また夏を追う

最近的笔记

Machine Boot

✏️ 🔧 11 min read

Envrionment Setting

My Computer information (which is required):

  • OS: Linux x86_64 (6.1.53-1-MANJARO)
  • CPU: 13th Gen Intel i5-13500H (16) @ 4.700GHz
  • GPU: Intel Raptor Lake-P [Iris Xe Graphics]
  • Memory: 16GB
  • Editor: VS Code
  • C Compiler: gcc(version 13.2.1) clang(version 16.0.6)
  • C++ Compiler: g++(version 13.2.1) clang++(version 16.0.6)
  • C/C++ build tools: cmake(version 3.27.6) xmake(version 2.8.3) make(version 4.4.1)
  • Rust: rustc 1.73.0-nightly
  • Python: Python 3.11.5
  • Java: openjdk 21 2023-09-19
  • Go: version go1.21.1 linux/amd64

How to build

Run command below in your shell:

Terminal window
git clone https://gitee.com/ipads-lab/chcore-lab-v2.git
cd chcore-lab-v2
make build

and then it will pull a image from docker hub, which is called chcore_builder.

If error like “failed to add the host (veth28449b0) ⇐> sandbox (veth45f4af9) pair interfaces: operation not supported.” occurs, try to restart your computer or docker

Now, you’ve already build a file named “kernel.img” in directory “build/“.

How to run

Simply run :

Terminal window
make qemu

to run Chcore

Note that there is no output, which is normal. It won’t give any output until you finish lab1

How to debug

You need install gdb-multiarch to debug. For example, in my manjaro, run :

Terminal window
yay -S gdb-multiarch

If you’re in Ubuntu, just run :

sudo apt-get install gdb-multiarch

After installing, using two bash(or anything else such as tmux), one run :

Terminal window
make qemu-gdb

to activate qemu’s gdb mode, the other run :

Terminal window
make gdb

to connect the remote target(i.e. the qemu’s gdb mode).

If you see this, then it’s success.

How to edit code

I strongly recommand VS Code, just cd into this directory and type code ., then you can use it to edit the code.

Machine Boot

The lab document is in docs/, follow it to finish the lab.

Pre

Modify Makefile like this :

qemu: build
  $(V)$(_QEMU) $(QEMU_OPTS)


qemu-gdb: build
  $(V)$(_QEMU) -S -gdb tcp::$(QEMU_GDB_PORT) $(QEMU_OPTS)

That would avoid run code but do not update it

Q1

阅读  _start  函数的开头，尝试说明 ChCore 是如何让其中一个核首先进入初始化流程，并让其他核暂停执行的。

  1. Read the cpu id from mpidr_el1 register to the x8 register
  2. Get the low 8-bit value of the register x8
  3. Check if low 8-bit is 0, if true, then jump to function primary, else, make a infnity loop to build a fence to stop others cores at this instruction
  4. function primary will enter a initial procedure, including set the exception level to EL1 and stack size, transfer control to the kernel

for more: register mpidr_el1 store the information of core, in multi-processor system, every processor has a unique value of mpidr_el1.

M2

练习题 2：在  arm64_elX_to_el1  函数的  LAB 1 TODO 1  处填写一行汇编代码，获取 CPU 当前异常级别。

mrs x9, CurrentEL

M3

Look at label .Lno_gic_sr, that is how to jump to el1 from el2.

So we write code like this :

  adr x9, .Ltarget
  msr elr_el3, x9
  mov x9, SPSR_ELX_DAIF | SPSR_ELX_EL1H
  msr spsr_el3, x9

to ensure our implementation is right, we use gdb below :

we will see that it jump back _start from function arm64_elX_to_el1

Q4

结合此前 ICS 课的知识，并参考  kernel.img  的反汇编（通过  aarch64-linux-gnu-objdump -S  可获得），说明为什么要在进入 C 函数之前设置启动栈。如果不设置，会发生什么？

We can see the assembly below :

   80010:  94001ffc   bl  88000 <arm64_elX_to_el1>
   80014:  580000a0   ldr  x0, 80028 <primary+0x18>
   80018:  91400400   add  x0, x0, #0x1, lsl #12
   8001c:  9100001f   mov  sp, x0
   80020:  940020e2   bl  883a8 <init_c>

note that =boot_cpu_stack is *0x80028(i.e. 0x00088980), and INIT_STACK_SIZE is 0x1000(i.e. 0x1, lsl #12)

Because init_c is a function which require a stack frame. That’s why we should initial the stack pointer and allocate stack space.

more details in rCore

Q5

在实验 1 中，其实不调用  clear_bss  也不影响内核的执行，请思考不清理  .bss  段在之后的何种情况下会导致内核无法工作。

Anser from GPT:

当不清零 .bss 段时，未显式初始化的全局变量和静态变量的初始值将是未定义的。这可能导致以下问题和错误：

  1. 未定义的行为：访问未初始化的变量可能导致未定义的行为。这包括读取随机的内存值、产生不可预测的结果以及可能导致程序崩溃或错误的行为。
  2. 依赖于零值的假设：在程序设计中，有时会假设未初始化的全局变量或静态变量的初始值为零。如果这些变量的初始值不是零，那么这些假设就会被破坏，可能导致程序逻辑错误或不一致。
  3. 安全问题：如果未初始化的变量被认为是零值，而实际上它们存储了其他敏感信息，可能会引发安全问题。例如，如果未初始化的变量被用于存储密码、密钥或其他敏感数据，那么不清零可能导致这些敏感信息泄露。
  4. 跨平台兼容性问题：不同的编译器和操作系统对未初始化变量的处理可能不同。在某些平台上，编译器或操作系统可能为未初始化变量提供默认的零值，而在其他平台上可能不会。因此，不清零 .bss 段可能导致在不同的平台上出现不一致的行为，影响跨平台的兼容性。

M6

void uart_send_string(char *str)
{
        /* LAB 1 TODO 3 BEGIN */


        while (*str != '\0') {
                early_uart_send((unsigned int)(*str));
                str++;
        }


        /* LAB 1 TODO 3 END */
}

M7

I wrote this in first

orr x8, x8, #0x82f

which enable all flags.

However, we can see that below here:

  bic     x8, x8, #SCTLR_EL1_A
  bic     x8, x8, #SCTLR_EL1_SA0
  bic     x8, x8, #SCTLR_EL1_SA
  orr     x8, x8, #SCTLR_EL1_nAA

it reminds us that we should just enable one flag here, i.e. flag M, so the correct code is :

orr x8, x8, #SCTLR_EL1_M

How to Grade

Remember to install expect， in Manjaro you can :

Terminal window
yay -S expect

then type :

Terminal window
make grade

you will see something like this:

🕸️ 关系图谱