->states[state_count - 1].enter_s2idle =
+                                       sbi_enter_s2idle_domain_idle_state;
+       sbi_cpuidle_use_cpuhp = true;
+
+       return 0;
+}
+
+static int sbi_cpuidle_dt_init_states(struct device *dev,
+                                       struct cpuidle_driver *drv,
+                                       unsigned int cpu,
+                                       unsigned int state_count)
+{
+       struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+       struct device_node *state_node;
+       struct device_node *cpu_node;
+       u32 *states;
+       int i, ret;
+
+       cpu_node = of_cpu_device_node_get(cpu);
+       if (!cpu_node)
+               return -ENODEV;
+
+       states = devm_kcalloc(dev, state_count, sizeof(*states), GFP_KERNEL);
+       if (!states) {
+               ret = -ENOMEM;
+               goto fail;
+       }
+
+       /* Parse SBI specific details from state DT nodes */
+       for (i = 1; i < state_count; i++) {
+               state_node = of_get_cpu_state_node(cpu_node, i - 1);
+               if (!state_node)
+                       break;
+
+               ret = sbi_dt_parse_state_node(state_node, &states[i]);
+               of_node_put(state_node);
+
+               if (ret)
+                       return ret;
+
+               pr_debug("sbi-state %#x index %d\n", states[i], i);
+       }
+       if (i != state_count) {
+               ret = -ENODEV;
+               goto fail;
+       }
+
+       /* Initialize optional data, used for the hierarchical topology. */
+       ret = sbi_dt_cpu_init_topology(drv, data, state_count, cpu);
+       if (ret < 0)
+               return ret;
+
+       /* Store states in the per-cpu struct. */
+       data->states = states;
+
+fail:
+       of_node_put(cpu_node);
+
+       return ret;
+}
+
+static void sbi_cpuidle_deinit_cpu(int cpu)
+{
+       struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+
+       dt_idle_detach_cpu(data->dev);
+       sbi_cpuidle_use_cpuhp = false;
+}
+
+static int sbi_cpuidle_init_cpu(struct device *dev, int cpu)
+{
+       struct cpuidle_driver *drv;
+       unsigned int state_count = 0;
+       int ret = 0;
+
+       drv = devm_kzalloc(dev, sizeof(*drv), GFP_KERNEL);
+       if (!drv)
+               return -ENOMEM;
+
+       drv->name = "sbi_cpuidle";
+       drv->owner = THIS_MODULE;
+       drv->cpumask = (struct cpumask *)cpumask_of(cpu);
+
+       /* RISC-V architectural WFI to be represented as state index 0. */
+       drv->states[0].enter = sbi_cpuidle_enter_state;
+       drv->states[0].exit_latency = 1;
+       drv->states[0].target_residency = 1;
+       drv->states[0].power_usage = UINT_MAX;
+       strcpy(drv->states[0].name, "WFI");
+       strcpy(drv->states[0].desc, "RISC-V WFI");
+
+       /*
+        * If no DT idle states are detected (ret == 0) let the driver
+        * initialization fail accordingly since there is no reason to
+        * initialize the idle driver if only wfi is supported, the
+        * default archictectural back-end already executes wfi
+        * on idle entry.
+        */
+       ret = dt_init_idle_driver(drv, sbi_cpuidle_state_match, 1);
+       if (ret <= 0) {
+               pr_debug("HART%ld: failed to parse DT idle states\n",
+                        cpuid_to_hartid_map(cpu));
+               return ret ? : -ENODEV;
+       }
+       state_count = ret + 1; /* Include WFI state as well */
+
+       /* Initialize idle states from DT. */
+       ret = sbi_cpuidle_dt_init_states(dev, drv, cpu, state_count);
+       if (ret) {
+               pr_err("HART%ld: failed to init idle states\n",
+                      cpuid_to_hartid_map(cpu));
+               return ret;
+       }
+
+       ret = cpuidle_register(drv, NULL);
+       if (ret)
+               goto deinit;
+
+       cpuidle_cooling_register(drv);
+
+       return 0;
+deinit:
+       sbi_cpuidle_deinit_cpu(cpu);
+       return ret;
+}
+
+static void sbi_cpuidle_domain_sync_state(struct device *dev)
+{
+       /*
+        * All devices have now been attached/probed to the PM domain
+        * topology, hence it's fine to allow domain states to be picked.
+        */
+       sbi_cpuidle_pd_allow_domain_state = true;
+}
+
+#ifdef CONFIG_DT_IDLE_GENPD
+
+static int sbi_cpuidle_pd_power_off(struct generic_pm_domain *pd)
+{
+       struct genpd_power_state *state = &pd->states[pd->state_idx];
+       u32 *pd_state;
+
+       if (!state->data)
+               return 0;
+
+       if (!sbi_cpuidle_pd_allow_domain_state)
+               return -EBUSY;
+
+       /* OSI mode is enabled, set the corresponding domain state. */
+       pd_state = state->data;
+       sbi_set_domain_state(*pd_state);
+
+       return 0;
+}
+
+struct sbi_pd_provider {
+       struct list_head link;
+       struct device_node *node;
+};
+
+static LIST_HEAD(sbi_pd_providers);
+
+static int sbi_pd_init(struct device_node *np)
+{
+       struct generic_pm_domain *pd;
+       struct sbi_pd_provider *pd_provider;
+       struct dev_power_governor *pd_gov;
+       int ret = -ENOMEM, state_count = 0;
+
+       pd = dt_idle_pd_alloc(np, sbi_dt_parse_state_node);
+       if (!pd)
+               goto out;
+
+       pd_provider = kzalloc(sizeof(*pd_provider), GFP_KERNEL);
+       if (!pd_provider)
+               goto free_pd;
+
+       pd->flags |= GENPD_FLAG_IRQ_SAFE | GENPD_FLAG_CPU_DOMAIN;
+
+       /* Allow power off when OSI is available. */
+       if (sbi_cpuidle_use_osi)
+               pd->power_off = sbi_cpuidle_pd_power_off;
+       else
+               pd->flags |= GENPD_FLAG_ALWAYS_ON;
+
+       /* Use governor for CPU PM domains if it has some states to manage. */
+       pd_gov = state_count > 0 ? &pm_domain_cpu_gov : NULL;
+
+       ret = pm_genpd_init(pd, pd_gov, false);
+       if (ret)
+               goto free_pd_prov;
+
+       ret = of_genpd_add_provider_simple(np, pd);
+       if (ret)
+               goto remove_pd;
+
+       pd_provider->node = of_node_get(np);
+       list_add(&pd_provider->link, &sbi_pd_providers);
+
+       pr_debug("init PM domain %s\n", pd->name);
+       return 0;
+
+remove_pd:
+       pm_genpd_remove(pd);
+free_pd_prov:
+       kfree(pd_provider);
+free_pd:
+       dt_idle_pd_free(pd);
+out:
+       pr_err("failed to init PM domain ret=%d %pOF\n", ret, np);
+       return ret;
+}
+
+static void sbi_pd_remove(void)
+{
+       struct sbi_pd_provider *pd_provider, *it;
+       struct generic_pm_domain *genpd;
+
+       list_for_each_entry_safe(pd_provider, it, &sbi_pd_providers, link) {
+               of_genpd_del_provider(pd_provider->node);
+
+               genpd = of_genpd_remove_last(pd_provider->node);
+               if (!IS_ERR(genpd))
+                       kfree(genpd);
+
+               of_node_put(pd_provider->node);
+               list_del(&pd_provider->link);
+               kfree(pd_provider);
+       }
+}
+
+static int sbi_genpd_probe(struct device_node *np)
+{
+       struct device_node *node;
+       int ret = 0, pd_count = 0;
+
+       if (!np)
+               return -ENODEV;
+
+       /*
+        * Parse child nodes for the "#power-domain-cells" property and
+        * initialize a genpd/genpd-of-provider pair when it's found.
+        */
+       for_each_child_of_node(np, node) {
+               if (!of_find_property(node, "#power-domain-cells", NULL))
+                       continue;
+
+               ret = sbi_pd_init(node);
+               if (ret)
+                       goto put_node;
+
+               pd_count++;
+       }
+
+       /* Bail out if not using the hierarchical CPU topology. */
+       if (!pd_count)
+               goto no_pd;
+
+       /* Link genpd masters/subdomains to model the CPU topology. */
+       ret = dt_idle_pd_init_topology(np);
+       if (ret)
+               goto remove_pd;
+
+       return 0;
+
+put_node:
+       of_node_put(node);
+remove_pd:
+       sbi_pd_remove();
+       pr_err("failed to create CPU PM domains ret=%d\n", ret);
+no_pd:
+       return ret;
+}
+
+#else
+
+static inline int sbi_genpd_probe(struct device_node *np)
+{
+       return 0;
+}
+
+#endif
+
+static int sbi_cpuidle_probe(struct platform_device *pdev)
+{
+       int cpu, ret;
+       struct cpuidle_driver *drv;
+       struct cpuidle_device *dev;
+       struct device_node *np, *pds_node;
+
+       /* Detect OSI support based on CPU DT nodes */
+       sbi_cpuidle_use_osi = true;
+       for_each_possible_cpu(cpu) {
+               np = of_cpu_device_node_get(cpu);
+               if (np &&
+                   of_find_property(np, "power-domains", NULL) &&
+                   of_find_property(np, "power-domain-names", NULL)) {
+                       continue;
+               } else {
+                       sbi_cpuidle_use_osi = false;
+                       break;
+               }
+       }
+
+       /* Populate generic power domains from DT nodes */
+       pds_node = of_find_node_by_path("/cpus/power-domains");
+       if (pds_node) {
+               ret = sbi_genpd_probe(pds_node);
+               of_node_put(pds_node);
+               if (ret)
+                       return ret;
+       }
+
+       /* Initialize CPU idle driver for each CPU */
+       for_each_possible_cpu(cpu) {
+               ret = sbi_cpuidle_init_cpu(&pdev->dev, cpu);
+               if (ret) {
+                       pr_debug("HART%ld: idle driver init failed\n",
+                                cpuid_to_hartid_map(cpu));
+                       goto out_fail;
+               }
+       }
+
+       /* Setup CPU hotplut notifiers */
+       sbi_idle_init_cpuhp();
+
+       pr_info("idle driver registered for all CPUs\n");
+
+       return 0;
+
+out_fail:
+       while (--cpu >= 0) {
+               dev = per_cpu(cpuidle_devices, cpu);
+               drv = cpuidle_get_cpu_driver(dev);
+               cpuidle_unregister(drv);
+               sbi_cpuidle_deinit_cpu(cpu);
+       }
+
+       return ret;
+}
+
+static struct platform_driver sbi_cpuidle_driver = {
+       .probe = sbi_cpuidle_probe,
+       .driver = {
+               .name = "sbi-cpuidle",
+               .sync_state = sbi_cpuidle_domain_sync_state,
+       },
+};
+
+static int __init sbi_cpuidle_init(void)
+{
+       int ret;
+       struct platform_device *pdev;
+
+       /*
+        * The SBI HSM suspend function is only available when:
+        * 1) SBI version is 0.3 or higher
+        * 2) SBI HSM extension is available
+        */
+       if ((sbi_spec_version < sbi_mk_version(0, 3)) ||
+           sbi_probe_extension(SBI_EXT_HSM) <= 0) {
+               pr_info("HSM suspend not available\n");
+               return 0;
+       }
+
+       ret = platform_driver_register(&sbi_cpuidle_driver);
+       if (ret)
+               return ret;
+
+       pdev = platform_device_register_simple("sbi-cpuidle",
+                                               -1, NULL, 0);
+       if (IS_ERR(pdev)) {
+               platform_driver_unregister(&sbi_cpuidle_driver);
+               return PTR_ERR(pdev);
+       }
+
+       return 0;
+}
+device_initcall(sbi_cpuidle_init);
--

2.25.1

Quoted 723 lines · Anup Patel

From: Anup Patel <[email protected]>

The RISC-V SBI HSM extension provides HSM suspend call which can be used by Linux RISC-V to enter platform specific low-power state.

This patch adds a CPU idle driver based on RISC-V SBI calls which will populate idle states from device tree and use SBI calls to entry these idle states.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

---

MAINTAINERS                         |   7 +
drivers/cpuidle/Kconfig             |   5 +
drivers/cpuidle/Kconfig.riscv       |  15 +
drivers/cpuidle/Makefile            |   4 +
drivers/cpuidle/cpuidle-riscv-sbi.c | 627 ++++++++++++++++++++++++++++
5 files changed, 658 insertions(+)

create mode 100644 drivers/cpuidle/Kconfig.riscv create mode 100644 drivers/cpuidle/cpuidle-riscv-sbi.c

MAINTAINERS +0 -0

index 39ece23e8d93..2ff0055a26a7 100644
@@ -5058,6 +5058,13 @@ S:       Supported

F:     drivers/cpuidle/dt_idle_genpd.c
F:     drivers/cpuidle/dt_idle_genpd.h

MAINTAINERS +7 -0

+CPUIDLE DRIVER - RISC-V SBI
+M:     Anup Patel <[email protected]>
+L:     linux-pm at vger.kernel.org
+L:     linux-riscv at lists.infradead.org
+S:     Maintained
+F:     drivers/cpuidle/cpuidle-riscv-sbi.c
+

CRAMFS FILESYSTEM

M:     Nicolas Pitre <[email protected]>
S:     Maintained

drivers/cpuidle/Kconfig +0 -0

index f1afe7ab6b54..ff71dd662880 100644
@@ -66,6 +66,11 @@ depends on PPC

source "drivers/cpuidle/Kconfig.powerpc"
endmenu

drivers/cpuidle/Kconfig +5 -0

+menu "RISC-V CPU Idle Drivers"
+depends on RISCV
+source "drivers/cpuidle/Kconfig.riscv"
+endmenu
+

config HALTPOLL_CPUIDLE
tristate "Halt poll cpuidle driver"
depends on X86 && KVM_GUEST

drivers/cpuidle/Kconfig.riscv new file +15 -0

new file mode 100644
index 000000000000..78518c26af74
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# RISC-V CPU Idle drivers
+#
+
+config RISCV_SBI_CPUIDLE
+       bool "RISC-V SBI CPU idle Driver"
+       depends on RISCV_SBI
+       select DT_IDLE_STATES
+       select CPU_IDLE_MULTIPLE_DRIVERS
+       select DT_IDLE_GENPD if PM_GENERIC_DOMAINS_OF
+       help
+         Select this option to enable RISC-V SBI firmware based CPU idle
+         driver for RISC-V systems. This drivers also supports hierarchical
+         DT based layout of the idle state.

drivers/cpuidle/Makefile +0 -0

index 11a26cef279f..d103342b7cfc 100644
@@ -35,3 +35,7 @@ obj-$(CONFIG_MIPS_CPS_CPUIDLE)                += cpuidle-cps.o
 # POWERPC drivers

obj-$(CONFIG_PSERIES_CPUIDLE) += cpuidle-pseries.o
obj-$(CONFIG_POWERNV_CPUIDLE) += cpuidle-powernv.o

drivers/cpuidle/Makefile +4 -0

+
+###############################################################################
+# RISC-V drivers
+obj-$(CONFIG_RISCV_SBI_CPUIDLE)                += cpuidle-riscv-sbi.o

drivers/cpuidle/cpuidle-riscv-sbi.c new file +627 -1

new file mode 100644
index 000000000000..b459eda2cd37
@@ -0,0 +1,627 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * RISC-V SBI CPU idle driver.
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (c) 2022 Ventana Micro Systems Inc.
+ */
+
+#define pr_fmt(fmt) "cpuidle-riscv-sbi: " fmt
+
+#include <linux/cpuidle.h>
+#include <linux/cpumask.h>
+#include <linux/cpu_pm.h>
+#include <linux/cpu_cooling.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <linux/pm_runtime.h>
+#include <asm/cpuidle.h>
+#include <asm/sbi.h>
+#include <asm/suspend.h>
+
+#include "dt_idle_states.h"
+#include "dt_idle_genpd.h"
+
+struct sbi_cpuidle_data {
+       u32 *states;
+       struct device *dev;
+};
+
+struct sbi_domain_state {
+       bool available;
+       u32 state;
+};
+
+static DEFINE_PER_CPU_READ_MOSTLY(struct sbi_cpuidle_data, sbi_cpuidle_data);
+static DEFINE_PER_CPU(struct sbi_domain_state, domain_state);
+static bool sbi_cpuidle_use_osi;
+static bool sbi_cpuidle_use_cpuhp;
+static bool sbi_cpuidle_pd_allow_domain_state;
+
+static inline void sbi_set_domain_state(u32 state)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       data->available = true;
+       data->state = state;
+}
+
+static inline u32 sbi_get_domain_state(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       return data->state;
+}
+
+static inline void sbi_clear_domain_state(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       data->available = false;
+}
+
+static inline bool sbi_is_domain_state_available(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       return data->available;
+}
+
+static int sbi_suspend_finisher(unsigned long suspend_type,
+                               unsigned long resume_addr,
+                               unsigned long opaque)
+{
+       struct sbiret ret;
+
+       ret = sbi_ecall(SBI_EXT_HSM, SBI_EXT_HSM_HART_SUSPEND,
+                       suspend_type, resume_addr, opaque, 0, 0, 0);
+
+       return (ret.error) ? sbi_err_map_linux_errno(ret.error) : 0;
+}
+
+static int sbi_suspend(u32 state)
+{
+       if (state & SBI_HSM_SUSP_NON_RET_BIT)
+               return cpu_suspend(state, sbi_suspend_finisher);
+       else
+               return sbi_suspend_finisher(state, 0, 0);
+}
+
+static int sbi_cpuidle_enter_state(struct cpuidle_device *dev,
+                                  struct cpuidle_driver *drv, int idx)
+{
+       u32 *states = __this_cpu_read(sbi_cpuidle_data.states);
+
+       return CPU_PM_CPU_IDLE_ENTER_PARAM(sbi_suspend, idx, states[idx]);
+}
+
+static int __sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+                                         struct cpuidle_driver *drv, int idx,
+                                         bool s2idle)
+{
+       struct sbi_cpuidle_data *data = this_cpu_ptr(&sbi_cpuidle_data);
+       u32 *states = data->states;
+       struct device *pd_dev = data->dev;
+       u32 state;
+       int ret;
+
+       ret = cpu_pm_enter();
+       if (ret)
+               return -1;
+
+       /* Do runtime PM to manage a hierarchical CPU toplogy. */
+       rcu_irq_enter_irqson();
+       if (s2idle)
+               dev_pm_genpd_suspend(pd_dev);
+       else
+               pm_runtime_put_sync_suspend(pd_dev);
+       rcu_irq_exit_irqson();
+
+       if (sbi_is_domain_state_available())
+               state = sbi_get_domain_state();
+       else
+               state = states[idx];
+
+       ret = sbi_suspend(state) ? -1 : idx;
+
+       rcu_irq_enter_irqson();
+       if (s2idle)
+               dev_pm_genpd_resume(pd_dev);
+       else
+               pm_runtime_get_sync(pd_dev);
+       rcu_irq_exit_irqson();
+
+       cpu_pm_exit();
+
+       /* Clear the domain state to start fresh when back from idle. */
+       sbi_clear_domain_state();
+       return ret;
+}
+
+static int sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+                                      struct cpuidle_driver *drv, int idx)
+{
+       return __sbi_enter_domain_idle_state(dev, drv, idx, false);
+}
+
+static int sbi_enter_s2idle_domain_idle_state(struct cpuidle_device *dev,
+                                             struct cpuidle_driver *drv,
+                                             int idx)
+{
+       return __sbi_enter_domain_idle_state(dev, drv, idx, true);
+}
+
+static int sbi_cpuidle_cpuhp_up(unsigned int cpu)
+{
+       struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+       if (pd_dev)
+               pm_runtime_get_sync(pd_dev);
+
+       return 0;
+}
+
+static int sbi_cpuidle_cpuhp_down(unsigned int cpu)
+{
+       struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+       if (pd_dev) {
+               pm_runtime_put_sync(pd_dev);
+               /* Clear domain state to start fresh at next online. */
+               sbi_clear_domain_state();
+       }
+
+       return 0;
+}
+
+static void sbi_idle_init_cpuhp(void)
+{
+       int err;
+
+       if (!sbi_cpuidle_use_cpuhp)
+               return;
+
+       err = cpuhp_setup_state_nocalls(CPUHP_AP_CPU_PM_STARTING,
+                                       "cpuidle/sbi:online",
+                                       sbi_cpuidle_cpuhp_up,
+                                       sbi_cpuidle_cpuhp_down);
+       if (err)
+               pr_warn("Failed %d while setup cpuhp state\n", err);
+}
+
+static const struct of_device_id sbi_cpuidle_state_match[] = {
+       { .compatible = "riscv,idle-state",
+         .data = sbi_cpuidle_enter_state },
+       { },
+};
+
+static bool sbi_suspend_state_is_valid(u32 state)
+{
+       if (state > SBI_HSM_SUSPEND_RET_DEFAULT &&
+           state < SBI_HSM_SUSPEND_RET_PLATFORM)
+               return false;
+       if (state > SBI_HSM_SUSPEND_NON_RET_DEFAULT &&
+           state < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
+               return false;
+       return true;
+}
+
+static int sbi_dt_parse_state_node(struct device_node *np, u32 *state)
+{
+       int err = of_property_read_u32(np, "riscv,sbi-suspend-param", state);
+
+       if (err) {
+               pr_warn("%pOF missing riscv,sbi-suspend-param property\n", np);
+               return err;
+       }
+
+       if (!sbi_suspend_state_is_valid(*state)) {
+               pr_warn("Invalid SBI suspend state %#x\n", *state);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int sbi_dt_cpu_init_topology(struct cpuidle_driver *drv,
+                                    struct sbi_cpuidle_data *data,
+                                    unsigned int state_count, int cpu)
+{
+       /* Currently limit the hierarchical topology to be used in OSI mode. */
+       if (!sbi_cpuidle_use_osi)
+               return 0;
+
+       data->dev = dt_idle_attach_cpu(cpu, "sbi");
+       if (IS_ERR_OR_NULL(data->dev))
+               return PTR_ERR_OR_ZERO(data->dev);
+
+       /*
+        * Using the deepest state for the CPU to trigger a potential selection
+        * of a shared state for the domain, assumes the domain states are all
+        * deeper states.
+        */
+       drv->states[state_count - 1].enter = sbi_enter_domain_idle_state;
+       drv->states[state_count - 1].enter_s2idle =
+                                       sbi_enter_s2idle_domain_idle_state;
+       sbi_cpuidle_use_cpuhp = true;
+
+       return 0;
+}
+
+static int sbi_cpuidle_dt_init_states(struct device *dev,
+                                       struct cpuidle_driver *drv,
+                                       unsigned int cpu,
+                                       unsigned int state_count)
+{
+       struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+       struct device_node *state_node;
+       struct device_node *cpu_node;
+       u32 *states;
+       int i, ret;
+
+       cpu_node = of_cpu_device_node_get(cpu);
+       if (!cpu_node)
+               return -ENODEV;
+
+       states = devm_kcalloc(dev, state_count, sizeof(*states), GFP_KERNEL);
+       if (!states) {
+               ret = -ENOMEM;
+               goto fail;
+       }
+
+       /* Parse SBI specific details from state DT nodes */
+       for (i = 1; i < state_count; i++) {
+               state_node = of_get_cpu_state_node(cpu_node, i - 1);
+               if (!state_node)
+                       break;
+
+               ret = sbi_dt_parse_state_node(state_node, &states[i]);
+               of_node_put(state_node);
+
+               if (ret)
+                       return ret;
+
+               pr_debug("sbi-state %#x index %d\n", states[i], i);
+       }
+       if (i != state_count) {
+               ret = -ENODEV;
+               goto fail;
+       }
+
+       /* Initialize optional data, used for the hierarchical topology. */
+       ret = sbi_dt_cpu_init_topology(drv, data, state_count, cpu);
+       if (ret < 0)
+               return ret;
+
+       /* Store states in the per-cpu struct. */
+       data->states = states;
+
+fail:
+       of_node_put(cpu_node);
+
+       return ret;
+}
+
+static void sbi_cpuidle_deinit_cpu(int cpu)
+{
+       struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+
+       dt_idle_detach_cpu(data->dev);
+       sbi_cpuidle_use_cpuhp = false;
+}
+
+static int sbi_cpuidle_init_cpu(struct device *dev, int cpu)
+{
+       struct cpuidle_driver *drv;
+       unsigned int state_count = 0;
+       int ret = 0;
+
+       drv = devm_kzalloc(dev, sizeof(*drv), GFP_KERNEL);
+       if (!drv)
+               return -ENOMEM;
+
+       drv->name = "sbi_cpuidle";
+       drv->owner = THIS_MODULE;
+       drv->cpumask = (struct cpumask *)cpumask_of(cpu);
+
+       /* RISC-V architectural WFI to be represented as state index 0. */
+       drv->states[0].enter = sbi_cpuidle_enter_state;
+       drv->states[0].exit_latency = 1;
+       drv->states[0].target_residency = 1;
+       drv->states[0].power_usage = UINT_MAX;
+       strcpy(drv->states[0].name, "WFI");
+       strcpy(drv->states[0].desc, "RISC-V WFI");
+
+       /*
+        * If no DT idle states are detected (ret == 0) let the driver
+        * initialization fail accordingly since there is no reason to
+        * initialize the idle driver if only wfi is supported, the
+        * default archictectural back-end already executes wfi
+        * on idle entry.
+        */
+       ret = dt_init_idle_driver(drv, sbi_cpuidle_state_match, 1);
+       if (ret <= 0) {
+               pr_debug("HART%ld: failed to parse DT idle states\n",
+                        cpuid_to_hartid_map(cpu));
+               return ret ? : -ENODEV;
+       }
+       state_count = ret + 1; /* Include WFI state as well */
+
+       /* Initialize idle states from DT. */
+       ret = sbi_cpuidle_dt_init_states(dev, drv, cpu, state_count);
+       if (ret) {
+               pr_err("HART%ld: failed to init idle states\n",
+                      cpuid_to_hartid_map(cpu));
+               return ret;
+       }
+
+       ret = cpuidle_register(drv, NULL);
+       if (ret)
+               goto deinit;
+
+       cpuidle_cooling_register(drv);
+
+       return 0;
+deinit:
+       sbi_cpuidle_deinit_cpu(cpu);
+       return ret;
+}
+
+static void sbi_cpuidle_domain_sync_state(struct device *dev)
+{
+       /*
+        * All devices have now been attached/probed to the PM domain
+        * topology, hence it's fine to allow domain states to be picked.
+        */
+       sbi_cpuidle_pd_allow_domain_state = true;
+}
+
+#ifdef CONFIG_DT_IDLE_GENPD
+
+static int sbi_cpuidle_pd_power_off(struct generic_pm_domain *pd)
+{
+       struct genpd_power_state *state = &pd->states[pd->state_idx];
+       u32 *pd_state;
+
+       if (!state->data)
+               return 0;
+
+       if (!sbi_cpuidle_pd_allow_domain_state)
+               return -EBUSY;
+
+       /* OSI mode is enabled, set the corresponding domain state. */
+       pd_state = state->data;
+       sbi_set_domain_state(*pd_state);
+
+       return 0;
+}
+
+struct sbi_pd_provider {
+       struct list_head link;
+       struct device_node *node;
+};
+
+static LIST_HEAD(sbi_pd_providers);
+
+static int sbi_pd_init(struct device_node *np)
+{
+       struct generic_pm_domain *pd;
+       struct sbi_pd_provider *pd_provider;
+       struct dev_power_governor *pd_gov;
+       int ret = -ENOMEM, state_count = 0;
+
+       pd = dt_idle_pd_alloc(np, sbi_dt_parse_state_node);
+       if (!pd)
+               goto out;
+
+       pd_provider = kzalloc(sizeof(*pd_provider), GFP_KERNEL);
+       if (!pd_provider)
+               goto free_pd;
+
+       pd->flags |= GENPD_FLAG_IRQ_SAFE | GENPD_FLAG_CPU_DOMAIN;
+
+       /* Allow power off when OSI is available. */
+       if (sbi_cpuidle_use_osi)
+               pd->power_off = sbi_cpuidle_pd_power_off;
+       else
+               pd->flags |= GENPD_FLAG_ALWAYS_ON;
+
+       /* Use governor for CPU PM domains if it has some states to manage. */
+       pd_gov = state_count > 0 ? &pm_domain_cpu_gov : NULL;
+
+       ret = pm_genpd_init(pd, pd_gov, false);
+       if (ret)
+               goto free_pd_prov;
+
+       ret = of_genpd_add_provider_simple(np, pd);
+       if (ret)
+               goto remove_pd;
+
+       pd_provider->node = of_node_get(np);
+       list_add(&pd_provider->link, &sbi_pd_providers);
+
+       pr_debug("init PM domain %s\n", pd->name);
+       return 0;
+
+remove_pd:
+       pm_genpd_remove(pd);
+free_pd_prov:
+       kfree(pd_provider);
+free_pd:
+       dt_idle_pd_free(pd);
+out:
+       pr_err("failed to init PM domain ret=%d %pOF\n", ret, np);
+       return ret;
+}
+
+static void sbi_pd_remove(void)
+{
+       struct sbi_pd_provider *pd_provider, *it;
+       struct generic_pm_domain *genpd;
+
+       list_for_each_entry_safe(pd_provider, it, &sbi_pd_providers, link) {
+               of_genpd_del_provider(pd_provider->node);
+
+               genpd = of_genpd_remove_last(pd_provider->node);
+               if (!IS_ERR(genpd))
+                       kfree(genpd);
+
+               of_node_put(pd_provider->node);
+               list_del(&pd_provider->link);
+               kfree(pd_provider);
+       }
+}
+
+static int sbi_genpd_probe(struct device_node *np)
+{
+       struct device_node *node;
+       int ret = 0, pd_count = 0;
+
+       if (!np)
+               return -ENODEV;
+
+       /*
+        * Parse child nodes for the "#power-domain-cells" property and
+        * initialize a genpd/genpd-of-provider pair when it's found.
+        */
+       for_each_child_of_node(np, node) {
+               if (!of_find_property(node, "#power-domain-cells", NULL))
+                       continue;
+
+               ret = sbi_pd_init(node);
+               if (ret)
+                       goto put_node;
+
+               pd_count++;
+       }
+
+       /* Bail out if not using the hierarchical CPU topology. */
+       if (!pd_count)
+               goto no_pd;
+
+       /* Link genpd masters/subdomains to model the CPU topology. */
+       ret = dt_idle_pd_init_topology(np);
+       if (ret)
+               goto remove_pd;
+
+       return 0;
+
+put_node:
+       of_node_put(node);
+remove_pd:
+       sbi_pd_remove();
+       pr_err("failed to create CPU PM domains ret=%d\n", ret);
+no_pd:
+       return ret;
+}
+
+#else
+
+static inline int sbi_genpd_probe(struct device_node *np)
+{
+       return 0;
+}
+
+#endif
+
+static int sbi_cpuidle_probe(struct platform_device *pdev)
+{
+       int cpu, ret;
+       struct cpuidle_driver *drv;
+       struct cpuidle_device *dev;
+       struct device_node *np, *pds_node;
+
+       /* Detect OSI support based on CPU DT nodes */
+       sbi_cpuidle_use_osi = true;
+       for_each_possible_cpu(cpu) {
+               np = of_cpu_device_node_get(cpu);
+               if (np &&
+                   of_find_property(np, "power-domains", NULL) &&
+                   of_find_property(np, "power-domain-names", NULL)) {
+                       continue;
+               } else {
+                       sbi_cpuidle_use_osi = false;
+                       break;
+               }
+       }
+
+       /* Populate generic power domains from DT nodes */
+       pds_node = of_find_node_by_path("/cpus/power-domains");
+       if (pds_node) {
+               ret = sbi_genpd_probe(pds_node);
+               of_node_put(pds_node);
+               if (ret)
+                       return ret;
+       }
+
+       /* Initialize CPU idle driver for each CPU */
+       for_each_possible_cpu(cpu) {
+               ret = sbi_cpuidle_init_cpu(&pdev->dev, cpu);
+               if (ret) {
+                       pr_debug("HART%ld: idle driver init failed\n",
+                                cpuid_to_hartid_map(cpu));
+                       goto out_fail;
+               }
+       }
+
+       /* Setup CPU hotplut notifiers */
+       sbi_idle_init_cpuhp();
+
+       pr_info("idle driver registered for all CPUs\n");
+
+       return 0;
+
+out_fail:
+       while (--cpu >= 0) {
+               dev = per_cpu(cpuidle_devices, cpu);
+               drv = cpuidle_get_cpu_driver(dev);
+               cpuidle_unregister(drv);
+               sbi_cpuidle_deinit_cpu(cpu);
+       }
+
+       return ret;
+}
+
+static struct platform_driver sbi_cpuidle_driver = {
+       .probe = sbi_cpuidle_probe,
+       .driver = {
+               .name = "sbi-cpuidle",
+               .sync_state = sbi_cpuidle_domain_sync_state,
+       },
+};
+
+static int __init sbi_cpuidle_init(void)
+{
+       int ret;
+       struct platform_device *pdev;
+
+       /*
+        * The SBI HSM suspend function is only available when:
+        * 1) SBI version is 0.3 or higher
+        * 2) SBI HSM extension is available
+        */
+       if ((sbi_spec_version < sbi_mk_version(0, 3)) ||
+           sbi_probe_extension(SBI_EXT_HSM) <= 0) {
+               pr_info("HSM suspend not available\n");
+               return 0;
+       }
+
+       ret = platform_driver_register(&sbi_cpuidle_driver);
+       if (ret)
+               return ret;
+
+       pdev = platform_device_register_simple("sbi-cpuidle",
+                                               -1, NULL, 0);
+       if (IS_ERR(pdev)) {
+               platform_driver_unregister(&sbi_cpuidle_driver);
+               return PTR_ERR(pdev);
+       }
+
+       return 0;
+}
+device_initcall(sbi_cpuidle_init);
--

Quoted 2 lines

2.25.1

For the SBI part,

Acked-by Atish Patra <[email protected]>

FYI..
SBI HSM suspend was included in SBI v0.3. The current version of the SBI specification (v1.0-rc2)
is already frozen as per the RVI guidelines. All the comments received during the public review period
have been addressed as well.

Quoted 731 lines · Atish Patra

Quoted 723 lines · Anup Patel

From: Anup Patel <[email protected]>

The RISC-V SBI HSM extension provides HSM suspend call which can be used by Linux RISC-V to enter platform specific low-power state.

This patch adds a CPU idle driver based on RISC-V SBI calls which will populate idle states from device tree and use SBI calls to entry these idle states.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

---

MAINTAINERS                         |   7 +
drivers/cpuidle/Kconfig             |   5 +
drivers/cpuidle/Kconfig.riscv       |  15 +
drivers/cpuidle/Makefile            |   4 +
drivers/cpuidle/cpuidle-riscv-sbi.c | 627 ++++++++++++++++++++++++++++
5 files changed, 658 insertions(+)

create mode 100644 drivers/cpuidle/Kconfig.riscv create mode 100644 drivers/cpuidle/cpuidle-riscv-sbi.c

MAINTAINERS +0 -0

index 39ece23e8d93..2ff0055a26a7 100644
@@ -5058,6 +5058,13 @@ S:       Supported

F:     drivers/cpuidle/dt_idle_genpd.c
F:     drivers/cpuidle/dt_idle_genpd.h

MAINTAINERS +7 -0

+CPUIDLE DRIVER - RISC-V SBI
+M:     Anup Patel <[email protected]>
+L:     linux-pm at vger.kernel.org
+L:     linux-riscv at lists.infradead.org
+S:     Maintained
+F:     drivers/cpuidle/cpuidle-riscv-sbi.c
+

CRAMFS FILESYSTEM

M:     Nicolas Pitre <[email protected]>
S:     Maintained

drivers/cpuidle/Kconfig +0 -0

index f1afe7ab6b54..ff71dd662880 100644
@@ -66,6 +66,11 @@ depends on PPC

source "drivers/cpuidle/Kconfig.powerpc"
endmenu

drivers/cpuidle/Kconfig +5 -0

+menu "RISC-V CPU Idle Drivers"
+depends on RISCV
+source "drivers/cpuidle/Kconfig.riscv"
+endmenu
+

config HALTPOLL_CPUIDLE
tristate "Halt poll cpuidle driver"
depends on X86 && KVM_GUEST

drivers/cpuidle/Kconfig.riscv new file +15 -0

new file mode 100644
index 000000000000..78518c26af74
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# RISC-V CPU Idle drivers
+#
+
+config RISCV_SBI_CPUIDLE
+       bool "RISC-V SBI CPU idle Driver"
+       depends on RISCV_SBI
+       select DT_IDLE_STATES
+       select CPU_IDLE_MULTIPLE_DRIVERS
+       select DT_IDLE_GENPD if PM_GENERIC_DOMAINS_OF
+       help
+         Select this option to enable RISC-V SBI firmware based CPU idle
+         driver for RISC-V systems. This drivers also supports hierarchical
+         DT based layout of the idle state.

drivers/cpuidle/Makefile +0 -0

index 11a26cef279f..d103342b7cfc 100644
@@ -35,3 +35,7 @@ obj-$(CONFIG_MIPS_CPS_CPUIDLE)                += cpuidle-cps.o
 # POWERPC drivers

obj-$(CONFIG_PSERIES_CPUIDLE) += cpuidle-pseries.o
obj-$(CONFIG_POWERNV_CPUIDLE) += cpuidle-powernv.o

drivers/cpuidle/Makefile +4 -0

+
+###############################################################################
+# RISC-V drivers
+obj-$(CONFIG_RISCV_SBI_CPUIDLE)                += cpuidle-riscv-sbi.o

drivers/cpuidle/cpuidle-riscv-sbi.c new file +627 -1

new file mode 100644
index 000000000000..b459eda2cd37
@@ -0,0 +1,627 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * RISC-V SBI CPU idle driver.
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (c) 2022 Ventana Micro Systems Inc.
+ */
+
+#define pr_fmt(fmt) "cpuidle-riscv-sbi: " fmt
+
+#include <linux/cpuidle.h>
+#include <linux/cpumask.h>
+#include <linux/cpu_pm.h>
+#include <linux/cpu_cooling.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <linux/pm_runtime.h>
+#include <asm/cpuidle.h>
+#include <asm/sbi.h>
+#include <asm/suspend.h>
+
+#include "dt_idle_states.h"
+#include "dt_idle_genpd.h"
+
+struct sbi_cpuidle_data {
+       u32 *states;
+       struct device *dev;
+};
+
+struct sbi_domain_state {
+       bool available;
+       u32 state;
+};
+
+static DEFINE_PER_CPU_READ_MOSTLY(struct sbi_cpuidle_data, sbi_cpuidle_data);
+static DEFINE_PER_CPU(struct sbi_domain_state, domain_state);
+static bool sbi_cpuidle_use_osi;
+static bool sbi_cpuidle_use_cpuhp;
+static bool sbi_cpuidle_pd_allow_domain_state;
+
+static inline void sbi_set_domain_state(u32 state)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       data->available = true;
+       data->state = state;
+}
+
+static inline u32 sbi_get_domain_state(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       return data->state;
+}
+
+static inline void sbi_clear_domain_state(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       data->available = false;
+}
+
+static inline bool sbi_is_domain_state_available(void)
+{
+       struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+       return data->available;
+}
+
+static int sbi_suspend_finisher(unsigned long suspend_type,
+                               unsigned long resume_addr,
+                               unsigned long opaque)
+{
+       struct sbiret ret;
+
+       ret = sbi_ecall(SBI_EXT_HSM, SBI_EXT_HSM_HART_SUSPEND,
+                       suspend_type, resume_addr, opaque, 0, 0, 0);
+
+       return (ret.error) ? sbi_err_map_linux_errno(ret.error) : 0;
+}
+
+static int sbi_suspend(u32 state)
+{
+       if (state & SBI_HSM_SUSP_NON_RET_BIT)
+               return cpu_suspend(state, sbi_suspend_finisher);
+       else
+               return sbi_suspend_finisher(state, 0, 0);
+}
+
+static int sbi_cpuidle_enter_state(struct cpuidle_device *dev,
+                                  struct cpuidle_driver *drv, int idx)
+{
+       u32 *states = __this_cpu_read(sbi_cpuidle_data.states);
+
+       return CPU_PM_CPU_IDLE_ENTER_PARAM(sbi_suspend, idx, states[idx]);
+}
+
+static int __sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+                                         struct cpuidle_driver *drv, int idx,
+                                         bool s2idle)
+{
+       struct sbi_cpuidle_data *data = this_cpu_ptr(&sbi_cpuidle_data);
+       u32 *states = data->states;
+       struct device *pd_dev = data->dev;
+       u32 state;
+       int ret;
+
+       ret = cpu_pm_enter();
+       if (ret)
+               return -1;
+
+       /* Do runtime PM to manage a hierarchical CPU toplogy. */
+       rcu_irq_enter_irqson();
+       if (s2idle)
+               dev_pm_genpd_suspend(pd_dev);
+       else
+               pm_runtime_put_sync_suspend(pd_dev);
+       rcu_irq_exit_irqson();
+
+       if (sbi_is_domain_state_available())
+               state = sbi_get_domain_state();
+       else
+               state = states[idx];
+
+       ret = sbi_suspend(state) ? -1 : idx;
+
+       rcu_irq_enter_irqson();
+       if (s2idle)
+               dev_pm_genpd_resume(pd_dev);
+       else
+               pm_runtime_get_sync(pd_dev);
+       rcu_irq_exit_irqson();
+
+       cpu_pm_exit();
+
+       /* Clear the domain state to start fresh when back from idle. */
+       sbi_clear_domain_state();
+       return ret;
+}
+
+static int sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+                                      struct cpuidle_driver *drv, int idx)
+{
+       return __sbi_enter_domain_idle_state(dev, drv, idx, false);
+}
+
+static int sbi_enter_s2idle_domain_idle_state(struct cpuidle_device *dev,
+                                             struct cpuidle_driver *drv,
+                                             int idx)
+{
+       return __sbi_enter_domain_idle_state(dev, drv, idx, true);
+}
+
+static int sbi_cpuidle_cpuhp_up(unsigned int cpu)
+{
+       struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+       if (pd_dev)
+               pm_runtime_get_sync(pd_dev);
+
+       return 0;
+}
+
+static int sbi_cpuidle_cpuhp_down(unsigned int cpu)
+{
+       struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+       if (pd_dev) {
+               pm_runtime_put_sync(pd_dev);
+               /* Clear domain state to start fresh at next online. */
+               sbi_clear_domain_state();
+       }
+
+       return 0;
+}
+
+static void sbi_idle_init_cpuhp(void)
+{
+       int err;
+
+       if (!sbi_cpuidle_use_cpuhp)
+               return;
+
+       err = cpuhp_setup_state_nocalls(CPUHP_AP_CPU_PM_STARTING,
+                                       "cpuidle/sbi:online",
+                                       sbi_cpuidle_cpuhp_up,
+                                       sbi_cpuidle_cpuhp_down);
+       if (err)
+               pr_warn("Failed %d while setup cpuhp state\n", err);
+}
+
+static const struct of_device_id sbi_cpuidle_state_match[] = {
+       { .compatible = "riscv,idle-state",
+         .data = sbi_cpuidle_enter_state },
+       { },
+};
+
+static bool sbi_suspend_state_is_valid(u32 state)
+{
+       if (state > SBI_HSM_SUSPEND_RET_DEFAULT &&
+           state < SBI_HSM_SUSPEND_RET_PLATFORM)
+               return false;
+       if (state > SBI_HSM_SUSPEND_NON_RET_DEFAULT &&
+           state < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
+               return false;
+       return true;
+}
+
+static int sbi_dt_parse_state_node(struct device_node *np, u32 *state)
+{
+       int err = of_property_read_u32(np, "riscv,sbi-suspend-param", state);
+
+       if (err) {
+               pr_warn("%pOF missing riscv,sbi-suspend-param property\n", np);
+               return err;
+       }
+
+       if (!sbi_suspend_state_is_valid(*state)) {
+               pr_warn("Invalid SBI suspend state %#x\n", *state);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int sbi_dt_cpu_init_topology(struct cpuidle_driver *drv,
+                                    struct sbi_cpuidle_data *data,
+                                    unsigned int state_count, int cpu)
+{
+       /* Currently limit the hierarchical topology to be used in OSI mode. */
+       if (!sbi_cpuidle_use_osi)
+               return 0;
+
+       data->dev = dt_idle_attach_cpu(cpu, "sbi");
+       if (IS_ERR_OR_NULL(data->dev))
+               return PTR_ERR_OR_ZERO(data->dev);
+
+       /*
+        * Using the deepest state for the CPU to trigger a potential selection
+        * of a shared state for the domain, assumes the domain states are all
+        * deeper states.
+        */
+       drv->states[state_count - 1].enter = sbi_enter_domain_idle_state;
+       drv->states[state_count - 1].enter_s2idle =
+                                       sbi_enter_s2idle_domain_idle_state;
+       sbi_cpuidle_use_cpuhp = true;
+
+       return 0;
+}
+
+static int sbi_cpuidle_dt_init_states(struct device *dev,
+                                       struct cpuidle_driver *drv,
+                                       unsigned int cpu,
+                                       unsigned int state_count)
+{
+       struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+       struct device_node *state_node;
+       struct device_node *cpu_node;
+       u32 *states;
+       int i, ret;
+
+       cpu_node = of_cpu_device_node_get(cpu);
+       if (!cpu_node)
+               return -ENODEV;
+
+       states = devm_kcalloc(dev, state_count, sizeof(*states), GFP_KERNEL);
+       if (!states) {
+               ret = -ENOMEM;
+               goto fail;
+       }
+
+       /* Parse SBI specific details from state DT nodes */
+       for (i = 1; i < state_count; i++) {
+               state_node = of_get_cpu_state_node(cpu_node, i - 1);
+               if (!state_node)
+                       break;
+
+               ret = sbi_dt_parse_state_node(state_node, &states[i]);
+               of_node_put(state_node);
+
+               if (ret)
+                       return ret;
+
+               pr_debug("sbi-state %#x index %d\n", states[i], i);
+       }
+       if (i != state_count) {
+               ret = -ENODEV;
+               goto fail;
+       }
+
+       /* Initialize optional data, used for the hierarchical topology. */
+       ret = sbi_dt_cpu_init_topology(drv, data, state_count, cpu);
+       if (ret < 0)
+               return ret;
+
+       /* Store states in the per-cpu struct. */
+       data->states = states;
+
+fail:
+       of_node_put(cpu_node);
+
+       return ret;
+}
+
+static void sbi_cpuidle_deinit_cpu(int cpu)
+{
+       struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+
+       dt_idle_detach_cpu(data->dev);
+       sbi_cpuidle_use_cpuhp = false;
+}
+
+static int sbi_cpuidle_init_cpu(struct device *dev, int cpu)
+{
+       struct cpuidle_driver *drv;
+       unsigned int state_count = 0;
+       int ret = 0;
+
+       drv = devm_kzalloc(dev, sizeof(*drv), GFP_KERNEL);
+       if (!drv)
+               return -ENOMEM;
+
+       drv->name = "sbi_cpuidle";
+       drv->owner = THIS_MODULE;
+       drv->cpumask = (struct cpumask *)cpumask_of(cpu);
+
+       /* RISC-V architectural WFI to be represented as state index 0. */
+       drv->states[0].enter = sbi_cpuidle_enter_state;
+       drv->states[0].exit_latency = 1;
+       drv->states[0].target_residency = 1;
+       drv->states[0].power_usage = UINT_MAX;
+       strcpy(drv->states[0].name, "WFI");
+       strcpy(drv->states[0].desc, "RISC-V WFI");
+
+       /*
+        * If no DT idle states are detected (ret == 0) let the driver
+        * initialization fail accordingly since there is no reason to
+        * initialize the idle driver if only wfi is supported, the
+        * default archictectural back-end already executes wfi
+        * on idle entry.
+        */
+       ret = dt_init_idle_driver(drv, sbi_cpuidle_state_match, 1);
+       if (ret <= 0) {
+               pr_debug("HART%ld: failed to parse DT idle states\n",
+                        cpuid_to_hartid_map(cpu));
+               return ret ? : -ENODEV;
+       }
+       state_count = ret + 1; /* Include WFI state as well */
+
+       /* Initialize idle states from DT. */
+       ret = sbi_cpuidle_dt_init_states(dev, drv, cpu, state_count);
+       if (ret) {
+               pr_err("HART%ld: failed to init idle states\n",
+                      cpuid_to_hartid_map(cpu));
+               return ret;
+       }
+
+       ret = cpuidle_register(drv, NULL);
+       if (ret)
+               goto deinit;
+
+       cpuidle_cooling_register(drv);
+
+       return 0;
+deinit:
+       sbi_cpuidle_deinit_cpu(cpu);
+       return ret;
+}
+
+static void sbi_cpuidle_domain_sync_state(struct device *dev)
+{
+       /*
+        * All devices have now been attached/probed to the PM domain
+        * topology, hence it's fine to allow domain states to be picked.
+        */
+       sbi_cpuidle_pd_allow_domain_state = true;
+}
+
+#ifdef CONFIG_DT_IDLE_GENPD
+
+static int sbi_cpuidle_pd_power_off(struct generic_pm_domain *pd)
+{
+       struct genpd_power_state *state = &pd->states[pd->state_idx];
+       u32 *pd_state;
+
+       if (!state->data)
+               return 0;
+
+       if (!sbi_cpuidle_pd_allow_domain_state)
+               return -EBUSY;
+
+       /* OSI mode is enabled, set the corresponding domain state. */
+       pd_state = state->data;
+       sbi_set_domain_state(*pd_state);
+
+       return 0;
+}
+
+struct sbi_pd_provider {
+       struct list_head link;
+       struct device_node *node;
+};
+
+static LIST_HEAD(sbi_pd_providers);
+
+static int sbi_pd_init(struct device_node *np)
+{
+       struct generic_pm_domain *pd;
+       struct sbi_pd_provider *pd_provider;
+       struct dev_power_governor *pd_gov;
+       int ret = -ENOMEM, state_count = 0;
+
+       pd = dt_idle_pd_alloc(np, sbi_dt_parse_state_node);
+       if (!pd)
+               goto out;
+
+       pd_provider = kzalloc(sizeof(*pd_provider), GFP_KERNEL);
+       if (!pd_provider)
+               goto free_pd;
+
+       pd->flags |= GENPD_FLAG_IRQ_SAFE | GENPD_FLAG_CPU_DOMAIN;
+
+       /* Allow power off when OSI is available. */
+       if (sbi_cpuidle_use_osi)
+               pd->power_off = sbi_cpuidle_pd_power_off;
+       else
+               pd->flags |= GENPD_FLAG_ALWAYS_ON;
+
+       /* Use governor for CPU PM domains if it has some states to manage. */
+       pd_gov = state_count > 0 ? &pm_domain_cpu_gov : NULL;
+
+       ret = pm_genpd_init(pd, pd_gov, false);
+       if (ret)
+               goto free_pd_prov;
+
+       ret = of_genpd_add_provider_simple(np, pd);
+       if (ret)
+               goto remove_pd;
+
+       pd_provider->node = of_node_get(np);
+       list_add(&pd_provider->link, &sbi_pd_providers);
+
+       pr_debug("init PM domain %s\n", pd->name);
+       return 0;
+
+remove_pd:
+       pm_genpd_remove(pd);
+free_pd_prov:
+       kfree(pd_provider);
+free_pd:
+       dt_idle_pd_free(pd);
+out:
+       pr_err("failed to init PM domain ret=%d %pOF\n", ret, np);
+       return ret;
+}
+
+static void sbi_pd_remove(void)
+{
+       struct sbi_pd_provider *pd_provider, *it;
+       struct generic_pm_domain *genpd;
+
+       list_for_each_entry_safe(pd_provider, it, &sbi_pd_providers, link) {
+               of_genpd_del_provider(pd_provider->node);
+
+               genpd = of_genpd_remove_last(pd_provider->node);
+               if (!IS_ERR(genpd))
+                       kfree(genpd);
+
+               of_node_put(pd_provider->node);
+               list_del(&pd_provider->link);
+               kfree(pd_provider);
+       }
+}
+
+static int sbi_genpd_probe(struct device_node *np)
+{
+       struct device_node *node;
+       int ret = 0, pd_count = 0;
+
+       if (!np)
+               return -ENODEV;
+
+       /*
+        * Parse child nodes for the "#power-domain-cells" property and
+        * initialize a genpd/genpd-of-provider pair when it's found.
+        */
+       for_each_child_of_node(np, node) {
+               if (!of_find_property(node, "#power-domain-cells", NULL))
+                       continue;
+
+               ret = sbi_pd_init(node);
+               if (ret)
+                       goto put_node;
+
+               pd_count++;
+       }
+
+       /* Bail out if not using the hierarchical CPU topology. */
+       if (!pd_count)
+               goto no_pd;
+
+       /* Link genpd masters/subdomains to model the CPU topology. */
+       ret = dt_idle_pd_init_topology(np);
+       if (ret)
+               goto remove_pd;
+
+       return 0;
+
+put_node:
+       of_node_put(node);
+remove_pd:
+       sbi_pd_remove();
+       pr_err("failed to create CPU PM domains ret=%d\n", ret);
+no_pd:
+       return ret;
+}
+
+#else
+
+static inline int sbi_genpd_probe(struct device_node *np)
+{
+       return 0;
+}
+
+#endif
+
+static int sbi_cpuidle_probe(struct platform_device *pdev)
+{
+       int cpu, ret;
+       struct cpuidle_driver *drv;
+       struct cpuidle_device *dev;
+       struct device_node *np, *pds_node;
+
+       /* Detect OSI support based on CPU DT nodes */
+       sbi_cpuidle_use_osi = true;
+       for_each_possible_cpu(cpu) {
+               np = of_cpu_device_node_get(cpu);
+               if (np &&
+                   of_find_property(np, "power-domains", NULL) &&
+                   of_find_property(np, "power-domain-names", NULL)) {
+                       continue;
+               } else {
+                       sbi_cpuidle_use_osi = false;
+                       break;
+               }
+       }
+
+       /* Populate generic power domains from DT nodes */
+       pds_node = of_find_node_by_path("/cpus/power-domains");
+       if (pds_node) {
+               ret = sbi_genpd_probe(pds_node);
+               of_node_put(pds_node);
+               if (ret)
+                       return ret;
+       }
+
+       /* Initialize CPU idle driver for each CPU */
+       for_each_possible_cpu(cpu) {
+               ret = sbi_cpuidle_init_cpu(&pdev->dev, cpu);
+               if (ret) {
+                       pr_debug("HART%ld: idle driver init failed\n",
+                                cpuid_to_hartid_map(cpu));
+                       goto out_fail;
+               }
+       }
+
+       /* Setup CPU hotplut notifiers */
+       sbi_idle_init_cpuhp();
+
+       pr_info("idle driver registered for all CPUs\n");
+
+       return 0;
+
+out_fail:
+       while (--cpu >= 0) {
+               dev = per_cpu(cpuidle_devices, cpu);
+               drv = cpuidle_get_cpu_driver(dev);
+               cpuidle_unregister(drv);
+               sbi_cpuidle_deinit_cpu(cpu);
+       }
+
+       return ret;
+}
+
+static struct platform_driver sbi_cpuidle_driver = {
+       .probe = sbi_cpuidle_probe,
+       .driver = {
+               .name = "sbi-cpuidle",
+               .sync_state = sbi_cpuidle_domain_sync_state,
+       },
+};
+
+static int __init sbi_cpuidle_init(void)
+{
+       int ret;
+       struct platform_device *pdev;
+
+       /*
+        * The SBI HSM suspend function is only available when:
+        * 1) SBI version is 0.3 or higher
+        * 2) SBI HSM extension is available
+        */
+       if ((sbi_spec_version < sbi_mk_version(0, 3)) ||
+           sbi_probe_extension(SBI_EXT_HSM) <= 0) {
+               pr_info("HSM suspend not available\n");
+               return 0;
+       }
+
+       ret = platform_driver_register(&sbi_cpuidle_driver);
+       if (ret)
+               return ret;
+
+       pdev = platform_device_register_simple("sbi-cpuidle",
+                                               -1, NULL, 0);
+       if (IS_ERR(pdev)) {
+               platform_driver_unregister(&sbi_cpuidle_driver);
+               return PTR_ERR(pdev);
+       }
+
+       return 0;
+}
+device_initcall(sbi_cpuidle_init);
--

Quoted 2 lines

2.25.1

For the SBI part,

Acked-by Atish Patra <[email protected]>

Quoted 6 lines

FYI..
SBI HSM suspend was included in SBI v0.3. The current version of the SBI specification (v1.0-rc2)
is already frozen as per the RVI guidelines. All the comments received during the public review period
have been addressed as well.

Yet not all comments from *before* the public review period.

Jess

From: Anup Patel <[email protected]>

The RISC-V SBI HSM extension provides HSM suspend call which can be used by Linux RISC-V to enter platform specific low-power state.

This patch adds a CPU idle driver based on RISC-V SBI calls which will populate idle states from device tree and use SBI calls to entry these idle states.

Signed-off-by Anup Patel <[email protected]>

Signed-off-by Anup Patel <[email protected]>

---
 MAINTAINERS                         |   7 +
 drivers/cpuidle/Kconfig             |   5 +
 drivers/cpuidle/Kconfig.riscv       |  15 +
 drivers/cpuidle/Makefile            |   4 +
 drivers/cpuidle/cpuidle-riscv-sbi.c | 627 ++++++++++++++++++++++++++++
 5 files changed, 658 insertions(+)
 create mode 100644 drivers/cpuidle/Kconfig.riscv
 create mode 100644 drivers/cpuidle/cpuidle-riscv-sbi.c

MAINTAINERS +7 -0

index 39ece23e8d93..2ff0055a26a7 100644
@@ -5058,6 +5058,13 @@ S:	Supported
 F:	drivers/cpuidle/dt_idle_genpd.c
 F:	drivers/cpuidle/dt_idle_genpd.h
 
+CPUIDLE DRIVER - RISC-V SBI
+M:	Anup Patel <[email protected]>
+L:	linux-pm at vger.kernel.org
+L:	linux-riscv at lists.infradead.org
+S:	Maintained
+F:	drivers/cpuidle/cpuidle-riscv-sbi.c
+
 CRAMFS FILESYSTEM
 M:	Nicolas Pitre <[email protected]>
 S:	Maintained

drivers/cpuidle/Kconfig +5 -0

index f1afe7ab6b54..ff71dd662880 100644
@@ -66,6 +66,11 @@ depends on PPC
 source "drivers/cpuidle/Kconfig.powerpc"
 endmenu
 
+menu "RISC-V CPU Idle Drivers"
+depends on RISCV
+source "drivers/cpuidle/Kconfig.riscv"
+endmenu
+
 config HALTPOLL_CPUIDLE
 	tristate "Halt poll cpuidle driver"
 	depends on X86 && KVM_GUEST

drivers/cpuidle/Kconfig.riscv new file +15 -0

new file mode 100644
index 000000000000..78518c26af74
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# RISC-V CPU Idle drivers
+#
+
+config RISCV_SBI_CPUIDLE
+	bool "RISC-V SBI CPU idle Driver"
+	depends on RISCV_SBI
+	select DT_IDLE_STATES
+	select CPU_IDLE_MULTIPLE_DRIVERS
+	select DT_IDLE_GENPD if PM_GENERIC_DOMAINS_OF
+	help
+	  Select this option to enable RISC-V SBI firmware based CPU idle
+	  driver for RISC-V systems. This drivers also supports hierarchical
+	  DT based layout of the idle state.

drivers/cpuidle/Makefile +4 -0

index 11a26cef279f..d103342b7cfc 100644
@@ -35,3 +35,7 @@ obj-$(CONFIG_MIPS_CPS_CPUIDLE)		+= cpuidle-cps.o
 # POWERPC drivers
 obj-$(CONFIG_PSERIES_CPUIDLE)		+= cpuidle-pseries.o
 obj-$(CONFIG_POWERNV_CPUIDLE)		+= cpuidle-powernv.o
+
+###############################################################################
+# RISC-V drivers
+obj-$(CONFIG_RISCV_SBI_CPUIDLE)		+= cpuidle-riscv-sbi.o

drivers/cpuidle/cpuidle-riscv-sbi.c new file +627 -0

new file mode 100644
index 000000000000..b459eda2cd37
@@ -0,0 +1,627 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * RISC-V SBI CPU idle driver.
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (c) 2022 Ventana Micro Systems Inc.
+ */
+
+#define pr_fmt(fmt) "cpuidle-riscv-sbi: " fmt
+
+#include <linux/cpuidle.h>
+#include <linux/cpumask.h>
+#include <linux/cpu_pm.h>
+#include <linux/cpu_cooling.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <linux/pm_runtime.h>
+#include <asm/cpuidle.h>
+#include <asm/sbi.h>
+#include <asm/suspend.h>
+
+#include "dt_idle_states.h"
+#include "dt_idle_genpd.h"
+
+struct sbi_cpuidle_data {
+	u32 *states;
+	struct device *dev;
+};
+
+struct sbi_domain_state {
+	bool available;
+	u32 state;
+};
+
+static DEFINE_PER_CPU_READ_MOSTLY(struct sbi_cpuidle_data, sbi_cpuidle_data);
+static DEFINE_PER_CPU(struct sbi_domain_state, domain_state);
+static bool sbi_cpuidle_use_osi;
+static bool sbi_cpuidle_use_cpuhp;
+static bool sbi_cpuidle_pd_allow_domain_state;
+
+static inline void sbi_set_domain_state(u32 state)
+{
+	struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+	data->available = true;
+	data->state = state;
+}
+
+static inline u32 sbi_get_domain_state(void)
+{
+	struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+	return data->state;
+}
+
+static inline void sbi_clear_domain_state(void)
+{
+	struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+	data->available = false;
+}
+
+static inline bool sbi_is_domain_state_available(void)
+{
+	struct sbi_domain_state *data = this_cpu_ptr(&domain_state);
+
+	return data->available;
+}
+
+static int sbi_suspend_finisher(unsigned long suspend_type,
+				unsigned long resume_addr,
+				unsigned long opaque)
+{
+	struct sbiret ret;
+
+	ret = sbi_ecall(SBI_EXT_HSM, SBI_EXT_HSM_HART_SUSPEND,
+			suspend_type, resume_addr, opaque, 0, 0, 0);
+
+	return (ret.error) ? sbi_err_map_linux_errno(ret.error) : 0;
+}
+
+static int sbi_suspend(u32 state)
+{
+	if (state & SBI_HSM_SUSP_NON_RET_BIT)
+		return cpu_suspend(state, sbi_suspend_finisher);
+	else
+		return sbi_suspend_finisher(state, 0, 0);
+}
+
+static int sbi_cpuidle_enter_state(struct cpuidle_device *dev,
+				   struct cpuidle_driver *drv, int idx)
+{
+	u32 *states = __this_cpu_read(sbi_cpuidle_data.states);
+
+	return CPU_PM_CPU_IDLE_ENTER_PARAM(sbi_suspend, idx, states[idx]);
+}
+
+static int __sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+					  struct cpuidle_driver *drv, int idx,
+					  bool s2idle)
+{
+	struct sbi_cpuidle_data *data = this_cpu_ptr(&sbi_cpuidle_data);
+	u32 *states = data->states;
+	struct device *pd_dev = data->dev;
+	u32 state;
+	int ret;
+
+	ret = cpu_pm_enter();
+	if (ret)
+		return -1;
+
+	/* Do runtime PM to manage a hierarchical CPU toplogy. */
+	rcu_irq_enter_irqson();
+	if (s2idle)
+		dev_pm_genpd_suspend(pd_dev);
+	else
+		pm_runtime_put_sync_suspend(pd_dev);
+	rcu_irq_exit_irqson();
+
+	if (sbi_is_domain_state_available())
+		state = sbi_get_domain_state();
+	else
+		state = states[idx];
+
+	ret = sbi_suspend(state) ? -1 : idx;
+
+	rcu_irq_enter_irqson();
+	if (s2idle)
+		dev_pm_genpd_resume(pd_dev);
+	else
+		pm_runtime_get_sync(pd_dev);
+	rcu_irq_exit_irqson();
+
+	cpu_pm_exit();
+
+	/* Clear the domain state to start fresh when back from idle. */
+	sbi_clear_domain_state();
+	return ret;
+}
+
+static int sbi_enter_domain_idle_state(struct cpuidle_device *dev,
+				       struct cpuidle_driver *drv, int idx)
+{
+	return __sbi_enter_domain_idle_state(dev, drv, idx, false);
+}
+
+static int sbi_enter_s2idle_domain_idle_state(struct cpuidle_device *dev,
+					      struct cpuidle_driver *drv,
+					      int idx)
+{
+	return __sbi_enter_domain_idle_state(dev, drv, idx, true);
+}
+
+static int sbi_cpuidle_cpuhp_up(unsigned int cpu)
+{
+	struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+	if (pd_dev)
+		pm_runtime_get_sync(pd_dev);
+
+	return 0;
+}
+
+static int sbi_cpuidle_cpuhp_down(unsigned int cpu)
+{
+	struct device *pd_dev = __this_cpu_read(sbi_cpuidle_data.dev);
+
+	if (pd_dev) {
+		pm_runtime_put_sync(pd_dev);
+		/* Clear domain state to start fresh at next online. */
+		sbi_clear_domain_state();
+	}
+
+	return 0;
+}
+
+static void sbi_idle_init_cpuhp(void)
+{
+	int err;
+
+	if (!sbi_cpuidle_use_cpuhp)
+		return;
+
+	err = cpuhp_setup_state_nocalls(CPUHP_AP_CPU_PM_STARTING,
+					"cpuidle/sbi:online",
+					sbi_cpuidle_cpuhp_up,
+					sbi_cpuidle_cpuhp_down);
+	if (err)
+		pr_warn("Failed %d while setup cpuhp state\n", err);
+}
+
+static const struct of_device_id sbi_cpuidle_state_match[] = {
+	{ .compatible = "riscv,idle-state",
+	  .data = sbi_cpuidle_enter_state },
+	{ },
+};
+
+static bool sbi_suspend_state_is_valid(u32 state)
+{
+	if (state > SBI_HSM_SUSPEND_RET_DEFAULT &&
+	    state < SBI_HSM_SUSPEND_RET_PLATFORM)
+		return false;
+	if (state > SBI_HSM_SUSPEND_NON_RET_DEFAULT &&
+	    state < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
+		return false;
+	return true;
+}
+
+static int sbi_dt_parse_state_node(struct device_node *np, u32 *state)
+{
+	int err = of_property_read_u32(np, "riscv,sbi-suspend-param", state);
+
+	if (err) {
+		pr_warn("%pOF missing riscv,sbi-suspend-param property\n", np);
+		return err;
+	}
+
+	if (!sbi_suspend_state_is_valid(*state)) {
+		pr_warn("Invalid SBI suspend state %#x\n", *state);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int sbi_dt_cpu_init_topology(struct cpuidle_driver *drv,
+				     struct sbi_cpuidle_data *data,
+				     unsigned int state_count, int cpu)
+{
+	/* Currently limit the hierarchical topology to be used in OSI mode. */
+	if (!sbi_cpuidle_use_osi)
+		return 0;
+
+	data->dev = dt_idle_attach_cpu(cpu, "sbi");
+	if (IS_ERR_OR_NULL(data->dev))
+		return PTR_ERR_OR_ZERO(data->dev);
+
+	/*
+	 * Using the deepest state for the CPU to trigger a potential selection
+	 * of a shared state for the domain, assumes the domain states are all
+	 * deeper states.
+	 */
+	drv->states[state_count - 1].enter = sbi_enter_domain_idle_state;
+	drv->states[state_count - 1].enter_s2idle =
+					sbi_enter_s2idle_domain_idle_state;
+	sbi_cpuidle_use_cpuhp = true;
+
+	return 0;
+}
+
+static int sbi_cpuidle_dt_init_states(struct device *dev,
+					struct cpuidle_driver *drv,
+					unsigned int cpu,
+					unsigned int state_count)
+{
+	struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+	struct device_node *state_node;
+	struct device_node *cpu_node;
+	u32 *states;
+	int i, ret;
+
+	cpu_node = of_cpu_device_node_get(cpu);
+	if (!cpu_node)
+		return -ENODEV;
+
+	states = devm_kcalloc(dev, state_count, sizeof(*states), GFP_KERNEL);
+	if (!states) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+
+	/* Parse SBI specific details from state DT nodes */
+	for (i = 1; i < state_count; i++) {
+		state_node = of_get_cpu_state_node(cpu_node, i - 1);
+		if (!state_node)
+			break;
+
+		ret = sbi_dt_parse_state_node(state_node, &states[i]);
+		of_node_put(state_node);
+
+		if (ret)
+			return ret;
+
+		pr_debug("sbi-state %#x index %d\n", states[i], i);
+	}
+	if (i != state_count) {
+		ret = -ENODEV;
+		goto fail;
+	}
+
+	/* Initialize optional data, used for the hierarchical topology. */
+	ret = sbi_dt_cpu_init_topology(drv, data, state_count, cpu);
+	if (ret < 0)
+		return ret;
+
+	/* Store states in the per-cpu struct. */
+	data->states = states;
+
+fail:
+	of_node_put(cpu_node);
+
+	return ret;
+}
+
+static void sbi_cpuidle_deinit_cpu(int cpu)
+{
+	struct sbi_cpuidle_data *data = per_cpu_ptr(&sbi_cpuidle_data, cpu);
+
+	dt_idle_detach_cpu(data->dev);
+	sbi_cpuidle_use_cpuhp = false;
+}
+
+static int sbi_cpuidle_init_cpu(struct device *dev, int cpu)
+{
+	struct cpuidle_driver *drv;
+	unsigned int state_count = 0;
+	int ret = 0;
+
+	drv = devm_kzalloc(dev, sizeof(*drv), GFP_KERNEL);
+	if (!drv)
+		return -ENOMEM;
+
+	drv->name = "sbi_cpuidle";
+	drv->owner = THIS_MODULE;
+	drv->cpumask = (struct cpumask *)cpumask_of(cpu);
+
+	/* RISC-V architectural WFI to be represented as state index 0. */
+	drv->states[0].enter = sbi_cpuidle_enter_state;
+	drv->states[0].exit_latency = 1;
+	drv->states[0].target_residency = 1;
+	drv->states[0].power_usage = UINT_MAX;
+	strcpy(drv->states[0].name, "WFI");
+	strcpy(drv->states[0].desc, "RISC-V WFI");
+
+	/*
+	 * If no DT idle states are detected (ret == 0) let the driver
+	 * initialization fail accordingly since there is no reason to
+	 * initialize the idle driver if only wfi is supported, the
+	 * default archictectural back-end already executes wfi
+	 * on idle entry.
+	 */
+	ret = dt_init_idle_driver(drv, sbi_cpuidle_state_match, 1);
+	if (ret <= 0) {
+		pr_debug("HART%ld: failed to parse DT idle states\n",
+			 cpuid_to_hartid_map(cpu));
+		return ret ? : -ENODEV;
+	}
+	state_count = ret + 1; /* Include WFI state as well */
+
+	/* Initialize idle states from DT. */
+	ret = sbi_cpuidle_dt_init_states(dev, drv, cpu, state_count);
+	if (ret) {
+		pr_err("HART%ld: failed to init idle states\n",
+		       cpuid_to_hartid_map(cpu));
+		return ret;
+	}
+
+	ret = cpuidle_register(drv, NULL);
+	if (ret)
+		goto deinit;
+
+	cpuidle_cooling_register(drv);
+
+	return 0;
+deinit:
+	sbi_cpuidle_deinit_cpu(cpu);
+	return ret;
+}
+
+static void sbi_cpuidle_domain_sync_state(struct device *dev)
+{
+	/*
+	 * All devices have now been attached/probed to the PM domain
+	 * topology, hence it's fine to allow domain states to be picked.
+	 */
+	sbi_cpuidle_pd_allow_domain_state = true;
+}
+
+#ifdef CONFIG_DT_IDLE_GENPD
+
+static int sbi_cpuidle_pd_power_off(struct generic_pm_domain *pd)
+{
+	struct genpd_power_state *state = &pd->states[pd->state_idx];
+	u32 *pd_state;
+
+	if (!state->data)
+		return 0;
+
+	if (!sbi_cpuidle_pd_allow_domain_state)
+		return -EBUSY;
+
+	/* OSI mode is enabled, set the corresponding domain state. */
+	pd_state = state->data;
+	sbi_set_domain_state(*pd_state);
+
+	return 0;
+}
+
+struct sbi_pd_provider {
+	struct list_head link;
+	struct device_node *node;
+};
+
+static LIST_HEAD(sbi_pd_providers);
+
+static int sbi_pd_init(struct device_node *np)
+{
+	struct generic_pm_domain *pd;
+	struct sbi_pd_provider *pd_provider;
+	struct dev_power_governor *pd_gov;
+	int ret = -ENOMEM, state_count = 0;
+
+	pd = dt_idle_pd_alloc(np, sbi_dt_parse_state_node);
+	if (!pd)
+		goto out;
+
+	pd_provider = kzalloc(sizeof(*pd_provider), GFP_KERNEL);
+	if (!pd_provider)
+		goto free_pd;
+
+	pd->flags |= GENPD_FLAG_IRQ_SAFE | GENPD_FLAG_CPU_DOMAIN;
+
+	/* Allow power off when OSI is available. */
+	if (sbi_cpuidle_use_osi)
+		pd->power_off = sbi_cpuidle_pd_power_off;
+	else
+		pd->flags |= GENPD_FLAG_ALWAYS_ON;
+
+	/* Use governor for CPU PM domains if it has some states to manage. */
+	pd_gov = state_count > 0 ? &pm_domain_cpu_gov : NULL;
+
+	ret = pm_genpd_init(pd, pd_gov, false);
+	if (ret)
+		goto free_pd_prov;
+
+	ret = of_genpd_add_provider_simple(np, pd);
+	if (ret)
+		goto remove_pd;
+
+	pd_provider->node = of_node_get(np);
+	list_add(&pd_provider->link, &sbi_pd_providers);
+
+	pr_debug("init PM domain %s\n", pd->name);
+	return 0;
+
+remove_pd:
+	pm_genpd_remove(pd);
+free_pd_prov:
+	kfree(pd_provider);
+free_pd:
+	dt_idle_pd_free(pd);
+out:
+	pr_err("failed to init PM domain ret=%d %pOF\n", ret, np);
+	return ret;
+}
+
+static void sbi_pd_remove(void)
+{
+	struct sbi_pd_provider *pd_provider, *it;
+	struct generic_pm_domain *genpd;
+
+	list_for_each_entry_safe(pd_provider, it, &sbi_pd_providers, link) {
+		of_genpd_del_provider(pd_provider->node);
+
+		genpd = of_genpd_remove_last(pd_provider->node);
+		if (!IS_ERR(genpd))
+			kfree(genpd);
+
+		of_node_put(pd_provider->node);
+		list_del(&pd_provider->link);
+		kfree(pd_provider);
+	}
+}
+
+static int sbi_genpd_probe(struct device_node *np)
+{
+	struct device_node *node;
+	int ret = 0, pd_count = 0;
+
+	if (!np)
+		return -ENODEV;
+
+	/*
+	 * Parse child nodes for the "#power-domain-cells" property and
+	 * initialize a genpd/genpd-of-provider pair when it's found.
+	 */
+	for_each_child_of_node(np, node) {
+		if (!of_find_property(node, "#power-domain-cells", NULL))
+			continue;
+
+		ret = sbi_pd_init(node);
+		if (ret)
+			goto put_node;
+
+		pd_count++;
+	}
+
+	/* Bail out if not using the hierarchical CPU topology. */
+	if (!pd_count)
+		goto no_pd;
+
+	/* Link genpd masters/subdomains to model the CPU topology. */
+	ret = dt_idle_pd_init_topology(np);
+	if (ret)
+		goto remove_pd;
+
+	return 0;
+
+put_node:
+	of_node_put(node);
+remove_pd:
+	sbi_pd_remove();
+	pr_err("failed to create CPU PM domains ret=%d\n", ret);
+no_pd:
+	return ret;
+}
+
+#else
+
+static inline int sbi_genpd_probe(struct device_node *np)
+{
+	return 0;
+}
+
+#endif
+
+static int sbi_cpuidle_probe(struct platform_device *pdev)
+{
+	int cpu, ret;
+	struct cpuidle_driver *drv;
+	struct cpuidle_device *dev;
+	struct device_node *np, *pds_node;
+
+	/* Detect OSI support based on CPU DT nodes */
+	sbi_cpuidle_use_osi = true;
+	for_each_possible_cpu(cpu) {
+		np = of_cpu_device_node_get(cpu);
+		if (np &&
+		    of_find_property(np, "power-domains", NULL) &&
+		    of_find_property(np, "power-domain-names", NULL)) {
+			continue;
+		} else {
+			sbi_cpuidle_use_osi = false;
+			break;
+		}
+	}
+
+	/* Populate generic power domains from DT nodes */
+	pds_node = of_find_node_by_path("/cpus/power-domains");
+	if (pds_node) {
+		ret = sbi_genpd_probe(pds_node);
+		of_node_put(pds_node);
+		if (ret)
+			return ret;
+	}
+
+	/* Initialize CPU idle driver for each CPU */
+	for_each_possible_cpu(cpu) {
+		ret = sbi_cpuidle_init_cpu(&pdev->dev, cpu);
+		if (ret) {
+			pr_debug("HART%ld: idle driver init failed\n",
+				 cpuid_to_hartid_map(cpu));
+			goto out_fail;
+		}
+	}
+
+	/* Setup CPU hotplut notifiers */
+	sbi_idle_init_cpuhp();
+
+	pr_info("idle driver registered for all CPUs\n");
+
+	return 0;
+
+out_fail:
+	while (--cpu >= 0) {
+		dev = per_cpu(cpuidle_devices, cpu);
+		drv = cpuidle_get_cpu_driver(dev);
+		cpuidle_unregister(drv);
+		sbi_cpuidle_deinit_cpu(cpu);
+	}
+
+	return ret;
+}
+
+static struct platform_driver sbi_cpuidle_driver = {
+	.probe = sbi_cpuidle_probe,
+	.driver = {
+		.name = "sbi-cpuidle",
+		.sync_state = sbi_cpuidle_domain_sync_state,
+	},
+};
+
+static int __init sbi_cpuidle_init(void)
+{
+	int ret;
+	struct platform_device *pdev;
+
+	/*
+	 * The SBI HSM suspend function is only available when:
+	 * 1) SBI version is 0.3 or higher
+	 * 2) SBI HSM extension is available
+	 */
+	if ((sbi_spec_version < sbi_mk_version(0, 3)) ||
+	    sbi_probe_extension(SBI_EXT_HSM) <= 0) {
+		pr_info("HSM suspend not available\n");
+		return 0;
+	}
+
+	ret = platform_driver_register(&sbi_cpuidle_driver);
+	if (ret)
+		return ret;
+
+	pdev = platform_device_register_simple("sbi-cpuidle",
+						-1, NULL, 0);
+	if (IS_ERR(pdev)) {
+		platform_driver_unregister(&sbi_cpuidle_driver);
+		return PTR_ERR(pdev);
+	}
+
+	return 0;
+}
+device_initcall(sbi_cpuidle_init);