跳转至

Lab3

Page Table

函数 作用
copyin copy data to user space
copyout copy data from user space
mappages install PTE for new mappings
walk find PA for VA

XV6 Book

  • 从3.5 开始， 都没读懂, 后续补上: <13-04-23, David Chen> -

implement pageing in practice

in main.c , it calls kvminit, and kvminit calls kvmmake. kvmmake returns a pointer (kernel_pagetable).

kvmmake first init allocate a PAGE for root pagetable, and then call mappages to allocate memory for hardware (At this point, paging is not started, so virtual address is directed mapped to physical address)

kvmmap calls mappages , mappages first align the start address and end address, and start allocate PA for VA at page intervals

  • mappages function makes me confusing
  • what is macro PAGEROUNDDOWN means
    • I guess this macro is used to align the size of a and size.

lab

Speed up system calls

某些现代操作系统通过一个只读区域来在用户与内核之间共享数据，以加速系统调用的执行速度。这样的策略能够减少系统在地址转换上花费的时间。为了帮助你更好的理解如何向 page table 中插入映射，你的第一个任务就是为 getpid 这个系统调用实现这样的优化。

Solution

The life cycle of a xv6 process is as follows.

We can call allocproc to create a new process. allocproc first use kalloc to allocate physical space for p->trapframe and p->usyscall .

  if((p->usyscall = (struct usyscall *)kalloc()) == 0){
    freeproc(p);
    release(&p->lock);
    return 0;
  }

  p->usyscall->pid = p->pid;

Then allocproc call proc_pagetabl to initialize the pagetable for the process.

proc_pagetable use uvmmap to map virtual memory to corresponding physical memory.

  // map a page for speed up syscall
  if(mappages(pagetable, USYSCALL, PGSIZE,
              (uint64)(p->usyscall), PTE_R) < 0){
    uvmunmap(pagetable, USYSCALL, 1, 0);
    uvmfree(pagetable, 0);
    return 0;
  }

  return pagetable;
}

We can call freeproc to free an running process. freeproc first use kfree to free process memory, Then call proc_freepagetabl to initialize free pagetable for the process.

  if(p->usyscall)
    kfree((void*)p->usyscall);
  p->usyscall = 0;

proc_pagetable use uvmummap to cancel mapping for virtual memory to corresponding physical memory.

And then call uvmfree to free pagetable.

void
proc_freepagetable(pagetable_t pagetable, uint64 sz)
{
  uvmunmap(pagetable, TRAMPOLINE, 1, 0);
  uvmunmap(pagetable, TRAPFRAME, 1, 0);
  uvmunmap(pagetable, USYSCALL, 1, 0);
  uvmfree(pagetable, sz);
}

Detecting which page have been accessed

and according to the PA, find corresponding VA for kernel. But How to do this. Or we insert a new mapping for this copyin and copyout confused me. I guess they work as follows. copyin use process pagetable to find PA of the argument

实现的时候，只需通过 walkaddr 找到 用户虚拟地址对应的 物理地址，再对物理地址进行操作即可。

How to write content to corresponding addr

  1. First we need to define PTE_A in riscv.h
  2. add stub for walk in def.h
  3. implement sys_pgaccess in sysproc.c
    • 遍历就好

注意我们不需要手动设置 PTE_A 位，硬件帮我们做好了 

int sys_pgaccess()
{
    // lab pgtbl: your code here.

  uint64 va0;
  int num;
  uint64 usr_cnt;

  // 从 trapframe 获取参数
  if(argaddr(0, &va0) > 0) {
    return -1;
  }

  if(argint(1, &num) > 0) {
    return -1;
  }

  if(argaddr(2, &usr_cnt) > 0) {
    return -1;
  } 

  uint64 k_cnt = 0;

  // 向下取整，保证地址对齐

  va0 = PGROUNDDOWN(va0);
  struct proc* p = myproc();

  // 从 va0 -> va0 + num * pagesize 遍历 页表
  // 查看其 PTE_A 是否为 1
  // 为 1 的话，置结果第 1 << num 位 为 1，
  // 再将 access 位 置 0
  for (int i = 0; i < num; i++) {
    pte_t* pte = walk(p->pagetable, va0, 0);
    if (*pte & PTE_A) {
      k_cnt |= (1 << i);
      *pte = *pte & (~PTE_A);
    }
    va0 += PGSIZE;
  }

  copyout(p->pagetable, usr_cnt, (char* )&k_cnt, sizeof(uint64) / sizeof(char));

  return 0;
}

<++>

```json

颜色主题调整

评论区~