[arm64] add hw thread self (#8942)

* [libcpu] arm64: Add hardware thread_self support

This patch introduces hardware-based thread self-identification
for the AArch64 architecture. It optimizes thread management by
using hardware registers to store and access the current thread's
pointer, reducing overhead and improving overall performance.

Changes include:
- Added `ARCH_USING_HW_THREAD_SELF` configuration option.
- Modified `rtdef.h`, `rtsched.h` to conditionally include
  `critical_switch_flag` based on the new config.
- Updated context management in `context_gcc.S`, `cpuport.h`
  to support hardware-based thread self.
- Enhanced `scheduler_mp.c` and `thread.c` to leverage the new
  hardware thread self feature.

These modifications ensure better scheduling and thread handling,
particularly in multi-core environments, by minimizing the
software overhead associated with thread management.

Signed-off-by: Shell <smokewood@qq.com>

* fixup: address suggestion

* fixup: rt_current_thread as global

* scheduler: add cpu object for UP scheduler

Also, maintain the rt_current_thread in cpu object on UP scheduler.

---------

Signed-off-by: Shell <smokewood@qq.com>

include/rtdef.h

@@ -740,8 +740,10 @@ struct rt_cpu
         struct rt_thread        *current_thread;
 
         rt_uint8_t              irq_switch_flag:1;
-        rt_uint8_t              critical_switch_flag:1;
         rt_uint8_t              sched_lock_flag:1;
+#ifndef ARCH_USING_HW_THREAD_SELF
+        rt_uint8_t              critical_switch_flag:1;
+#endif /* ARCH_USING_HW_THREAD_SELF */
 
         rt_uint8_t              current_priority;
         rt_list_t               priority_table[RT_THREAD_PRIORITY_MAX];
@@ -763,10 +765,19 @@ struct rt_cpu
     struct rt_cpu_usage_stats   cpu_stat;
 #endif
 };
-typedef struct rt_cpu *rt_cpu_t;
+
+#else /* !RT_USING_SMP */
+struct rt_cpu
+{
+    struct rt_thread            *current_thread;
+};
 
 #endif /* RT_USING_SMP */
 
+typedef struct rt_cpu *rt_cpu_t;
+/* Noted: As API to reject writing to this variable from application codes */
+#define rt_current_thread rt_thread_self()
+
 struct rt_thread;
 
 #ifdef RT_USING_SMART

include/rtsched.h

@@ -56,6 +56,10 @@ struct rt_sched_thread_ctx
     rt_uint8_t                  sched_flag_locked:1;    /**< calling thread have the scheduler locked */
     rt_uint8_t                  sched_flag_ttmr_set:1;  /**< thread timer is start */
 
+#ifdef ARCH_USING_HW_THREAD_SELF
+    rt_uint8_t                  critical_switch_flag:1; /**< critical switch pending */
+#endif /* ARCH_USING_HW_THREAD_SELF */
+
 #ifdef RT_USING_SMP
     rt_uint8_t                  bind_cpu;               /**< thread is bind to cpu */
     rt_uint8_t                  oncpu;                  /**< process on cpu */
@@ -170,6 +174,7 @@ rt_err_t rt_sched_thread_timer_stop(struct rt_thread *thread);
 rt_err_t rt_sched_thread_timer_start(struct rt_thread *thread);
 void rt_sched_insert_thread(struct rt_thread *thread);
 void rt_sched_remove_thread(struct rt_thread *thread);
+struct rt_thread *rt_sched_thread_self(void);
 
 #endif /* defined(__RT_KERNEL_SOURCE__) || defined(__RT_IPC_SOURCE__) */
 

include/rtthread.h

@@ -669,6 +669,12 @@ rt_err_t  rt_device_control(rt_device_t dev, int cmd, void *arg);
 void rt_interrupt_enter(void);
 void rt_interrupt_leave(void);
 
+/**
+ * CPU object
+ */
+struct rt_cpu *rt_cpu_self(void);
+struct rt_cpu *rt_cpu_index(int index);
+
 #ifdef RT_USING_SMP
 
 /*
@@ -679,9 +685,6 @@ rt_base_t rt_cpus_lock(void);
 void rt_cpus_unlock(rt_base_t level);
 void rt_cpus_lock_status_restore(struct rt_thread *thread);
 
-struct rt_cpu *rt_cpu_self(void);
-struct rt_cpu *rt_cpu_index(int index);
-
 #ifdef RT_USING_DEBUG
     rt_base_t rt_cpu_get_id(void);
 #else /* !RT_USING_DEBUG */

libcpu/Kconfig

@@ -14,6 +14,8 @@ if ARCH_ARMV8 && ARCH_CPU_64BIT
         default y
     config ARCH_USING_GENERIC_CPUID
         bool "Using generic cpuid implemenation"
+        select ARCH_USING_HW_THREAD_SELF
+        default y if RT_USING_OFW
         default n
     endmenu
 endif
@@ -270,3 +272,7 @@ config ARCH_HOST_SIMULATOR
 config ARCH_CPU_STACK_GROWS_UPWARD
     bool
     default n
+
+config ARCH_USING_HW_THREAD_SELF
+    bool
+    default n

libcpu/aarch64/common/context_gcc.S

@@ -8,6 +8,7 @@
  * 2021-05-18     Jesven       the first version
  * 2023-06-24     WangXiaoyao  Support backtrace for user thread
  * 2024-01-06     Shell        Fix barrier on irq_disable/enable
+ * 2024-01-18     Shell        fix implicit dependency of cpuid management
  */
 
 #ifndef __ASSEMBLY__
@@ -27,15 +28,35 @@ rt_thread_switch_interrupt_flag: .zero  8
 #endif
 
 .text
-.weak rt_hw_cpu_id_set
+
+/**
+ * #ifdef RT_USING_OFW
+ * void rt_hw_cpu_id_set(long cpuid)
+ * #else
+ * void rt_hw_cpu_id_set(void)
+ * #endif
+ */
 .type rt_hw_cpu_id_set, @function
 rt_hw_cpu_id_set:
-    mrs x0, mpidr_el1           /* MPIDR_EL1: Multi-Processor Affinity Register */
+#ifdef ARCH_USING_GENERIC_CPUID
+    .globl  rt_hw_cpu_id_set
+#else /* !ARCH_USING_GENERIC_CPUID */
+    .weak   rt_hw_cpu_id_set
+#endif /* ARCH_USING_GENERIC_CPUID */
+
+#ifndef RT_USING_OFW
+    mrs     x0, mpidr_el1           /* MPIDR_EL1: Multi-Processor Affinity Register */
 #ifdef ARCH_ARM_CORTEX_A55
-    lsr x0, x0, #8
-#endif
-    and x0, x0, #15
-    msr tpidr_el1, x0
+    lsr     x0, x0, #8
+#endif /* ARCH_ARM_CORTEX_A55 */
+    and     x0, x0, #15
+#endif /* !RT_USING_OFW */
+
+#ifdef ARCH_USING_HW_THREAD_SELF
+    msr     tpidrro_el0, x0
+#else /* !ARCH_USING_HW_THREAD_SELF */
+    msr     tpidr_el1, x0
+#endif /* ARCH_USING_HW_THREAD_SELF */
     ret
 
 /*

libcpu/aarch64/common/cpu.c

@@ -231,6 +231,24 @@ int rt_hw_cpu_boot_secondary(int num_cpus, rt_uint64_t *cpu_hw_ids, struct cpu_o
 
 #endif /*RT_USING_SMP*/
 
+/**
+ * Generic hw-cpu-id
+ */
+#ifdef ARCH_USING_GENERIC_CPUID
+
+int rt_hw_cpu_id(void)
+{
+#if RT_CPUS_NR > 1
+    long cpuid;
+    __asm__ volatile("mrs %0, tpidrro_el0":"=r"(cpuid));
+    return cpuid;
+#else
+    return 0;
+#endif /* RT_CPUS_NR > 1 */
+}
+
+#endif /* ARCH_USING_GENERIC_CPUID */
+
 /**
  * @addtogroup ARM CPU
  */

libcpu/aarch64/common/cpuport.h

@@ -7,6 +7,7 @@
  * Date           Author       Notes
  * 2023-10-25     Shell        Move ffs to cpuport, add general implementation
  *                             by inline assembly
+ * 2024-01-18     Shell        support rt_hw_thread_self to improve overall performance
  */
 
 #ifndef  CPUPORT_H__
@@ -27,31 +28,6 @@ typedef struct
     rt_uint32_t value;
 } rt_hw_spinlock_t;
 
-/**
- * Generic hw-cpu-id
- */
-#ifdef ARCH_USING_GENERIC_CPUID
-
-#if RT_CPUS_NR > 1
-
-rt_inline int rt_hw_cpu_id(void)
-{
-    long cpuid;
-    __asm__ volatile("mrs %0, tpidr_el1":"=r"(cpuid));
-    return cpuid;
-}
-
-#else
-
-rt_inline int rt_hw_cpu_id(void)
-{
-    return 0;
-}
-
-#endif /* RT_CPUS_NR > 1 */
-
-#endif /* ARCH_USING_GENERIC_CPUID */
-
 #endif /* RT_USING_SMP */
 
 #define rt_hw_barrier(cmd, ...) \
@@ -107,4 +83,20 @@ rt_inline int __rt_ffs(int value)
 
 #endif /* RT_USING_CPU_FFS */
 
+#ifdef ARCH_USING_HW_THREAD_SELF
+rt_inline struct rt_thread *rt_hw_thread_self(void)
+{
+    struct rt_thread *thread;
+    __asm__ volatile ("mrs %0, tpidr_el1":"=r"(thread));
+
+    return thread;
+}
+
+rt_inline void rt_hw_thread_set_self(struct rt_thread *thread)
+{
+    __asm__ volatile ("msr tpidr_el1, %0"::"r"(thread));
+}
+
+#endif /* ARCH_USING_HW_THREAD_SELF */
+
 #endif  /*CPUPORT_H__*/

libcpu/aarch64/cortex-a/entry_point.S

@@ -7,6 +7,7 @@
  * 2020-01-15     bigmagic     the first version
  * 2020-08-10     SummerGift   support clang compiler
  * 2023-04-29     GuEe-GUI     support kernel's ARM64 boot header
+ * 2024-01-18     Shell        fix implicit dependency of cpuid management
  */
 
 #ifndef __ASSEMBLY__
@@ -95,6 +96,10 @@ _start:
     /* Save cpu stack */
     get_phy stack_top, .boot_cpu_stack_top
     /* Save cpu id temp */
+#ifdef ARCH_USING_HW_THREAD_SELF
+    msr     tpidrro_el0, xzr
+    /* Save thread self */
+#endif /* ARCH_USING_HW_THREAD_SELF */
     msr     tpidr_el1, xzr
 
     bl      init_cpu_el
@@ -149,11 +154,10 @@ _secondary_cpu_entry:
 
     /* Get cpu id success */
     sub     x0, x2, #1
-    msr     tpidr_el1, x0           /* Save cpu id global */
-#else
-    bl      rt_hw_cpu_id_set
-    mrs     x0, tpidr_el1
 #endif /* RT_USING_OFW */
+    /* Save cpu id global */
+    bl      rt_hw_cpu_id_set
+    bl      rt_hw_cpu_id
 
     /* Set current cpu's stack top */
     sub     x0, x0, #1

src/cpu_up.c

@@ -6,10 +6,14 @@
  * Change Logs:
  * Date           Author       Notes
  * 2024-04-19     Shell        Fixup UP irq spinlock
+ * 2024-05-22     Shell        Add UP cpu object and
+ *                             maintain the rt_current_thread inside it
  */
 #include <rthw.h>
 #include <rtthread.h>
 
+static struct rt_cpu _cpu;
+
 /**
  * @brief   Initialize a static spinlock object.
  *
@@ -80,3 +84,25 @@ void rt_spin_unlock_irqrestore(struct rt_spinlock *lock, rt_base_t level)
     rt_exit_critical_safe(critical_level);
     rt_hw_interrupt_enable(level);
 }
+
+/**
+ * @brief   This fucntion will return current cpu object.
+ *
+ * @return  Return a pointer to the current cpu object.
+ */
+struct rt_cpu *rt_cpu_self(void)
+{
+    return &_cpu;
+}
+
+/**
+ * @brief   This fucntion will return the cpu object corresponding to index.
+ *
+ * @param   index is the index of target cpu object.
+ *
+ * @return  Return a pointer to the cpu object corresponding to index.
+ */
+struct rt_cpu *rt_cpu_index(int index)
+{
+    return index == 0 ? &_cpu : RT_NULL;
+}