From a94fb0cd58092e938fb1f3d2510d17831031284c Mon Sep 17 00:00:00 2001
From: Wang Chen <wangchen20@iscas.ac.cn>
Date: Sat, 17 Jun 2023 21:46:13 +0800
Subject: [PATCH] stack pointer aligment

Following the standard RISC-V calling convention, make sure
the stack pointer sp is always 16-byte aligned.

Fixed an issue, make sure the sp of task point to the bottom
of the stack, while originally we waste one byte (forgive my
stupid ~~~).

Signed-off-by: Wang Chen <wangchen20@iscas.ac.cn>
---
 code/os/00-bootstrap/start.S     | 3 +++
 code/os/01-helloRVOS/start.S     | 3 +++
 code/os/02-memanagement/start.S  | 3 +++
 code/os/03-contextswitch/sched.c | 8 ++++++--
 code/os/03-contextswitch/start.S | 3 +++
 code/os/04-multitask/sched.c     | 8 ++++++--
 code/os/04-multitask/start.S     | 3 +++
 code/os/05-traps/sched.c         | 8 ++++++--
 code/os/05-traps/start.S         | 3 +++
 code/os/06-interrupts/sched.c    | 8 ++++++--
 code/os/06-interrupts/start.S    | 3 +++
 code/os/07-hwtimer/sched.c       | 8 ++++++--
 code/os/07-hwtimer/start.S       | 3 +++
 code/os/08-preemptive/sched.c    | 8 ++++++--
 code/os/08-preemptive/start.S    | 3 +++
 code/os/09-lock/sched.c          | 8 ++++++--
 code/os/09-lock/start.S          | 3 +++
 code/os/10-swtimer/sched.c       | 8 ++++++--
 code/os/10-swtimer/start.S       | 3 +++
 code/os/11-syscall/sched.c       | 8 ++++++--
 code/os/11-syscall/start.S       | 3 +++
 21 files changed, 90 insertions(+), 18 deletions(-)

diff --git a/code/os/00-bootstrap/start.S b/code/os/00-bootstrap/start.S
index 5c4e011..ae25bb9 100644
--- a/code/os/00-bootstrap/start.S
+++ b/code/os/00-bootstrap/start.S
@@ -26,6 +26,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/01-helloRVOS/start.S b/code/os/01-helloRVOS/start.S
index 5c4e011..ae25bb9 100644
--- a/code/os/01-helloRVOS/start.S
+++ b/code/os/01-helloRVOS/start.S
@@ -26,6 +26,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/02-memanagement/start.S b/code/os/02-memanagement/start.S
index c7ad5e4..5bdeff1 100644
--- a/code/os/02-memanagement/start.S
+++ b/code/os/02-memanagement/start.S
@@ -36,6 +36,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/03-contextswitch/sched.c b/code/os/03-contextswitch/sched.c
index b2933d0..e1ef938 100644
--- a/code/os/03-contextswitch/sched.c
+++ b/code/os/03-contextswitch/sched.c
@@ -4,7 +4,11 @@
 extern void switch_to(struct context *next);
 
 #define STACK_SIZE 1024
-uint8_t task_stack[STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[STACK_SIZE];
 struct context ctx_task;
 
 static void w_mscratch(reg_t x)
@@ -17,7 +21,7 @@ void sched_init()
 {
 	w_mscratch(0);
 
-	ctx_task.sp = (reg_t) &task_stack[STACK_SIZE - 1];
+	ctx_task.sp = (reg_t) &task_stack[STACK_SIZE];
 	ctx_task.ra = (reg_t) user_task0;
 }
 
diff --git a/code/os/03-contextswitch/start.S b/code/os/03-contextswitch/start.S
index c7ad5e4..5bdeff1 100644
--- a/code/os/03-contextswitch/start.S
+++ b/code/os/03-contextswitch/start.S
@@ -36,6 +36,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/04-multitask/sched.c b/code/os/04-multitask/sched.c
index 1bb3bfd..bac0e02 100644
--- a/code/os/04-multitask/sched.c
+++ b/code/os/04-multitask/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -51,7 +55,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].ra = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/04-multitask/start.S b/code/os/04-multitask/start.S
index c7ad5e4..5bdeff1 100644
--- a/code/os/04-multitask/start.S
+++ b/code/os/04-multitask/start.S
@@ -36,6 +36,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/05-traps/sched.c b/code/os/05-traps/sched.c
index 7e53dd4..fa8359b 100644
--- a/code/os/05-traps/sched.c
+++ b/code/os/05-traps/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -46,7 +50,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].ra = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/05-traps/start.S b/code/os/05-traps/start.S
index c7ad5e4..5bdeff1 100644
--- a/code/os/05-traps/start.S
+++ b/code/os/05-traps/start.S
@@ -36,6 +36,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/06-interrupts/sched.c b/code/os/06-interrupts/sched.c
index 7e53dd4..fa8359b 100644
--- a/code/os/06-interrupts/sched.c
+++ b/code/os/06-interrupts/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -46,7 +50,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].ra = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/06-interrupts/start.S b/code/os/06-interrupts/start.S
index c7ad5e4..5bdeff1 100644
--- a/code/os/06-interrupts/start.S
+++ b/code/os/06-interrupts/start.S
@@ -36,6 +36,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/07-hwtimer/sched.c b/code/os/07-hwtimer/sched.c
index 7e53dd4..fa8359b 100644
--- a/code/os/07-hwtimer/sched.c
+++ b/code/os/07-hwtimer/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -46,7 +50,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].ra = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/07-hwtimer/start.S b/code/os/07-hwtimer/start.S
index c7ad5e4..5bdeff1 100644
--- a/code/os/07-hwtimer/start.S
+++ b/code/os/07-hwtimer/start.S
@@ -36,6 +36,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/08-preemptive/sched.c b/code/os/08-preemptive/sched.c
index 26ac265..7567eda 100644
--- a/code/os/08-preemptive/sched.c
+++ b/code/os/08-preemptive/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -49,7 +53,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].pc = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/08-preemptive/start.S b/code/os/08-preemptive/start.S
index 0b730a9..13cc266 100644
--- a/code/os/08-preemptive/start.S
+++ b/code/os/08-preemptive/start.S
@@ -46,6 +46,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/09-lock/sched.c b/code/os/09-lock/sched.c
index 26ac265..7567eda 100644
--- a/code/os/09-lock/sched.c
+++ b/code/os/09-lock/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -49,7 +53,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].pc = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/09-lock/start.S b/code/os/09-lock/start.S
index 0b730a9..13cc266 100644
--- a/code/os/09-lock/start.S
+++ b/code/os/09-lock/start.S
@@ -46,6 +46,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/10-swtimer/sched.c b/code/os/10-swtimer/sched.c
index 26ac265..7567eda 100644
--- a/code/os/10-swtimer/sched.c
+++ b/code/os/10-swtimer/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -49,7 +53,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].pc = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/10-swtimer/start.S b/code/os/10-swtimer/start.S
index 0b730a9..13cc266 100644
--- a/code/os/10-swtimer/start.S
+++ b/code/os/10-swtimer/start.S
@@ -46,6 +46,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks
 
diff --git a/code/os/11-syscall/sched.c b/code/os/11-syscall/sched.c
index 26ac265..7567eda 100644
--- a/code/os/11-syscall/sched.c
+++ b/code/os/11-syscall/sched.c
@@ -5,7 +5,11 @@ extern void switch_to(struct context *next);
 
 #define MAX_TASKS 10
 #define STACK_SIZE 1024
-uint8_t task_stack[MAX_TASKS][STACK_SIZE];
+/*
+ * In the standard RISC-V calling convention, the stack pointer sp
+ * is always 16-byte aligned.
+ */
+uint8_t __attribute__((aligned(16))) task_stack[MAX_TASKS][STACK_SIZE];
 struct context ctx_tasks[MAX_TASKS];
 
 /*
@@ -49,7 +53,7 @@ void schedule()
 int task_create(void (*start_routin)(void))
 {
 	if (_top < MAX_TASKS) {
-		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE - 1];
+		ctx_tasks[_top].sp = (reg_t) &task_stack[_top][STACK_SIZE];
 		ctx_tasks[_top].pc = (reg_t) start_routin;
 		_top++;
 		return 0;
diff --git a/code/os/11-syscall/start.S b/code/os/11-syscall/start.S
index ba34f12..75d7808 100644
--- a/code/os/11-syscall/start.S
+++ b/code/os/11-syscall/start.S
@@ -70,6 +70,9 @@ park:
 	wfi
 	j	park
 
+	# In the standard RISC-V calling convention, the stack pointer sp
+	# is always 16-byte aligned.
+.align 16
 stacks:
 	.skip	STACK_SIZE * MAXNUM_CPU # allocate space for all the harts stacks