feat: arm64: generic implementation of vector irq (#9336)

feat: overall implementation of vector irq

This patch generalize the irq handling on up/mp system by adding the
`rt_hw_irq_exit()` & `rt_hw_vector_irq_sched()` API.

Changes:
- Added `rt_hw_irq_exit()` and `rt_hw_vector_irq_sched()` APIs for unified IRQ management.
- Refactored assembly code for both UP and MP systems to use the new IRQ handling flow.
- Removed redundant code and optimized exception handling paths.

Signed-off-by: Shell <[email protected]>

7 files changed, 97 insertions, 109 deletions

diff --git a/libcpu/aarch64/common/include/asm-generic.h b/libcpu/aarch64/common/include/asm-generic.h
index 362ef8f525..d2d32ce73f 100644
--- a/libcpu/aarch64/common/include/asm-generic.h
+++ b/libcpu/aarch64/common/include/asm-generic.h
@@ -23,6 +23,14 @@
     .cfi_endproc;               \
     .size name, .-name;
 
+#define TRACE_SYMBOL(name)
+
+.macro NEVER_RETURN
+#ifdef RT_USING_DEBUG
+    b       .
+#endif /* RT_USING_DEBUG */
+.endm
+
 .macro GET_THREAD_SELF, dst:req
 #ifdef ARCH_USING_HW_THREAD_SELF
     mrs     x0, tpidr_el1
diff --git a/libcpu/aarch64/common/include/vector_gcc.h b/libcpu/aarch64/common/include/vector_gcc.h
index 8fa5c04aec..31c532316e 100644
--- a/libcpu/aarch64/common/include/vector_gcc.h
+++ b/libcpu/aarch64/common/include/vector_gcc.h
@@ -6,6 +6,7 @@
  * Change Logs:
  * Date           Author       Notes
  * 2024-03-28     Shell        Move vector handling codes from context_gcc.S
+ * 2024-04-08     Shell        Optimizing exception switch between u-space/kernel,
  */
 
 #ifndef __ARM64_INC_VECTOR_H__
@@ -45,8 +46,6 @@
     mrs     x2, elr_el1
 
     stp     x2, x3, [sp, #-0x10]!
-
-    mov     x0, sp   /* Move SP into X0 for saving. */
 .endm
 
 #ifdef RT_USING_SMP
@@ -55,60 +54,29 @@
 #include "../up/context_gcc.h"
 #endif
 
-.macro RESTORE_IRQ_CONTEXT_WITHOUT_MMU_SWITCH
-    /* the SP is already ok */
-    ldp     x2, x3, [sp], #0x10  /* SPSR and ELR. */
-
-    tst     x3, #0x1f
-    msr     spsr_el1, x3
-    msr     elr_el1, x2
-
-    ldp     x29, x30, [sp], #0x10
-    msr     sp_el0, x29
-    ldp     x28, x29, [sp], #0x10
-    msr     fpcr, x28
-    msr     fpsr, x29
-    ldp     x28, x29, [sp], #0x10
-    ldp     x26, x27, [sp], #0x10
-    ldp     x24, x25, [sp], #0x10
-    ldp     x22, x23, [sp], #0x10
-    ldp     x20, x21, [sp], #0x10
-    ldp     x18, x19, [sp], #0x10
-    ldp     x16, x17, [sp], #0x10
-    ldp     x14, x15, [sp], #0x10
-    ldp     x12, x13, [sp], #0x10
-    ldp     x10, x11, [sp], #0x10
-    ldp     x8, x9, [sp], #0x10
-    ldp     x6, x7, [sp], #0x10
-    ldp     x4, x5, [sp], #0x10
-    ldp     x2, x3, [sp], #0x10
-    ldp     x0, x1, [sp], #0x10
-    RESTORE_FPU sp
+.macro SAVE_USER_CTX, eframex, tmpx
 #ifdef RT_USING_SMART
-    beq     arch_ret_to_user
-#endif
-    eret
-.endm
-
-.macro SAVE_USER_CTX
-    mrs     x1, spsr_el1
-    and     x1, x1, 0xf
-    cmp     x1, xzr
-
-    bne     1f
-    bl      lwp_uthread_ctx_save
-    ldp     x0, x1, [sp]
+    mrs     \tmpx, spsr_el1
+    and     \tmpx, \tmpx, 0xf
+    cbz     \tmpx, 1f
+    b       2f
 1:
+    mov     x0, \eframex
+    bl      lwp_uthread_ctx_save
+2:
+#endif /* RT_USING_SMART */
 .endm
 
-.macro RESTORE_USER_CTX, ctx
-    ldr     x1, [\ctx, #CONTEXT_OFFSET_SPSR_EL1]
-    and     x1, x1, 0x1f
-    cmp     x1, xzr
-
-    bne     1f
-    bl      lwp_uthread_ctx_restore
+.macro RESTORE_USER_CTX, eframex, tmpx
+#ifdef RT_USING_SMART
+    ldr     \tmpx, [\eframex, #CONTEXT_OFFSET_SPSR_EL1]
+    and     \tmpx, \tmpx, 0x1f
+    cbz     \tmpx, 1f
+    b       2f
 1:
+    bl      lwp_uthread_ctx_restore
+2:
+#endif /* RT_USING_SMART */
 .endm
 
 #endif /* __ARM64_INC_VECTOR_H__ */
diff --git a/libcpu/aarch64/common/mp/context_gcc.S b/libcpu/aarch64/common/mp/context_gcc.S
index 355899e34d..adc3587651 100644
--- a/libcpu/aarch64/common/mp/context_gcc.S
+++ b/libcpu/aarch64/common/mp/context_gcc.S
@@ -129,5 +129,7 @@ rt_hw_context_switch_interrupt:
     b       _context_switch_exit
 
 _context_switch_exit:
+    .local _context_switch_exit
+
     clrex
     RESTORE_CONTEXT_SWITCH
diff --git a/libcpu/aarch64/common/mp/vector_gcc.S b/libcpu/aarch64/common/mp/vector_gcc.S
index 2b3ebd584e..911098fcec 100644
--- a/libcpu/aarch64/common/mp/vector_gcc.S
+++ b/libcpu/aarch64/common/mp/vector_gcc.S
@@ -12,41 +12,23 @@
 #define __ASSEMBLY__
 #endif
 
-#include "../include/vector_gcc.h"
+#include "vector_gcc.h"
 #include "context_gcc.h"
 
 .section .text
 
-.globl vector_fiq
 vector_fiq:
+    .globl vector_fiq
     b       .
 
 .globl rt_hw_irq_exit
 
-START_POINT(vector_irq)
-    SAVE_IRQ_CONTEXT
-    stp     x0, x1, [sp, #-0x10]!   /* X0 is thread sp */
-
-    bl      rt_interrupt_enter
-    ldp     x0, x1, [sp]
-
-#ifdef RT_USING_SMART
-    SAVE_USER_CTX
-#endif /* RT_USING_SMART */
-
-    bl      rt_hw_trap_irq
-
-#ifdef RT_USING_SMART
-    ldp     x0, x1, [sp]
-    RESTORE_USER_CTX x0
-#endif /* RT_USING_SMART */
-
-    bl      rt_interrupt_leave
+/**
+ * void rt_hw_vector_irq_sched(void *eframe)
+ * @brief do IRQ scheduling
+ */
+rt_hw_vector_irq_sched:
+    .globl rt_hw_vector_irq_sched
 
-    ldp     x0, x1, [sp], #0x10
     bl      rt_scheduler_do_irq_switch
-
-rt_hw_irq_exit:
-    RESTORE_IRQ_CONTEXT
-
-START_POINT_END(vector_irq)
+    b       rt_hw_irq_exit
diff --git a/libcpu/aarch64/common/up/context_gcc.S b/libcpu/aarch64/common/up/context_gcc.S
index 59e0dbd093..1a2933c447 100644
--- a/libcpu/aarch64/common/up/context_gcc.S
+++ b/libcpu/aarch64/common/up/context_gcc.S
@@ -45,6 +45,7 @@ rt_hw_context_switch_to:
     clrex
     ldr     x0, [x0]
     RESTORE_CONTEXT_SWITCH x0
+    NEVER_RETURN
 
 /*
  * void rt_hw_context_switch(rt_ubase_t from, rt_ubase_t to);
@@ -62,6 +63,7 @@ rt_hw_context_switch:
     ldr    x0, [x1]            // get new task stack pointer
 
     RESTORE_CONTEXT_SWITCH x0
+    NEVER_RETURN
 
 .globl rt_thread_switch_interrupt_flag
 .globl rt_interrupt_from_thread
@@ -115,3 +117,4 @@ rt_hw_context_switch_interrupt_do:
     ldr     x0,  [x4]       // get new task's stack pointer
 
     RESTORE_CONTEXT_SWITCH x0
+    NEVER_RETURN
diff --git a/libcpu/aarch64/common/up/vector_gcc.S b/libcpu/aarch64/common/up/vector_gcc.S
index 0b2ddeb8d0..26428fa597 100644
--- a/libcpu/aarch64/common/up/vector_gcc.S
+++ b/libcpu/aarch64/common/up/vector_gcc.S
@@ -22,28 +22,24 @@
 
 .section .text
 
+vector_fiq:
     .align  8
     .globl vector_fiq
-vector_fiq:
+
     SAVE_IRQ_CONTEXT
     bl      rt_hw_trap_fiq
     RESTORE_IRQ_CONTEXT
 
-.globl      rt_interrupt_enter
-.globl      rt_interrupt_leave
 .globl      rt_thread_switch_interrupt_flag
-.globl      rt_interrupt_from_thread
-.globl      rt_interrupt_to_thread
 .globl      rt_hw_context_switch_interrupt_do
 
+/**
+ * void rt_hw_vector_irq_sched(void *eframe)
+ * @brief do IRQ scheduling
+ */
+rt_hw_vector_irq_sched:
+    .globl rt_hw_vector_irq_sched
     .align  8
-    .globl vector_irq
-vector_irq:
-    SAVE_IRQ_CONTEXT
-
-    bl      rt_interrupt_enter
-    bl      rt_hw_trap_irq
-    bl      rt_interrupt_leave
 
     /**
      * if rt_thread_switch_interrupt_flag set, jump to
@@ -52,12 +48,13 @@ vector_irq:
     ldr     x1, =rt_thread_switch_interrupt_flag
     ldr     x2, [x1]
     cmp     x2, #1
-    b.ne    vector_irq_exit
+    bne     1f
 
-    mov     x2,  #0         // clear flag
+    /* clear flag */
+    mov     x2,  #0
     str     x2,  [x1]
 
     bl      rt_hw_context_switch_interrupt_do
 
-vector_irq_exit:
-    RESTORE_IRQ_CONTEXT_WITHOUT_MMU_SWITCH
+1:
+    b       rt_hw_irq_exit
diff --git a/libcpu/aarch64/common/vector_gcc.S b/libcpu/aarch64/common/vector_gcc.S
index 6475e19a01..aaa172fcf9 100644
--- a/libcpu/aarch64/common/vector_gcc.S
+++ b/libcpu/aarch64/common/vector_gcc.S
@@ -67,32 +67,60 @@ system_vectors:
     b vector_serror
 
 #include "include/vector_gcc.h"
+#define EFRAMEX x19
 
 START_POINT(vector_exception)
     SAVE_IRQ_CONTEXT
-    stp     x0, x1, [sp, #-0x10]!
-#ifdef RT_USING_SMART
-    SAVE_USER_CTX
-#endif
+    mov     EFRAMEX, sp
+
+    SAVE_USER_CTX EFRAMEX, x0
 
+    mov     x0, EFRAMEX
     bl      rt_hw_trap_exception
-#ifdef RT_USING_SMART
-    ldp     x0, x1, [sp]
-    RESTORE_USER_CTX x0
-#endif
+    RESTORE_USER_CTX EFRAMEX, x0
 
-    ldp     x0, x1, [sp], #0x10
-    RESTORE_IRQ_CONTEXT_WITHOUT_MMU_SWITCH
+    RESTORE_IRQ_CONTEXT
 START_POINT_END(vector_exception)
 
 START_POINT(vector_serror)
     SAVE_IRQ_CONTEXT
+    mov     EFRAMEX, sp
 
-#ifdef RT_USING_SMART
-    SAVE_USER_CTX
-#endif
+    SAVE_USER_CTX EFRAMEX, x0
 
-    stp     x0, x1, [sp, #-0x10]!
+    mov     x0, EFRAMEX
     bl      rt_hw_trap_serror
-    b       .
+
+    RESTORE_USER_CTX EFRAMEX, x0
+
+    NEVER_RETURN
 START_POINT_END(vector_serror)
+
+START_POINT(vector_irq)
+    SAVE_IRQ_CONTEXT
+    mov     EFRAMEX, sp
+
+    /* trace IRQ level */
+    bl      rt_interrupt_enter
+
+    SAVE_USER_CTX EFRAMEX, x0
+
+    /* handline IRQ */
+    mov     x0, EFRAMEX
+    bl      rt_hw_trap_irq
+
+    RESTORE_USER_CTX EFRAMEX, x0
+
+    /* restore IRQ level */
+    bl      rt_interrupt_leave
+
+    mov     x0, EFRAMEX
+    bl      rt_hw_vector_irq_sched
+
+    b       rt_hw_irq_exit
+START_POINT_END(vector_irq)
+
+rt_hw_irq_exit:
+    .globl rt_hw_irq_exit
+
+    RESTORE_IRQ_CONTEXT