fix bug: 'cargo qemu --ch 2 --release' is hung

[chyyuu]

在 release 模式下，cargo qemu --ch 2 --release 卡住的原因有两个：

1. kernel-context/src/lib.rs 中的 execute 函数

原代码：

core::arch::asm!(
    "csrrw {sscratch}, sscratch, {sscratch}  // 交换后 {sscratch} 寄存器变成旧的 sscratch 值
     ...
     csrw sscratch, {sscratch}               // 用错误的值恢复 sscratch
     csrr {sepc}, sepc
    ",
    sscratch = in(reg) self,                 // 编译器在 asm 后仍用这个寄存器访问 self.sepc
    sepc = inlateout(reg) self.sepc,         // 但寄存器已经不是 self 了！
);

在 csrrw 指令后，存放 self 的寄存器被交换成了旧的 sscratch 值。但编译器在 asm 块后仍然用该寄存器去写 self.sepc，导致写入到错误的内存地址。

修复：使用独立的寄存器保存上下文指针和旧 sscratch 值：

        // 保存 self 指针和 sepc，避免 release 模式下 csrrw 破坏寄存器后的问题
        let ctx_ptr = self as *mut Self;
        let mut sepc = self.sepc;
        let old_sscratch: usize;
        core::arch::asm!(
            "   csrrw {old_ss}, sscratch, {ctx}
                ...
                csrw  sscratch, {old_ss}
                csrr  {sepc}   , sepc
            ",
            ctx           = in       (reg) ctx_ptr,
            old_ss        = out      (reg) old_sscratch,
            sepc          = inlateout(reg) sepc,
            ...
        );
        (*ctx_ptr).sepc = sepc;

2. ch2/src/main.rs 中的用户栈初始化

原代码 let mut user_stack = [0usize; 256]; 在 release 模式下，2048 字节的零初始化数组会触发某些优化问题。

修复：使用 MaybeUninit 跳过零初始化：

        // 设置用户栈（使用 MaybeUninit 避免 release 模式下零初始化的问题）
        let mut user_stack: core::mem::MaybeUninit<[usize; 256]> = core::mem::MaybeUninit::uninit();
        let user_stack_ptr = user_stack.as_mut_ptr() as *mut usize;
        *ctx.sp_mut() = unsafe { user_stack_ptr.add(256) } as usize;

[Copilot]

Pull request overview

This PR fixes a critical bug that causes cargo qemu --ch 2 --release to hang in release mode. The issue stems from aggressive compiler optimizations that break assumptions in inline assembly and stack initialization.

Key changes:

• Fixed register corruption in LocalContext::execute() by using separate variables for context pointer and old sscratch value
• Added #[inline(never)] to prevent inlining issues with inline assembly
• Replaced zero-initialized user stack with MaybeUninit to avoid optimization issues

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
kernel-context/src/lib.rs	Fixed register corruption bug in execute() function by separating context pointer from sscratch register and preventing function inlining
ch2/src/main.rs	Changed user stack from zero-initialized array to MaybeUninit and added black_box to prevent compiler optimizations

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The variable old_sscratch is already used in the inline assembly as an output operand (old_ss), so this line is redundant. The compiler won't produce an "unused variable" warning for variables that are used in inline assembly operands.

Suggested change

let _ = old_sscratch; // suppress unused warning