[libcpu/aarch64] add gicv3 support and bsp/rockchip/rk3568 (#5722)

* [libcpu/aarch64] add smp support

* [libcpu/aarch64] rt_hw_trap_irq get irq instead of iar when using gicv2

* [libcpu/aarch64] disable irq/fiq when switch thread

* [libcpu/aarch64] add gtimer frq set and stack align

* [libcpu/aarch64] add gicv3 support and bsp/rockchip/rk3568

9 files changed, 1024 insertions, 55 deletions

diff --git a/libcpu/aarch64/common/SConscript b/libcpu/aarch64/common/SConscript
index 915b9d29aa..72fb8e26c3 100644
--- a/libcpu/aarch64/common/SConscript
+++ b/libcpu/aarch64/common/SConscript
@@ -6,10 +6,6 @@ Import('rtconfig')
 
 cwd     = GetCurrentDir()
 src     = Glob('*.c') + Glob('*.cpp') + Glob('*.S')
-
-if not GetDepend('BSP_USING_GIC'):
-    SrcRemove(src, 'gic.c')
-
 CPPPATH = [cwd]
 
 group = DefineGroup('common', src, depend = [''], CPPPATH = CPPPATH)
diff --git a/libcpu/aarch64/common/gic.c b/libcpu/aarch64/common/gic.c
index 7a2c368d23..95a710c480 100644
--- a/libcpu/aarch64/common/gic.c
+++ b/libcpu/aarch64/common/gic.c
@@ -9,20 +9,25 @@
  * 2014-04-03     Grissiom     many enhancements
  * 2018-11-22     Jesven       add rt_hw_ipi_send()
  *                             add rt_hw_ipi_handler_install()
+ * 2022-03-08     GuEe-GUI     add BSP bind SPI CPU self support
  */
 
 #include <rthw.h>
 #include <rtthread.h>
 
+#if defined(BSP_USING_GIC) && defined(BSP_USING_GICV2)
+
 #include <gic.h>
 #include <cpuport.h>
 
+#include <board.h>
+
 struct arm_gic
 {
     rt_uint64_t offset;         /* the first interrupt index in the vector table */
 
     rt_uint64_t dist_hw_base;   /* the base address of the gic distributor */
-    rt_uint64_t cpu_hw_base;    /* the base addrees of the gic cpu interface */
+    rt_uint64_t cpu_hw_base;    /* the base address of the gic cpu interface */
 };
 
 /* 'ARM_GIC_MAX_NR' is the number of cores */
@@ -30,33 +35,33 @@ static struct arm_gic _gic_table[ARM_GIC_MAX_NR];
 
 /** Macro to access the Generic Interrupt Controller Interface (GICC)
 */
-#define GIC_CPU_CTRL(hw_base)               __REG32((hw_base) + 0x00U)
-#define GIC_CPU_PRIMASK(hw_base)            __REG32((hw_base) + 0x04U)
-#define GIC_CPU_BINPOINT(hw_base)           __REG32((hw_base) + 0x08U)
-#define GIC_CPU_INTACK(hw_base)             __REG32((hw_base) + 0x0cU)
-#define GIC_CPU_EOI(hw_base)                __REG32((hw_base) + 0x10U)
-#define GIC_CPU_RUNNINGPRI(hw_base)         __REG32((hw_base) + 0x14U)
-#define GIC_CPU_HIGHPRI(hw_base)            __REG32((hw_base) + 0x18U)
-#define GIC_CPU_IIDR(hw_base)               __REG32((hw_base) + 0xFCU)
+#define GIC_CPU_CTRL(hw_base)               HWREG32((hw_base) + 0x00U)
+#define GIC_CPU_PRIMASK(hw_base)            HWREG32((hw_base) + 0x04U)
+#define GIC_CPU_BINPOINT(hw_base)           HWREG32((hw_base) + 0x08U)
+#define GIC_CPU_INTACK(hw_base)             HWREG32((hw_base) + 0x0cU)
+#define GIC_CPU_EOI(hw_base)                HWREG32((hw_base) + 0x10U)
+#define GIC_CPU_RUNNINGPRI(hw_base)         HWREG32((hw_base) + 0x14U)
+#define GIC_CPU_HIGHPRI(hw_base)            HWREG32((hw_base) + 0x18U)
+#define GIC_CPU_IIDR(hw_base)               HWREG32((hw_base) + 0xFCU)
 
 /** Macro to access the Generic Interrupt Controller Distributor (GICD)
 */
-#define GIC_DIST_CTRL(hw_base)              __REG32((hw_base) + 0x000U)
-#define GIC_DIST_TYPE(hw_base)              __REG32((hw_base) + 0x004U)
-#define GIC_DIST_IGROUP(hw_base, n)         __REG32((hw_base) + 0x080U + ((n)/32U) * 4U)
-#define GIC_DIST_ENABLE_SET(hw_base, n)     __REG32((hw_base) + 0x100U + ((n)/32U) * 4U)
-#define GIC_DIST_ENABLE_CLEAR(hw_base, n)   __REG32((hw_base) + 0x180U + ((n)/32U) * 4U)
-#define GIC_DIST_PENDING_SET(hw_base, n)    __REG32((hw_base) + 0x200U + ((n)/32U) * 4U)
-#define GIC_DIST_PENDING_CLEAR(hw_base, n)  __REG32((hw_base) + 0x280U + ((n)/32U) * 4U)
-#define GIC_DIST_ACTIVE_SET(hw_base, n)     __REG32((hw_base) + 0x300U + ((n)/32U) * 4U)
-#define GIC_DIST_ACTIVE_CLEAR(hw_base, n)   __REG32((hw_base) + 0x380U + ((n)/32U) * 4U)
-#define GIC_DIST_PRI(hw_base, n)            __REG32((hw_base) + 0x400U +  ((n)/4U) * 4U)
-#define GIC_DIST_TARGET(hw_base, n)         __REG32((hw_base) + 0x800U +  ((n)/4U) * 4U)
-#define GIC_DIST_CONFIG(hw_base, n)         __REG32((hw_base) + 0xc00U + ((n)/16U) * 4U)
-#define GIC_DIST_SOFTINT(hw_base)           __REG32((hw_base) + 0xf00U)
-#define GIC_DIST_CPENDSGI(hw_base, n)       __REG32((hw_base) + 0xf10U + ((n)/4U) * 4U)
-#define GIC_DIST_SPENDSGI(hw_base, n)       __REG32((hw_base) + 0xf20U + ((n)/4U) * 4U)
-#define GIC_DIST_ICPIDR2(hw_base)           __REG32((hw_base) + 0xfe8U)
+#define GIC_DIST_CTRL(hw_base)              HWREG32((hw_base) + 0x000U)
+#define GIC_DIST_TYPE(hw_base)              HWREG32((hw_base) + 0x004U)
+#define GIC_DIST_IGROUP(hw_base, n)         HWREG32((hw_base) + 0x080U + ((n)/32U) * 4U)
+#define GIC_DIST_ENABLE_SET(hw_base, n)     HWREG32((hw_base) + 0x100U + ((n)/32U) * 4U)
+#define GIC_DIST_ENABLE_CLEAR(hw_base, n)   HWREG32((hw_base) + 0x180U + ((n)/32U) * 4U)
+#define GIC_DIST_PENDING_SET(hw_base, n)    HWREG32((hw_base) + 0x200U + ((n)/32U) * 4U)
+#define GIC_DIST_PENDING_CLEAR(hw_base, n)  HWREG32((hw_base) + 0x280U + ((n)/32U) * 4U)
+#define GIC_DIST_ACTIVE_SET(hw_base, n)     HWREG32((hw_base) + 0x300U + ((n)/32U) * 4U)
+#define GIC_DIST_ACTIVE_CLEAR(hw_base, n)   HWREG32((hw_base) + 0x380U + ((n)/32U) * 4U)
+#define GIC_DIST_PRI(hw_base, n)            HWREG32((hw_base) + 0x400U +  ((n)/4U) * 4U)
+#define GIC_DIST_TARGET(hw_base, n)         HWREG32((hw_base) + 0x800U +  ((n)/4U) * 4U)
+#define GIC_DIST_CONFIG(hw_base, n)         HWREG32((hw_base) + 0xc00U + ((n)/16U) * 4U)
+#define GIC_DIST_SOFTINT(hw_base)           HWREG32((hw_base) + 0xf00U)
+#define GIC_DIST_CPENDSGI(hw_base, n)       HWREG32((hw_base) + 0xf10U + ((n)/4U) * 4U)
+#define GIC_DIST_SPENDSGI(hw_base, n)       HWREG32((hw_base) + 0xf20U + ((n)/4U) * 4U)
+#define GIC_DIST_ICPIDR2(hw_base)           HWREG32((hw_base) + 0xfe8U)
 
 static unsigned int _gic_max_irq;
 
@@ -179,7 +184,7 @@ void arm_gic_clear_pending_irq(rt_uint64_t index, int irq)
     }
 }
 
-void arm_gic_set_configuration(rt_uint64_t index, int irq, uint32_t config)
+void arm_gic_set_configuration(rt_uint64_t index, int irq, rt_uint32_t config)
 {
     rt_uint64_t icfgr;
     rt_uint64_t shift;
@@ -323,6 +328,8 @@ void arm_gic_send_sgi(rt_uint64_t index, int irq, rt_uint64_t target_list, rt_ui
 
     GIC_DIST_SOFTINT(_gic_table[index].dist_hw_base) =
         ((filter_list & 0x3U) << 24U) | ((target_list & 0xFFUL) << 16U) | (irq & 0x0FUL);
+
+    __DSB();
 }
 
 rt_uint64_t arm_gic_get_high_pending_irq(rt_uint64_t index)
@@ -341,8 +348,8 @@ rt_uint64_t arm_gic_get_interface_id(rt_uint64_t index)
 
 void arm_gic_set_group(rt_uint64_t index, int irq, rt_uint64_t group)
 {
-    uint32_t igroupr;
-    uint32_t shift;
+    rt_uint32_t igroupr;
+    rt_uint32_t shift;
 
     RT_ASSERT(index < ARM_GIC_MAX_NR);
     RT_ASSERT(group <= 1U);
@@ -373,6 +380,10 @@ int arm_gic_dist_init(rt_uint64_t index, rt_uint64_t dist_base, int irq_start)
     unsigned int gic_type, i;
     rt_uint64_t cpumask = 1U << 0U;
 
+#ifdef ARM_SPI_BIND_CPU_ID
+    cpumask = 1U << ARM_SPI_BIND_CPU_ID;
+#endif
+
     RT_ASSERT(index < ARM_GIC_MAX_NR);
 
     _gic_table[index].dist_hw_base = dist_base;
@@ -502,3 +513,4 @@ long gic_dump(void)
 }
 MSH_CMD_EXPORT(gic_dump, show gic status);
 
+#endif /* defined(BSP_USING_GIC) && defined(BSP_USING_GICV2) */
diff --git a/libcpu/aarch64/common/gic.h b/libcpu/aarch64/common/gic.h
index 54ab2bda18..7ea0d6d3af 100644
--- a/libcpu/aarch64/common/gic.h
+++ b/libcpu/aarch64/common/gic.h
@@ -11,9 +11,9 @@
 #ifndef __GIC_H__
 #define __GIC_H__
 
-#include <rthw.h>
-#include <stdint.h>
-#include <board.h>
+#include <rtdef.h>
+
+#if defined(BSP_USING_GIC) && defined(BSP_USING_GICV2)
 
 int arm_gic_get_active_irq(rt_uint64_t index);
 void arm_gic_ack(rt_uint64_t index, int irq);
@@ -25,7 +25,7 @@ rt_uint64_t arm_gic_get_pending_irq(rt_uint64_t index, int irq);
 void arm_gic_set_pending_irq(rt_uint64_t index, int irq);
 void arm_gic_clear_pending_irq(rt_uint64_t index, int irq);
 
-void arm_gic_set_configuration(rt_uint64_t index, int irq, uint32_t config);
+void arm_gic_set_configuration(rt_uint64_t index, int irq, rt_uint32_t config);
 rt_uint64_t arm_gic_get_configuration(rt_uint64_t index, int irq);
 
 void arm_gic_clear_active(rt_uint64_t index, int irq);
@@ -59,5 +59,7 @@ int arm_gic_cpu_init(rt_uint64_t index, rt_uint64_t cpu_base);
 void arm_gic_dump_type(rt_uint64_t index);
 void arm_gic_dump(rt_uint64_t index);
 
+#endif /* defined(BSP_USING_GIC) && defined(BSP_USING_GICV2) */
+
 #endif
 
diff --git a/libcpu/aarch64/common/gicv3.c b/libcpu/aarch64/common/gicv3.c
new file mode 100644
index 0000000000..67010f1d57
--- /dev/null
+++ b/libcpu/aarch64/common/gicv3.c
@@ -0,0 +1,849 @@
+/*
+ * Copyright (c) 2006-2022, RT-Thread Development Team
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ *
+ * Change Logs:
+ * Date           Author       Notes
+ * 2013-07-20     Bernard      first version
+ * 2014-04-03     Grissiom     many enhancements
+ * 2018-11-22     Jesven       add rt_hw_ipi_send()
+ *                             add rt_hw_ipi_handler_install()
+ * 2022-03-08     GuEe-GUI     add BSP bind SPI CPU self support
+ *                             add GICv3 AArch64 system register interface
+ *                             modify arm_gic_redist_init() args
+ *                             modify arm_gic_cpu_init() args
+ *                             modify arm_gic_send_affinity_sgi() args
+ *                             remove arm_gic_redist_address_set()
+ *                             remove arm_gic_cpu_interface_address_set()
+ *                             remove arm_gic_secondary_cpu_init()
+ *                             remove get_main_cpu_affval()
+ *                             remove arm_gic_cpumask_to_affval()
+ */
+
+#include <rthw.h>
+#include <rtthread.h>
+
+#if defined(BSP_USING_GIC) && defined(BSP_USING_GICV3)
+
+#include <gicv3.h>
+#include <cpuport.h>
+
+#include <board.h>
+
+#ifndef ARM_SPI_BIND_CPU_ID
+#define ARM_SPI_BIND_CPU_ID 0
+#endif
+
+#ifndef RT_USING_SMP
+#define RT_CPUS_NR 1
+extern int rt_hw_cpu_id(void);
+#else
+extern rt_uint64_t rt_cpu_mpidr_early[];
+#endif /* RT_USING_SMP */
+
+struct arm_gic
+{
+    rt_uint64_t offset;                     /* the first interrupt index in the vector table */
+    rt_uint64_t redist_hw_base[RT_CPUS_NR]; /* the pointer of the gic redistributor */
+    rt_uint64_t dist_hw_base;               /* the base address of the gic distributor */
+    rt_uint64_t cpu_hw_base[RT_CPUS_NR];    /* the base address of the gic cpu interface */
+};
+
+/* 'ARM_GIC_MAX_NR' is the number of cores */
+static struct arm_gic _gic_table[ARM_GIC_MAX_NR];
+static unsigned int _gic_max_irq;
+
+#define GET_GICV3_REG(reg, out) __asm__ volatile ("mrs %0, " reg:"=r"(out)::"memory");
+#define SET_GICV3_REG(reg, in)  __asm__ volatile ("msr " reg ", %0"::"r"(in):"memory");
+
+/* AArch64 System register interface to GICv3 */
+#define ICC_IAR0_EL1    "S3_0_C12_C8_0"
+#define ICC_IAR1_EL1    "S3_0_C12_C12_0"
+#define ICC_EOIR0_EL1   "S3_0_C12_C8_1"
+#define ICC_EOIR1_EL1   "S3_0_C12_C12_1"
+#define ICC_HPPIR0_EL1  "S3_0_C12_C8_2"
+#define ICC_HPPIR1_EL1  "S3_0_C12_C12_2"
+#define ICC_BPR0_EL1    "S3_0_C12_C8_3"
+#define ICC_BPR1_EL1    "S3_0_C12_C12_3"
+#define ICC_DIR_EL1     "S3_0_C12_C11_1"
+#define ICC_PMR_EL1     "S3_0_C4_C6_0"
+#define ICC_RPR_EL1     "S3_0_C12_C11_3"
+#define ICC_CTLR_EL1    "S3_0_C12_C12_4"
+#define ICC_CTLR_EL3    "S3_6_C12_C12_4"
+#define ICC_SRE_EL1     "S3_0_C12_C12_5"
+#define ICC_SRE_EL2     "S3_4_C12_C9_5"
+#define ICC_SRE_EL3     "S3_6_C12_C12_5"
+#define ICC_IGRPEN0_EL1 "S3_0_C12_C12_6"
+#define ICC_IGRPEN1_EL1 "S3_0_C12_C12_7"
+#define ICC_IGRPEN1_EL3 "S3_6_C12_C12_7"
+#define ICC_SGI0R_EL1   "S3_0_C12_C11_7"
+#define ICC_SGI1R_EL1   "S3_0_C12_C11_5"
+#define ICC_ASGI1R_EL1  "S3_0_C12_C11_6"
+
+/* Macro to access the Distributor Control Register (GICD_CTLR) */
+#define GICD_CTLR_RWP       (1 << 31)
+#define GICD_CTLR_E1NWF     (1 << 7)
+#define GICD_CTLR_DS        (1 << 6)
+#define GICD_CTLR_ARE_NS    (1 << 5)
+#define GICD_CTLR_ARE_S     (1 << 4)
+#define GICD_CTLR_ENGRP1S   (1 << 2)
+#define GICD_CTLR_ENGRP1NS  (1 << 1)
+#define GICD_CTLR_ENGRP0    (1 << 0)
+
+/* Macro to access the Redistributor Control Register (GICR_CTLR) */
+#define GICR_CTLR_UWP       (1 << 31)
+#define GICR_CTLR_DPG1S     (1 << 26)
+#define GICR_CTLR_DPG1NS    (1 << 25)
+#define GICR_CTLR_DPG0      (1 << 24)
+#define GICR_CTLR_RWP       (1 << 3)
+#define GICR_CTLR_IR        (1 << 2)
+#define GICR_CTLR_CES       (1 << 1)
+#define GICR_CTLR_EnableLPI (1 << 0)
+
+/* Macro to access the Generic Interrupt Controller Interface (GICC) */
+#define GIC_CPU_CTRL(hw_base)               HWREG32((hw_base) + 0x00U)
+#define GIC_CPU_PRIMASK(hw_base)            HWREG32((hw_base) + 0x04U)
+#define GIC_CPU_BINPOINT(hw_base)           HWREG32((hw_base) + 0x08U)
+#define GIC_CPU_INTACK(hw_base)             HWREG32((hw_base) + 0x0cU)
+#define GIC_CPU_EOI(hw_base)                HWREG32((hw_base) + 0x10U)
+#define GIC_CPU_RUNNINGPRI(hw_base)         HWREG32((hw_base) + 0x14U)
+#define GIC_CPU_HIGHPRI(hw_base)            HWREG32((hw_base) + 0x18U)
+#define GIC_CPU_IIDR(hw_base)               HWREG32((hw_base) + 0xFCU)
+
+/* Macro to access the Generic Interrupt Controller Distributor (GICD) */
+#define GIC_DIST_CTRL(hw_base)              HWREG32((hw_base) + 0x000U)
+#define GIC_DIST_TYPE(hw_base)              HWREG32((hw_base) + 0x004U)
+#define GIC_DIST_IGROUP(hw_base, n)         HWREG32((hw_base) + 0x080U + ((n) / 32U) * 4U)
+#define GIC_DIST_ENABLE_SET(hw_base, n)     HWREG32((hw_base) + 0x100U + ((n) / 32U) * 4U)
+#define GIC_DIST_ENABLE_CLEAR(hw_base, n)   HWREG32((hw_base) + 0x180U + ((n) / 32U) * 4U)
+#define GIC_DIST_PENDING_SET(hw_base, n)    HWREG32((hw_base) + 0x200U + ((n) / 32U) * 4U)
+#define GIC_DIST_PENDING_CLEAR(hw_base, n)  HWREG32((hw_base) + 0x280U + ((n) / 32U) * 4U)
+#define GIC_DIST_ACTIVE_SET(hw_base, n)     HWREG32((hw_base) + 0x300U + ((n) / 32U) * 4U)
+#define GIC_DIST_ACTIVE_CLEAR(hw_base, n)   HWREG32((hw_base) + 0x380U + ((n) / 32U) * 4U)
+#define GIC_DIST_PRI(hw_base, n)            HWREG32((hw_base) + 0x400U + ((n) / 4U) * 4U)
+#define GIC_DIST_TARGET(hw_base, n)         HWREG32((hw_base) + 0x800U + ((n) / 4U) * 4U)
+#define GIC_DIST_CONFIG(hw_base, n)         HWREG32((hw_base) + 0xc00U + ((n) / 16U) * 4U)
+#define GIC_DIST_SOFTINT(hw_base)           HWREG32((hw_base) + 0xf00U)
+#define GIC_DIST_CPENDSGI(hw_base, n)       HWREG32((hw_base) + 0xf10U + ((n) / 4U) * 4U)
+#define GIC_DIST_SPENDSGI(hw_base, n)       HWREG32((hw_base) + 0xf20U + ((n) / 4U) * 4U)
+#define GIC_DIST_ICPIDR2(hw_base)           HWREG32((hw_base) + 0xfe8U)
+#define GIC_DIST_IROUTER(hw_base, n)        HWREG64((hw_base) + 0x6000U + (n) * 8U)
+
+/* SGI base address is at 64K offset from Redistributor base address */
+#define GIC_RSGI_OFFSET 0x10000
+
+/* Macro to access the Generic Interrupt Controller Redistributor (GICR) */
+#define GIC_RDIST_CTRL(hw_base)             HWREG32((hw_base) + 0x000U)
+#define GIC_RDIST_IIDR(hw_base)             HWREG32((hw_base) + 0x004U)
+#define GIC_RDIST_TYPER(hw_base)            HWREG64((hw_base) + 0x008U)
+#define GIC_RDIST_TSTATUSR(hw_base)         HWREG32((hw_base) + 0x010U)
+#define GIC_RDIST_WAKER(hw_base)            HWREG32((hw_base) + 0x014U)
+#define GIC_RDIST_SETLPIR(hw_base)          HWREG32((hw_base) + 0x040U)
+#define GIC_RDIST_CLRLPIR(hw_base)          HWREG32((hw_base) + 0x048U)
+#define GIC_RDIST_PROPBASER(hw_base)        HWREG32((hw_base) + 0x070U)
+#define GIC_RDIST_PENDBASER(hw_base)        HWREG32((hw_base) + 0x078U)
+#define GIC_RDIST_INVLPIR(hw_base)          HWREG32((hw_base) + 0x0A0U)
+#define GIC_RDIST_INVALLR(hw_base)          HWREG32((hw_base) + 0x0B0U)
+#define GIC_RDIST_SYNCR(hw_base)            HWREG32((hw_base) + 0x0C0U)
+
+#define GIC_RDISTSGI_IGROUPR0(hw_base, n)   HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x080U + (n) * 4U)
+#define GIC_RDISTSGI_ISENABLER0(hw_base)    HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x100U)
+#define GIC_RDISTSGI_ICENABLER0(hw_base)    HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x180U)
+#define GIC_RDISTSGI_ISPENDR0(hw_base)      HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x200U)
+#define GIC_RDISTSGI_ICPENDR0(hw_base)      HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x280U)
+#define GIC_RDISTSGI_ISACTIVER0(hw_base)    HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x300U)
+#define GIC_RDISTSGI_ICACTIVER0(hw_base)    HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x380U)
+#define GIC_RDISTSGI_IPRIORITYR(hw_base, n) HWREG32((hw_base) + GIC_RSGI_OFFSET + 0x400U + ((n) / 4U) * 4U)
+#define GIC_RDISTSGI_ICFGR0(hw_base)        HWREG32((hw_base) + GIC_RSGI_OFFSET + 0xC00U)
+#define GIC_RDISTSGI_ICFGR1(hw_base)        HWREG32((hw_base) + GIC_RSGI_OFFSET + 0xC04U)
+#define GIC_RDISTSGI_IGRPMODR0(hw_base, n)  HWREG32((hw_base) + GIC_RSGI_OFFSET + 0xD00U + (n) * 4)
+#define GIC_RDISTSGI_NSACR(hw_base)         HWREG32((hw_base) + GIC_RSGI_OFFSET + 0xE00U)
+
+int arm_gic_get_active_irq(rt_uint64_t index)
+{
+    int irq;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    GET_GICV3_REG(ICC_IAR1_EL1, irq);
+
+    irq = (irq & 0x1ffffff) + _gic_table[index].offset;
+    return irq;
+}
+
+void arm_gic_ack(rt_uint64_t index, int irq)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+    RT_ASSERT(irq >= 0);
+
+    __DSB();
+    SET_GICV3_REG(ICC_EOIR1_EL1, irq);
+}
+
+void arm_gic_mask(rt_uint64_t index, int irq)
+{
+    rt_uint64_t mask = 1 << (irq % 32);
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq < 32)
+    {
+        rt_int32_t cpu_id = rt_hw_cpu_id();
+
+        GIC_RDISTSGI_ICENABLER0(_gic_table[index].redist_hw_base[cpu_id]) = mask;
+    }
+    else
+    {
+        GIC_DIST_ENABLE_CLEAR(_gic_table[index].dist_hw_base, irq) = mask;
+    }
+}
+
+void arm_gic_umask(rt_uint64_t index, int irq)
+{
+    rt_uint64_t mask = 1 << (irq % 32);
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq < 32)
+    {
+        rt_int32_t cpu_id = rt_hw_cpu_id();
+
+        GIC_RDISTSGI_ISENABLER0(_gic_table[index].redist_hw_base[cpu_id]) = mask;
+    }
+    else
+    {
+        GIC_DIST_ENABLE_SET(_gic_table[index].dist_hw_base, irq) = mask;
+    }
+}
+
+rt_uint64_t arm_gic_get_pending_irq(rt_uint64_t index, int irq)
+{
+    rt_uint64_t pend;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq >= 16)
+    {
+        pend = (GIC_DIST_PENDING_SET(_gic_table[index].dist_hw_base, irq) >> (irq % 32)) & 0x1;
+    }
+    else
+    {
+        /* INTID 0-15 Software Generated Interrupt */
+        pend = (GIC_DIST_SPENDSGI(_gic_table[index].dist_hw_base, irq) >> ((irq % 4) * 8)) & 0xff;
+        /* No CPU identification offered */
+        if (pend != 0)
+        {
+            pend = 1;
+        }
+        else
+        {
+            pend = 0;
+        }
+    }
+
+    return pend;
+}
+
+void arm_gic_set_pending_irq(rt_uint64_t index, int irq)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq >= 16)
+    {
+        GIC_DIST_PENDING_SET(_gic_table[index].dist_hw_base, irq) = 1 << (irq % 32);
+    }
+    else
+    {
+        /* INTID 0-15 Software Generated Interrupt */
+        /* Forward the interrupt to the CPU interface that requested it */
+        GIC_DIST_SOFTINT(_gic_table[index].dist_hw_base) = (irq | 0x02000000);
+    }
+}
+
+void arm_gic_clear_pending_irq(rt_uint64_t index, int irq)
+{
+    rt_uint64_t mask;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq >= 16)
+    {
+        mask = 1 << (irq % 32);
+        GIC_DIST_PENDING_CLEAR(_gic_table[index].dist_hw_base, irq) = mask;
+    }
+    else
+    {
+        mask = 1 << ((irq % 4) * 8);
+        GIC_DIST_CPENDSGI(_gic_table[index].dist_hw_base, irq) = mask;
+    }
+}
+
+void arm_gic_set_configuration(rt_uint64_t index, int irq, rt_uint32_t config)
+{
+    rt_uint64_t icfgr;
+    rt_uint64_t shift;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    icfgr = GIC_DIST_CONFIG(_gic_table[index].dist_hw_base, irq);
+    shift = (irq % 16) << 1;
+
+    icfgr &= (~(3 << shift));
+    icfgr |= (config << shift);
+
+    GIC_DIST_CONFIG(_gic_table[index].dist_hw_base, irq) = icfgr;
+}
+
+rt_uint64_t arm_gic_get_configuration(rt_uint64_t index, int irq)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    return (GIC_DIST_CONFIG(_gic_table[index].dist_hw_base, irq) >> ((irq % 16) >> 1));
+}
+
+void arm_gic_clear_active(rt_uint64_t index, int irq)
+{
+    rt_uint64_t mask = 1 << (irq % 32);
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    GIC_DIST_ACTIVE_CLEAR(_gic_table[index].dist_hw_base, irq) = mask;
+}
+
+/* Set up the cpu mask for the specific interrupt */
+void arm_gic_set_cpu(rt_uint64_t index, int irq, unsigned int cpumask)
+{
+    rt_uint64_t old_tgt;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    old_tgt = GIC_DIST_TARGET(_gic_table[index].dist_hw_base, irq);
+
+    old_tgt &= ~(0x0ff << ((irq % 4) * 8));
+    old_tgt |= cpumask << ((irq % 4) * 8);
+
+    GIC_DIST_TARGET(_gic_table[index].dist_hw_base, irq) = old_tgt;
+}
+
+rt_uint64_t arm_gic_get_target_cpu(rt_uint64_t index, int irq)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    return (GIC_DIST_TARGET(_gic_table[index].dist_hw_base, irq) >> ((irq % 4) * 8)) & 0xff;
+}
+
+void arm_gic_set_priority(rt_uint64_t index, int irq, rt_uint64_t priority)
+{
+    rt_uint64_t mask;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq < 32)
+    {
+        rt_int32_t cpu_id = rt_hw_cpu_id();
+
+        mask = GIC_RDISTSGI_IPRIORITYR(_gic_table[index].redist_hw_base[cpu_id], irq);
+        mask &= ~(0xff << ((irq % 4) * 8));
+        mask |= ((priority & 0xff) << ((irq % 4) * 8));
+        GIC_RDISTSGI_IPRIORITYR(_gic_table[index].redist_hw_base[cpu_id], irq) = mask;
+    }
+    else
+    {
+        mask = GIC_DIST_PRI(_gic_table[index].dist_hw_base, irq);
+        mask &= ~(0xff << ((irq % 4) * 8));
+        mask |= ((priority & 0xff) << ((irq % 4) * 8));
+        GIC_DIST_PRI(_gic_table[index].dist_hw_base, irq) = mask;
+    }
+}
+
+rt_uint64_t arm_gic_get_priority(rt_uint64_t index, int irq)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    if (irq < 32)
+    {
+        rt_int32_t cpu_id = rt_hw_cpu_id();
+
+        return (GIC_RDISTSGI_IPRIORITYR(_gic_table[index].redist_hw_base[cpu_id], irq) >> ((irq % 4) * 8)) & 0xff;
+    }
+    else
+    {
+        return (GIC_DIST_PRI(_gic_table[index].dist_hw_base, irq) >> ((irq % 4) * 8)) & 0xff;
+    }
+}
+
+void arm_gic_set_system_register_enable_mask(rt_uint64_t index, rt_uint64_t value)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    value &= 0xff;
+    /* set priority mask */
+    SET_GICV3_REG(ICC_SRE_EL1, value);
+    __ISB();
+}
+
+rt_uint64_t arm_gic_get_system_register_enable_mask(rt_uint64_t index)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+    rt_uint64_t value;
+
+    GET_GICV3_REG(ICC_SRE_EL1, value);
+    return value;
+}
+
+void arm_gic_set_interface_prior_mask(rt_uint64_t index, rt_uint64_t priority)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    priority &= 0xff;
+    /* set priority mask */
+    SET_GICV3_REG(ICC_PMR_EL1, priority);
+}
+
+rt_uint64_t arm_gic_get_interface_prior_mask(rt_uint64_t index)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+    rt_uint64_t priority;
+
+    GET_GICV3_REG(ICC_PMR_EL1, priority);
+    return priority;
+}
+
+void arm_gic_set_binary_point(rt_uint64_t index, rt_uint64_t binary_point)
+{
+    index = index;
+    binary_point &= 0x7;
+
+    SET_GICV3_REG(ICC_BPR1_EL1, binary_point);
+}
+
+rt_uint64_t arm_gic_get_binary_point(rt_uint64_t index)
+{
+    rt_uint64_t binary_point;
+
+    index = index;
+    GET_GICV3_REG(ICC_BPR1_EL1, binary_point);
+    return binary_point;
+}
+
+rt_uint64_t arm_gic_get_irq_status(rt_uint64_t index, int irq)
+{
+    rt_uint64_t pending, active;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    active = (GIC_DIST_ACTIVE_SET(_gic_table[index].dist_hw_base, irq) >> (irq % 32)) & 0x1;
+    pending = (GIC_DIST_PENDING_SET(_gic_table[index].dist_hw_base, irq) >> (irq % 32)) & 0x1;
+
+    return ((active << 1) | pending);
+}
+
+#ifdef RT_USING_SMP
+void arm_gic_send_affinity_sgi(rt_uint64_t index, int irq, rt_uint64_t cpu_masks[], rt_uint64_t routing_mode)
+{
+    const int cpu_mask_cpu_max_nr = sizeof(cpu_masks[0]) * 8;
+    rt_uint64_t int_id = (irq & 0xf) << 24;
+    rt_uint64_t irm = routing_mode << 40; /* Interrupt Routing Mode */
+
+    if (routing_mode == GICV3_ROUTED_TO_SPEC)
+    {
+        int cpu_id, cpu_mask_bit, i, cpu_masks_nr = RT_CPUS_NR / cpu_mask_cpu_max_nr;
+        rt_uint16_t target_list;
+        rt_uint64_t rs = 0; /* Range Selector */
+        rt_uint64_t affinity_val, next_affinity_val;
+
+        if (cpu_masks_nr * cpu_mask_cpu_max_nr != RT_CPUS_NR)
+        {
+            ++cpu_masks_nr;
+        }
+
+        for (i = cpu_id = 0; i < cpu_masks_nr;)
+        {
+            /* No cpu in this mask */
+            if (cpu_masks[i] == 0)
+            {
+                ++i;
+                cpu_id += cpu_mask_cpu_max_nr;
+                continue;
+            }
+
+            /* Get last cpu affinity value */
+            affinity_val = rt_cpu_mpidr_early[cpu_id] & 0xff00ffff00ULL;
+
+            /* Read 16 cpus information */
+            for (cpu_mask_bit = 0; cpu_mask_bit < 16; ++cpu_mask_bit, ++cpu_id)
+            {
+                /* MPIDR_EL1: aff3[39:32], aff2[23:16], aff1[15:8] */
+                next_affinity_val = rt_cpu_mpidr_early[cpu_id] & 0xff00ffff00ULL;
+
+                /* Affinity value is different, read end */
+                if (affinity_val != next_affinity_val)
+                {
+                    break;
+                }
+            }
+
+            /* Get all valid cpu mask */
+            target_list = (0xffff >> (16 - cpu_mask_bit)) & cpu_masks[i];
+            /* Clear read mask */
+            cpu_masks[i] >>= cpu_mask_bit;
+            /* ICC_SGI1R_EL1: aff3[55:48], aff2[39:32], aff1[23:16] */
+            affinity_val <<= 8;
+
+            __DSB();
+            /* Interrupts routed to the PEs specified by Aff3.Aff2.Aff1.<target list>. */
+            SET_GICV3_REG(ICC_SGI1R_EL1, affinity_val | (rs << 44) | irm | int_id | target_list);
+            __ISB();
+
+            /* Check if reset the range selector */
+            rs = affinity_val != next_affinity_val ? 0 : rs + 1;
+        }
+    }
+    else
+    {
+        __DSB();
+        /* Interrupts routed to all PEs in the system, excluding "self". */
+        SET_GICV3_REG(ICC_SGI1R_EL1, irm | int_id);
+        __ISB();
+    }
+}
+#endif /* RT_USING_SMP */
+
+rt_uint64_t arm_gic_get_high_pending_irq(rt_uint64_t index)
+{
+    rt_uint64_t irq;
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    index = index;
+    GET_GICV3_REG(ICC_HPPIR1_EL1, irq);
+
+    return irq;
+}
+
+rt_uint64_t arm_gic_get_interface_id(rt_uint64_t index)
+{
+    rt_uint64_t ret = 0;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    if (_gic_table[index].cpu_hw_base != RT_NULL)
+    {
+        ret = GIC_CPU_IIDR(_gic_table[index].cpu_hw_base);
+    }
+
+    return ret;
+}
+
+void arm_gic_set_group(rt_uint64_t index, int irq, rt_uint64_t group)
+{
+    rt_uint64_t igroupr;
+    rt_uint64_t shift;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+    RT_ASSERT(group <= 1);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    igroupr = GIC_DIST_IGROUP(_gic_table[index].dist_hw_base, irq);
+    shift = (irq % 32);
+    igroupr &= (~(1U << shift));
+    igroupr |= ((group & 0x1U) << shift);
+
+    GIC_DIST_IGROUP(_gic_table[index].dist_hw_base, irq) = igroupr;
+}
+
+rt_uint64_t arm_gic_get_group(rt_uint64_t index, int irq)
+{
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    irq = irq - _gic_table[index].offset;
+    RT_ASSERT(irq >= 0);
+
+    return (GIC_DIST_IGROUP(_gic_table[index].dist_hw_base, irq) >> (irq % 32)) & 0x1UL;
+}
+
+static int arm_gicv3_wait_rwp(rt_uint64_t index, rt_uint64_t irq)
+{
+    rt_uint64_t rwp_bit;
+    rt_uint64_t base;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    if (irq < 32)
+    {
+        rt_int32_t cpu_id = rt_hw_cpu_id();
+
+        base = _gic_table[index].redist_hw_base[cpu_id];
+        rwp_bit = GICR_CTLR_RWP;
+    }
+    else
+    {
+        base = _gic_table[index].dist_hw_base;
+        rwp_bit = GICD_CTLR_RWP;
+    }
+
+    while (HWREG32(base) & rwp_bit)
+    {
+    }
+
+    return 0;
+}
+
+int arm_gic_dist_init(rt_uint64_t index, rt_uint64_t dist_base, int irq_start)
+{
+    int i;
+    unsigned int gic_type;
+    rt_uint64_t main_cpu_affinity_val;
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    _gic_table[index].dist_hw_base = dist_base;
+    _gic_table[index].offset = irq_start;
+
+
+    /* Find out how many interrupts are supported. */
+    gic_type = GIC_DIST_TYPE(dist_base);
+    _gic_max_irq = ((gic_type & 0x1f) + 1) * 32;
+
+    /*
+     * The GIC only supports up to 1020 interrupt sources.
+     * Limit this to either the architected maximum, or the
+     * platform maximum.
+     */
+    if (_gic_max_irq > 1020)
+    {
+        _gic_max_irq = 1020;
+    }
+    if (_gic_max_irq > ARM_GIC_NR_IRQS) /* the platform maximum interrupts */
+    {
+        _gic_max_irq = ARM_GIC_NR_IRQS;
+    }
+
+    GIC_DIST_CTRL(dist_base) = 0;
+    /* Wait for register write pending */
+    arm_gicv3_wait_rwp(0, 32);
+
+    /* Set all global interrupts to be level triggered, active low. */
+    for (i = 32; i < _gic_max_irq; i += 16)
+    {
+        GIC_DIST_CONFIG(dist_base, i) = 0;
+    }
+
+    arm_gicv3_wait_rwp(0, 32);
+
+#ifdef RT_USING_SMP
+    main_cpu_affinity_val = rt_cpu_mpidr_early[ARM_SPI_BIND_CPU_ID];
+#else
+    __asm__ volatile ("mrs %0, mpidr_el1":"=r"(main_cpu_affinity_val));
+#endif
+
+    /* aff3[39:32], aff2[23:16], aff1[15:8], aff0[7:0] */
+    main_cpu_affinity_val &= 0xff00ffffffULL;
+
+    /* Set all global interrupts to this CPU only. */
+    for (i = 32; i < _gic_max_irq; i++)
+    {
+        GIC_DIST_IROUTER(dist_base, i) = main_cpu_affinity_val | (GICV3_ROUTED_TO_SPEC << 31);
+    }
+
+    arm_gicv3_wait_rwp(0, 32);
+
+    /* Set priority on spi interrupts. */
+    for (i = 32; i < _gic_max_irq; i += 4)
+    {
+        GIC_DIST_PRI(dist_base, i) = 0xa0a0a0a0;
+    }
+
+    arm_gicv3_wait_rwp(0, 32);
+    /* Disable all interrupts. */
+    for (i = 0; i < _gic_max_irq; i += 32)
+    {
+        GIC_DIST_PENDING_CLEAR(dist_base, i) = 0xffffffff;
+        GIC_DIST_ENABLE_CLEAR(dist_base, i) = 0xffffffff;
+    }
+
+    arm_gicv3_wait_rwp(0, 32);
+    /* All interrupts defaults to IGROUP1(IRQ). */
+    for (i = 0; i < _gic_max_irq; i += 32)
+    {
+        GIC_DIST_IGROUP(dist_base, i) = 0xffffffff;
+    }
+
+    arm_gicv3_wait_rwp(0, 32);
+
+    /*
+     * The Distributor control register (GICD_CTLR) must be configured to enable the interrupt groups and to set the routing mode.
+     * Enable Affinity routing (ARE bits) The ARE bits in GICD_CTLR control whether affinity routing is enabled.
+     * If affinity routing is not enabled, GICv3 can be configured for legacy operation.
+     * Whether affinity routing is enabled or not can be controlled separately for Secure and Non-secure state.
+     * Enables GICD_CTLR contains separate enable bits for Group 0, Secure Group 1 and Non-secure Group 1:
+     * GICD_CTLR.EnableGrp1S enables distribution of Secure Group 1 interrupts.
+     * GICD_CTLR.EnableGrp1NS enables distribution of Non-secure Group 1 interrupts.
+     * GICD_CTLR.EnableGrp0 enables distribution of Group 0 interrupts.
+     */
+    GIC_DIST_CTRL(dist_base) = GICD_CTLR_ARE_NS | GICD_CTLR_ENGRP1NS;
+
+    return 0;
+}
+
+int arm_gic_redist_init(rt_uint64_t index, rt_uint64_t redist_base)
+{
+    int i;
+    int cpu_id = rt_hw_cpu_id();
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    redist_base += cpu_id * (2 << 16);
+    _gic_table[index].redist_hw_base[cpu_id] = redist_base;
+
+    /* redistributor enable */
+    GIC_RDIST_WAKER(redist_base) &= ~(1 << 1);
+    while (GIC_RDIST_WAKER(redist_base) & (1 << 2))
+    {
+    }
+
+    /* Disable all sgi and ppi interrupt */
+    GIC_RDISTSGI_ICENABLER0(redist_base) = 0xffffffff;
+    arm_gicv3_wait_rwp(0, 0);
+
+    /* Clear all inetrrupt pending */
+    GIC_RDISTSGI_ICPENDR0(redist_base) = 0xffffffff;
+
+    /* the corresponding interrupt is Group 1 or Non-secure Group 1. */
+    GIC_RDISTSGI_IGROUPR0(redist_base, 0) = 0xffffffff;
+    GIC_RDISTSGI_IGRPMODR0(redist_base, 0) = 0xffffffff;
+
+    /* Configure default priorities for SGI 0:15 and PPI 16:31. */
+    for (i = 0; i < 32; i += 4)
+    {
+        GIC_RDISTSGI_IPRIORITYR(redist_base, i) = 0xa0a0a0a0U;
+    }
+
+    /* Trigger level for PPI interrupts*/
+    GIC_RDISTSGI_ICFGR1(redist_base) = 0;
+    return 0;
+}
+
+int arm_gic_cpu_init(rt_uint64_t index, rt_uint64_t cpu_base)
+{
+    rt_uint64_t value;
+    int cpu_id = rt_hw_cpu_id();
+
+    RT_ASSERT(index < ARM_GIC_MAX_NR);
+
+    _gic_table[index].cpu_hw_base[cpu_id] = cpu_base;
+
+    value = arm_gic_get_system_register_enable_mask(index);
+    value |= (1 << 0);
+    arm_gic_set_system_register_enable_mask(index, value);
+    SET_GICV3_REG(ICC_CTLR_EL1, 0);
+
+    arm_gic_set_interface_prior_mask(index, 0xff);
+
+    /* Enable group1 interrupt */
+    value = 1;
+    SET_GICV3_REG(ICC_IGRPEN1_EL1, value);
+
+    arm_gic_set_binary_point(0, 0);
+
+    /* ICC_BPR0_EL1 determines the preemption group for both Group 0 and Group 1 interrupts. */
+    value  = 1;         /* ICC_BPR0_EL1 determines the preemption group for both Group 0 and Group 1 interrupts.*/
+    value |= 1 << 18;   /* Targeted SGIs with affinity level 0 values of 0 - 255 are supported. */
+    SET_GICV3_REG(ICC_CTLR_EL1, value);
+
+    return 0;
+}
+
+void arm_gic_dump_type(rt_uint64_t index)
+{
+    unsigned int gic_type;
+
+    gic_type = GIC_DIST_TYPE(_gic_table[index].dist_hw_base);
+    rt_kprintf("GICv%d on %p, max IRQs: %d, %s security extension(%08x)\n",
+               (GIC_DIST_ICPIDR2(_gic_table[index].dist_hw_base) >> 4) & 0xf,
+               _gic_table[index].dist_hw_base,
+               _gic_max_irq,
+               gic_type & (1 << 10) ? "has" : "no",
+               gic_type);
+}
+
+void arm_gic_dump(rt_uint64_t index)
+{
+    int i;
+    unsigned int val;
+
+    val = arm_gic_get_high_pending_irq(0);
+    rt_kprintf("--- high pending priority: %d(%08x)\n", val, val);
+
+    rt_kprintf("--- hw mask ---\n");
+    for (i = 0; i < _gic_max_irq / 32; ++i)
+    {
+        rt_kprintf("0x%08x, ", GIC_DIST_ENABLE_SET(_gic_table[index].dist_hw_base, i * 32));
+    }
+
+    rt_kprintf("\b\b\n--- hw pending ---\n");
+    for (i = 0; i < _gic_max_irq / 32; ++i)
+    {
+        rt_kprintf("0x%08x, ", GIC_DIST_PENDING_SET(_gic_table[index].dist_hw_base, i * 32));
+    }
+
+    rt_kprintf("\b\b\n--- hw active ---\n");
+    for (i = 0; i < _gic_max_irq / 32; ++i)
+    {
+        rt_kprintf("0x%08x, ", GIC_DIST_ACTIVE_SET(_gic_table[index].dist_hw_base, i * 32));
+    }
+
+    rt_kprintf("\b\b\n");
+}
+
+long gic_dump(void)
+{
+    arm_gic_dump_type(0);
+    arm_gic_dump(0);
+
+    return 0;
+}
+MSH_CMD_EXPORT(gic_dump, show gic status);
+
+#endif /* defined(BSP_USING_GIC) && defined(BSP_USING_GICV3) */
diff --git a/libcpu/aarch64/common/gicv3.h b/libcpu/aarch64/common/gicv3.h
new file mode 100644
index 0000000000..0cb6bda03a
--- /dev/null
+++ b/libcpu/aarch64/common/gicv3.h
@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2006-2022, RT-Thread Development Team
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ *
+ * Change Logs:
+ * Date           Author       Notes
+ * 2013-07-20     Bernard      first version
+ * 2014-04-03     Grissiom     many enhancements
+ * 2018-11-22     Jesven       add rt_hw_ipi_send()
+ *                             add rt_hw_ipi_handler_install()
+ */
+
+#ifndef __GICV3_H__
+#define __GICV3_H__
+
+#include <rtdef.h>
+
+#if defined(BSP_USING_GIC) && defined(BSP_USING_GICV3)
+
+#define GICV3_ROUTED_TO_ALL   1UL
+#define GICV3_ROUTED_TO_SPEC  0UL
+
+int arm_gic_get_active_irq(rt_uint64_t index);
+void arm_gic_ack(rt_uint64_t index, int irq);
+
+void arm_gic_mask(rt_uint64_t index, int irq);
+void arm_gic_umask(rt_uint64_t index, int irq);
+
+rt_uint64_t arm_gic_get_pending_irq(rt_uint64_t index, int irq);
+void arm_gic_set_pending_irq(rt_uint64_t index, int irq);
+void arm_gic_clear_pending_irq(rt_uint64_t index, int irq);
+
+void arm_gic_set_configuration(rt_uint64_t index, int irq, rt_uint32_t config);
+rt_uint64_t arm_gic_get_configuration(rt_uint64_t index, int irq);
+
+void arm_gic_clear_active(rt_uint64_t index, int irq);
+
+void arm_gic_set_cpu(rt_uint64_t index, int irq, unsigned int cpumask);
+rt_uint64_t arm_gic_get_target_cpu(rt_uint64_t index, int irq);
+
+void arm_gic_set_priority(rt_uint64_t index, int irq, rt_uint64_t priority);
+rt_uint64_t arm_gic_get_priority(rt_uint64_t index, int irq);
+
+void arm_gic_set_interface_prior_mask(rt_uint64_t index, rt_uint64_t priority);
+rt_uint64_t arm_gic_get_interface_prior_mask(rt_uint64_t index);
+
+void arm_gic_set_binary_point(rt_uint64_t index, rt_uint64_t binary_point);
+rt_uint64_t arm_gic_get_binary_point(rt_uint64_t index);
+
+rt_uint64_t arm_gic_get_irq_status(rt_uint64_t index, int irq);
+
+#ifdef RT_USING_SMP
+void arm_gic_send_affinity_sgi(rt_uint64_t index, int irq, rt_uint64_t cpu_masks[], rt_uint64_t routing_mode);
+#endif
+
+rt_uint64_t arm_gic_get_high_pending_irq(rt_uint64_t index);
+
+rt_uint64_t arm_gic_get_interface_id(rt_uint64_t index);
+
+void arm_gic_set_group(rt_uint64_t index, int irq, rt_uint64_t group);
+rt_uint64_t arm_gic_get_group(rt_uint64_t index, int irq);
+
+int arm_gic_redist_address_set(rt_uint64_t index, rt_uint64_t redist_addr, int cpu_id);
+int arm_gic_cpu_interface_address_set(rt_uint64_t index, rt_uint64_t interface_addr, int cpu_id);
+
+int arm_gic_dist_init(rt_uint64_t index, rt_uint64_t dist_base, int irq_start);
+int arm_gic_redist_init(rt_uint64_t index, rt_uint64_t redist_base);
+int arm_gic_cpu_init(rt_uint64_t index, rt_uint64_t cpu_base);
+
+void arm_gic_dump_type(rt_uint64_t index);
+void arm_gic_dump(rt_uint64_t index);
+
+#endif /* defined(BSP_USING_GIC) && defined(BSP_USING_GICV3) */
+
+#endif
+
diff --git a/libcpu/aarch64/common/interrupt.c b/libcpu/aarch64/common/interrupt.c
index f31003b2c0..c5db2d46af 100644
--- a/libcpu/aarch64/common/interrupt.c
+++ b/libcpu/aarch64/common/interrupt.c
@@ -13,6 +13,7 @@
 #include <rtthread.h>
 #include "interrupt.h"
 #include "gic.h"
+#include "gicv3.h"
 #include "armv8.h"
 #include "mmu.h"
 #include "cpuport.h"
@@ -52,6 +53,9 @@ void rt_hw_interrupt_init(void)
     /* initialize ARM GIC */
     arm_gic_dist_init(0, platform_get_gic_dist_base(), GIC_IRQ_START);
     arm_gic_cpu_init(0, platform_get_gic_cpu_base());
+#ifdef BSP_USING_GICV3
+    arm_gic_redist_init(0, platform_get_gic_redist_base());
+#endif
 #endif
 }
 
@@ -346,8 +350,12 @@ rt_isr_handler_t rt_hw_interrupt_install(int vector, rt_isr_handler_t handler,
 void rt_hw_ipi_send(int ipi_vector, unsigned int cpu_mask)
 {
 #ifdef BSP_USING_GIC
+#ifdef BSP_USING_GICV2
     arm_gic_send_sgi(0, ipi_vector, cpu_mask, 0);
 #else
+    arm_gic_send_affinity_sgi(0, ipi_vector, (rt_uint64_t *)&cpu_mask, GICV3_ROUTED_TO_SPEC);
+#endif
+#else
     int i;
 
     __DSB();
@@ -359,8 +367,9 @@ void rt_hw_ipi_send(int ipi_vector, unsigned int cpu_mask)
             IPI_MAILBOX_SET(i) = 1 << ipi_vector;
         }
     }
-#endif
+
     __DSB();
+#endif
 }
 
 void rt_hw_ipi_handler_install(int ipi_vector, rt_isr_handler_t ipi_isr_handler)
diff --git a/libcpu/aarch64/common/psci.c b/libcpu/aarch64/common/psci.c
index 400db977ee..65e84ebf0a 100644
--- a/libcpu/aarch64/common/psci.c
+++ b/libcpu/aarch64/common/psci.c
@@ -29,11 +29,19 @@ static psci_call_handle psci_call = psci_smc_call;
 static uint64_t shutdown_args[3] = {0, 0, 0};
 static uint64_t reboot_args[3] = {0, 0, 0};
 
-void arm_psci_init(uint64_t *platform_shutdown_args, uint64_t *platform_reboot_args)
+void arm_psci_init(uint64_t method, uint64_t *platform_shutdown_args, uint64_t *platform_reboot_args)
 {
-    if (rt_hw_get_current_el() < 2)
+    switch (method)
     {
-        psci_call = psci_hvc_call;
+    case PSCI_METHOD_SMC:
+        psci_call = psci_smc_call;
+        break;
+    case PSCI_METHOD_HVC:
+        if (rt_hw_get_current_el() < 2)
+        {
+            psci_call = psci_hvc_call;
+        }
+        break;
     }
 
     if (platform_shutdown_args != RT_NULL)
diff --git a/libcpu/aarch64/common/psci.h b/libcpu/aarch64/common/psci.h
index e8b2e3129d..999489a801 100644
--- a/libcpu/aarch64/common/psci.h
+++ b/libcpu/aarch64/common/psci.h
@@ -18,6 +18,9 @@
  */
 #define PSCI_VER_0_2                        0x00000002
 
+#define PSCI_METHOD_SMC                     3
+#define PSCI_METHOD_HVC                     2
+
 /* PSCI 0.2 interface */
 #define PSCI_0_2_FN_BASE                    0x84000000
 #define PSCI_0_2_FN(n)                      (PSCI_0_2_FN_BASE + (n))
@@ -114,7 +117,7 @@
 #define PSCI_RET_DISABLED                   (-8)
 #define PSCI_RET_INVALID_ADDRESS            (-9)
 
-void arm_psci_init(uint64_t *platform_shutdown_args, uint64_t *platform_reboot_args);
+void arm_psci_init(uint64_t method, uint64_t *platform_shutdown_args, uint64_t *platform_reboot_args);
 
 uint32_t arm_psci_get_version();
 uint32_t arm_psci_get_affinity_info(uint64_t target_affinity, uint64_t lowest_affinity_level);
diff --git a/libcpu/aarch64/cortex-a/entry_point.S b/libcpu/aarch64/cortex-a/entry_point.S
index 5bf2b8a3a6..c605a41631 100644
--- a/libcpu/aarch64/cortex-a/entry_point.S
+++ b/libcpu/aarch64/cortex-a/entry_point.S
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2006-2020, RT-Thread Development Team
+ * Copyright (c) 2006-2022, RT-Thread Development Team
  *
  * SPDX-License-Identifier: Apache-2.0
  *
@@ -11,19 +11,28 @@
  */
 
 #include "rtconfig.h"
-.section ".text.entrypoint","ax"
 
 #define SECONDARY_STACK_SIZE 4096
 
+.section ".text.entrypoint","ax"
+
 .globl _start
 .globl secondary_cpu_start
 _start:
+#ifdef RT_USING_SMP
     mrs     x1, mpidr_el1
-    and     x1, x1, #0xff
-    cbnz    x1, cpu_idle            /* If cpu id > 0, stop slave cores */
+    adr     x4, .boot_cpu_mpidr
+    str     x1, [x4]
+    dsb     sy
+#endif
+
+    bl      __asm_flush_dcache_all  /* The kernel and data must flush to DDR */
 
 secondary_cpu_start:
 #ifdef RT_USING_SMP
+    adr     x4, .boot_cpu_mpidr
+    ldr     x4, [x4]
+    dsb     sy
     /* Read cpu mpidr_el1 */
     mrs     x1, mpidr_el1
 
@@ -34,6 +43,7 @@ secondary_cpu_start:
 cpu_id_confirm:
     add     x2, x2, #1              /* Next cpu id inc */
     ldr     x3, [x0], #8
+    dsb     sy
     cmp     x3, #0
     beq     cpu_idle                /* Mean that `rt_cpu_mpidr_early' table is end */
     cmp     x3, x1
@@ -42,7 +52,8 @@ cpu_id_confirm:
     /* Get cpu id success */
     sub     x0, x2, #1
     msr     tpidr_el1, x0           /* Save cpu id global */
-    cbz     x0, cpu_setup           /* Only go to cpu_setup when cpu id = 0 */
+    cmp     x3, x4                  /* If it is boot cpu */
+    beq     boot_cpu_setup
 
     /* Set current cpu's stack top */
     sub     x0, x0, #1
@@ -53,14 +64,9 @@ cpu_id_confirm:
     b       cpu_check_el
 #else
     msr     tpidr_el1, xzr
-    b       cpu_setup
 #endif /* RT_USING_SMP */
 
-cpu_idle:
-    wfe
-    b       cpu_idle
-
-cpu_setup:
+boot_cpu_setup:
     ldr     x1, =_start
 
 cpu_check_el:
@@ -75,7 +81,7 @@ cpu_check_el:
     mov     x2, #(1 << 0)           /* EL0 and EL1 are in Non-Secure state */
     orr     x2, x2, #(1 << 4)       /* RES1 */
     orr     x2, x2, #(1 << 5)       /* RES1 */
-    orr     x2, x2, #(1 << 7)       /* SMC instructions are undefined at EL1 and above */
+    bic     x2, x2, #(1 << 7)       /* SMC instructions are enabled at EL1 and above */
     orr     x2, x2, #(1 << 8)       /* HVC instructions are enabled at EL1 and above */
     orr     x2, x2, #(1 << 10)      /* The next lower level is AArch64 */
     msr     scr_el3, x2
@@ -91,7 +97,7 @@ cpu_check_el:
     eret
 
 cpu_not_in_el3:                     /* Running at EL2 or EL1 */
-    cmp     x0, #4                  /* EL1 = 0100  */
+    cmp     x0, #4                  /* EL1 = 0100 */
     beq     cpu_in_el1
 
 cpu_in_el2:
@@ -159,9 +165,16 @@ clean_bss_check:
 
 jump_to_entry:
     b       rtthread_startup
-    b       cpu_idle                /* For failsafe, halt this core too */
+
+cpu_idle:
+    wfe
+    b       cpu_idle
 
 #ifdef RT_USING_SMP
+.align 3
+.boot_cpu_mpidr:
+    .quad 0x0
+
 .align 12
 .secondary_cpu_stack:
 .space (SECONDARY_STACK_SIZE * (RT_CPUS_NR - 1))