mmap代码在这里。这是完成的最后一个实验了，因为网络部分在计算机网络实验中有所涉及，就没再做了。只能说，完结撒花吧！感谢xv6陪伴的这一个学期。
写在前面
都最后一个实验了，相关参考资料就不多提了吧~还是提一下：
我的博客OS实验xv6 6.S081 开坑中给出了一些有用的参考资料，大家可以参考。
实验介绍
老样子，先给实验指导书。
本次实验主要完成两个任务：
- 简单的
mmap - 简单的
munmap
开始！
什么是mmap呢？在Linux Shell中输入man mmap后会得到下述解答：
mmap主要的工作就是将文件描述符fd对应的文件映射到进程的虚拟内存中，映射的起始地址从addr开始，如果addr为0，则表明kernel应该自行为其分配地址，length表明映射的长度，prot表明文件的权限，本次实验只要求实现PROT_READ和PROT_WRITE，flags代表映射的类型，有很多类型，但在本次实验中我们只需要处理两种类型：MAP_SHARED和MAP_PRIVATE。其中可以简单的理解为MAP_SHARED要将映射到内存中的文件回写，而MAP_PRIVATE则不。
如何完成mmap呢？从Hints入手：
- Start by adding mmap and munmap system calls and associated flags (PROT_READ etc), in order to get user/mmaptest.c to compile. For now, just return errors from mmap and munmap. Run mmaptest, which will fail at the first mmap call.
- Fill in the page table lazily, in response to page faults. That is, mmap should not allocate physical memory or read the file. Instead, do that in page fault handling code in (or called by) usertrap, as in the lazy page allocation lab. The reason to be lazy is to ensure that mmap of a large file is fast, and that mmap of a file larger than physical memory is possible.
- Keep track of what mmap has mapped for each process. Define a structure corresponding to the VMA (virtual memory area) described in Lecture 15, recording the address, length, permissions, file, etc. for a virtual memory range created by mmap. Since the xv6 kernel doesn’t have a memory allocator in the kernel, it’s OK to declare a fixed-size array of VMAs and allocate from that array as needed.
- Implement mmap: find an unused region in the process’s address space in which to map the file, and add a VMA to the process’s table of mapped regions. The VMA should contain a pointer to a struct file for the file being mapped; mmap should increase the file’s reference count so that the structure doesn’t disappear when the file is closed (hint: see filedup). Run mmaptest: the first mmap should succeed, but the first access to the mmap-ed memory will cause a page fault and kill mmaptest.
- Add code to cause a page-fault in a mmap-ed region to allocate a page of physical memory, read 4096 bytes of the relevant file into that page, and map it into the user address space. Read the file with readi, which takes an offset argument at which to read in the file (but you will have to lock/unlock the inode passed to readi). Don’t forget to set the permissions correctly on the page. Run mmaptest; it should get to the first munmap.
- Implement munmap: find the VMA for the address range and unmap the specified pages (hint: use uvmunmap). If munmap removes all pages of a previous mmap, it should decrement the reference count of the corresponding struct file. If an unmapped page has been modified and the file is mapped MAP_SHARED, write the page back to the file. Look at filewrite for inspiration.
- Ideally your implementation would only write back MAP_SHARED pages that the program actually modified. The dirty bit (D) in the RISC-V PTE indicates whether a page has been written. However, mmaptest does not check that non-dirty pages are not written back; thus you can get away with writing pages back without looking at D bits.
- Modify exit to unmap the process’s mapped regions as if munmap had been called. Run mmaptest; mmap_test should pass, but probably not fork_test.
- Modify fork to ensure that the child has the same mapped regions as the parent. Don’t forget to increment the reference count for a VMA’s struct file. In the page fault handler of the child, it is OK to allocate a new physical page instead of sharing a page with the parent. The latter would be cooler, but it would require more implementation work. Run mmaptest; it should pass both mmap_test and fork_test.
下面，我们开始逐一解读这些Hints的含义：
- 为
mmap和munmap添加系统调用，并添加一些相关的标志位，例如PROT_READ等； mmap会返回映射到虚拟空间的起始地址，但该地址并没有进行页映射（mappages），也就是说mmap并不做内存映射的工作，因此当用户进程访问到这地址时会出现缺页异常，我们需要在trap.c中捕获该异常，并为其懒加载一个页面，将文件的内容写进去，这才真正完成了文件到内存的映射；- 要跟踪
mmap映射的每一块内存，这需要定义一个VMA数据结构，VMA要能够记录内存块的起始地址、长度、权限、相关文件等信息； - 实现
mmap，要在用户进程的地址空间中找到一块空闲的空间用于内存映射，这就需要好好设计VMA和proc； - 利用
readi将文件的内容读取到相应的内存地址中； - 实现
munmap，用uvmunmap释放页面，并查看当前flags，如果是MAP_SHARED，还应该将内存的内容回写到文件中； - 不必考察写脏位D；
- 修改
exit函数，使之能够像munmap一样释放所有mmap的区域； - 修改
fork，保证子进程和父进程有一样的映射区域。
首先是syscall的实现，这个就不再做赘述了。我们就从设计VMA开始吧，具体的VMA该如何设计是参考了Linux中的VMA
/**
*
* Define a structure corresponding to the VMA (virtual memory area) described in Lecture 15,
* recording the address, length, permissions, file, etc. for a virtual memory range created by mmap
*
* RETRIVEFROM:
* https://sites.google.com/site/knsathyawiki/example-page/chapter-15-the-process-address-space#TOC-Virtual-Memory-Areas
*/
struct VMA
{
/** 说明VMA是否可用，1为可用，0为已被占用 */
int vm_valid;
uint64 vm_start;
uint64 vm_end;
/** Flags */
int vm_flags;
/** 页面权限，可写？可读？ */
int vm_prot;
/** 指向某个文件 */
struct file* vm_file;
/** 文件描述符 */
int vm_fd;
};
接下来，我们修改proc中的内容，使之能够管理VMA，这里我们定义最多管理NVMA=100个VAM，代码如下：
/** Implementation of MMAP */
/** VMA管理数组 */
struct VMA vmas[NVMA];
好了，接下来就需要好好设计一下映射模式了：
How to find an unused region in the process’s address space in which to map the file？
我们将实现拉回到用户进程的内存布局中，如下图所示。可以看到，heap段便是我们可用的内存空间。我们可以让映射内存从trapframe之下开始向下增长。
如何记录当前已经增长到哪个位置了呢？我们则需要一个当前最大VMA地址指针来记录。最终方案如下图所示：
思路既成，那么实现起来就非常快了。首先在proc中添加字段：
/** 当前可用的最大的虚拟地址 从上往下*/
uint64 current_maxva;
接下来在allocproc中初始化添加的proc字段。这里要注意的是，xv6中提到：trampoline与trapframe均占用一个页面，因此初始的current_maxva为PGROUNDDOWN(MAXVA - 2 * PGSIZE)
#define VMASTART PGROUNDDOWN(MAXVA - 2 * PGSIZE)
static struct proc*
allocproc(void)
{
......
/** Implementation of MMAP */
for (int i = NVMA - 1; i >= 0; i--)
{
p->vmas[i].vm_valid = 1;
}
p->current_maxva = VMASTART;
return p;
}
下面开始正式实现mmap。基本思路是：首先找到第一个可用的VMA，然后再根据current_maxva和传入的参数为其设置各种字段。需要注意的是，vm_start应该小于vm_end，因为在之后的trap.c中我们会利用kalloc分配内存，kalloc分配的内存是向上增长的，而我们想要的正是让内存从vm_start向上增长。
uint64
sys_mmap(void){
uint64 addr;
int length;
int prot;
int flags;
int fd;
struct file* f;
int offset;
if(argaddr(0, &addr) < 0 || argint(1, &length) < 0 ||
argint(2, &prot) < 0 || argint(3, &flags) < 0 ||
argfd(4, &fd, &f) < 0 || argint(5, &offset) < 0){
return -1;
}
if(!f->writable && (prot & PROT_WRITE) && (flags & MAP_SHARED)){
return -1;
}
/**
*
* you can assume that addr will always be zero,
* meaning that the kernel should decide the virtual address at which to map the file
*
* offset it's the starting point in the file at which to map
*/
struct proc* p;
p = myproc();
struct VMA* vma = 0;
/** 从上往下找到第一个可用的vma */
for (int i = NVMA - 1; i >= 0; i--)
{
if(p->vmas[i].vm_valid){
vma = &p->vmas[i];
/** 置当前的imaxvma为i */
p->current_imaxvma = i;
break;
}
}
/**
* 1. VMA：START在下方
* 原因：kalloc是分配内存向上增长，因此start要在下方
* ->current_maxva 0x001xx END
* ............
* 0x000xx START
*
* ----------------------------------
* 2. 更新current_maxva
* 0x001xx END
* ............
* ->current_maxva 0x000xx START
* ----------------------------------
*/
if(vma){
/** 记得这里要用uint64，否则会做最高位拓展 */
printf("sys_mmap(): %p, length: %d\n",p->current_maxva, length);
uint64 vm_end = PGROUNDDOWN(p->current_maxva);
uint64 vm_start = PGROUNDDOWN(p->current_maxva - length);
printf("vm_start(): %p, vm_end: %p\n",vm_start, vm_end);
vma->vm_valid = 0;
vma->vm_fd = fd;
vma->vm_file = f;
vma->vm_flags = flags;
vma->vm_prot = prot;
vma->vm_end = vm_end;
vma->vm_start = vm_start;
/**
* mmap should increase the file's reference count
* so that the structure doesn't disappear when the file is closed (hint: see filedup).
*/
vma->vm_file->ref++;
p->current_maxva = vm_start;
}
else
{
return -1;
}
return vma->vm_start;
}
下面便转战trap.c了，四个关键：第一，捕获缺页异常；第二，懒加载一个页面；第三，根据出错的虚拟地址找到相应的VMA；第四，通过VMA中保存的file并利用readi将文件内容读取到页面内。代码如下：
if(r_scause() == 13 || r_scause() == 15){
/**
* Implement Lazy allocation for mmap
*
* REASON: That is, mmap should not allocate physical memory or read the file
* */
struct proc* p = myproc();
uint64 va = PGROUNDDOWN(r_stval());
printf("MAXVA: %p, va: %p, current_max: %p\n",MAXVA, va, p->current_maxva);
/** 找到虚拟地址对应的vma */
struct VMA* vma = 0; /* = &p->vmas[p->current_ivma]; */
for (int i = NVMA; i >= 0; i--)
{
if(p->vmas[i].vm_start <= va && va <= p->vmas[i].vm_end){
vma = &p->vmas[i];
break;
}
}
if(vma == 0){
printf("usertrap(): not find vma \n");
p->killed = 1;
goto end;
}
if(va > vma->vm_end){
printf("usertrap(): va is greater than vm_end \n");
p->killed = 1;
goto end;
}
/** 内存向上增长 */
char* mem = (char *)kalloc();
if(mem == 0){
printf("usertrap(): no mem left\n");
p->killed = 1;
goto end;
}
printf("walk va %p result : %d \n",va, walkaddr(p->pagetable, va));
memset(mem, 0, PGSIZE);
/** Don't forget to set the permissions correctly on the page */
if(mappages(p->pagetable, va, PGSIZE, (uint64)mem, vma->vm_prot|PTE_U|PTE_X) < 0){
printf("usertrap(): cannot map\n");
kfree(mem);
p->killed = 1;
goto end;
}
/** 利用readi将文件内容映射到虚拟地址上，映射的文件开始地址偏移为 va - vma->vm_start */
struct file* f = vma->vm_file;
int offset = va - vma->vm_start;
ilock(f->ip);
//1表示读入用户内存，
readi(f->ip, 1, va, offset, PGSIZE);
iunlock(f->ip);
}
else
{
printf("usertrap(): unexpected scause %p (%s) pid=%d\n", r_scause(), scause_desc(r_scause()), p->pid);
printf(" sepc=%p stval=%p\n", r_sepc(), r_stval());
p->killed = 1;
}
}
好，现在我们已经完成了mmap，接下来该实现munmap了。munmap就是释放mmap的空间，当发现mmap的flages是MAP_SHARED时我们要回写文件，否则不用（本次实验中是这样的）。另外我们我们还需要考虑一个问题，那就是内部碎片的问题，设想下面的情况。VMA[1]被释放了，但是我们的current_maxva并不能移动。那么在此后再进行mmap时，按照我们的算法这块被释放的内存将永远不会被用到了。
为了解决这个问题，我们需要在释放后进行内存紧缩的操作，如下图所示：
代码实现如下：
uint64
sys_munmap(void){
uint64 addr;
int length;
if(argaddr(0, &addr) < 0 || argint(1, &length) < 0){
return -1;
}
printf("### sys_munmap: \n");
printf("addr: %p, length:%d, current:%p\n", addr, length, myproc()->current_maxva);
struct proc* p = myproc();
for (int i = NVMA - 1; i >= 0; i--)
{
if(p->vmas[i].vm_start <= addr && addr <= p->vmas[i].vm_end){
struct VMA* vma = &p->vmas[i];
/**
* 1. If an unmapped page has been modified and the file is mapped MAP_SHARED,
* write the page back to the file. Look at filewrite for inspiration.
*
*
* 2. However, mmaptest does not check that non-dirty pages are not written back;
* thus you can get away with writing pages back without looking at D bits.
*
*
* 1. unmap的时候，只会从一个vma的起始开始，因此，可以默认p->vmas[i].vm_start = addr，因此
* 我们后面vma->vm_start += length操作。
* 2. 就是说，碰到MAP_SHARED就回写，不用理会“写脏D位”
*
* 3. 指针current_maxva基于current_imaxvma紧缩，这样是一个折中的办法，而不是一直向下增长
*
* */
/** 首先要判断 */
if(walkaddr(p->pagetable, vma->vm_start)){
if(vma->vm_flags == MAP_SHARED){
printf("sys_munmap(): write back \n");
/** 回写文件 */
filewrite(vma->vm_file, vma->vm_start, length);
}
uvmunmap(p->pagetable, vma->vm_start, length ,1);
}
vma->vm_start += length;
printf("vma_start: %p, vma_end: %p\n", vma->vm_start, vma->vm_end);
if(vma->vm_start == vma->vm_end){
vma->vm_file->ref--;
/** 置该块可用 */
vma->vm_valid = 1;
}
/** Shrink */
int j;
/** 紧缩 p->current_maxva */
for (j = p->current_imaxvma; j < NVMA; j++)
{
if(!p->vmas[j].vm_valid){
p->current_maxva = p->vmas[j].vm_start;
p->current_imaxvma = j;
break;
}
}
if(j == NVMA){
p->current_maxva = VMASTART;
}
return 0;
}
}
return -1;
}
然后，修改exit，与munmap的实现基本一样，不过因为它是释放所有VMA及其映射的内存，因此不需要考虑内存紧缩问题，代码如下：
void
exit(int status)
{
struct proc *p = myproc();
if(p == initproc)
panic("init exiting");
// Close all open files.
for(int fd = 0; fd < NOFILE; fd++){
if(p->ofile[fd]){
struct file *f = p->ofile[fd];
fileclose(f);
p->ofile[fd] = 0;
}
}
/**
* Unmap all the maped regions
* 从下往上回收
* */
struct VMA* vma;
for (int i = 0; i < NVMA; i++)
{
if(!p->vmas[i].vm_valid){
vma = &p->vmas[i];
vma->vm_valid = 1;
int totsz = vma->vm_end - vma->vm_start;
if(walkaddr(p->pagetable, vma->vm_start)){
if(vma->vm_flags == MAP_SHARED){
printf("sys_munmap(): write back \n");
filewrite(vma->vm_file, vma->vm_start, totsz);
}
uvmunmap(p->pagetable, vma->vm_start, totsz,1);
}
vma->vm_start += totsz;
if(vma->vm_start == vma->vm_end){
vma->vm_file->ref--;
}
}
}
p->current_maxva = VMASTART;
.......
}
最后，完成fork时proc中添加字段的复制：
int
fork(void)
{
int i, pid;
struct proc *np;
struct proc *p = myproc();
// Allocate process.
if((np = allocproc()) == 0){
return -1;
}
// Copy user memory from parent to child.
if(uvmcopy(p->pagetable, np->pagetable, p->sz) < 0){
freeproc(np);
release(&np->lock);
return -1;
}
np->sz = p->sz;
np->parent = p;
/** Modify fork to ensure that the child has the same mapped regions as the parent. */
np->current_maxva = p->current_maxva;
np->current_imaxvma = p->current_imaxvma;
for (int i = NVMA - 1; i >= 0; i--)
{
if(p->vmas[i].vm_file)
p->vmas[i].vm_file->ref++;
np->vmas[i].vm_end = p->vmas[i].vm_end;
np->vmas[i].vm_fd = p->vmas[i].vm_fd;
np->vmas[i].vm_file = p->vmas[i].vm_file;
np->vmas[i].vm_flags = p->vmas[i].vm_flags;
np->vmas[i].vm_prot = p->vmas[i].vm_prot;
np->vmas[i].vm_start = p->vmas[i].vm_start;
np->vmas[i].vm_valid = p->vmas[i].vm_valid;
}
......
}
好了，make grade测试结果如下。
这次的usertests运行得也太快了吧，心里不踏实，看一下xv6.out：
xv6 kernel is booting
virtio disk init 0
init: starting sh
$ usertests
usertests starting
test reparent2: OK
test pgbug: OK
test sbrkbugs: usertrap(): unexpected scause 0x000000000000000c (instruction page fault) pid=3207
sepc=0x00000000000044b0 stval=0x00000000000044b0
usertrap(): unexpected scause 0x000000000000000c (instruction page fault) pid=3208
sepc=0x00000000000044b0 stval=0x00000000000044b0
OK
test badarg: OK
test reparent: OK
test twochildren: OK
test forkfork: OK
test forkforkfork: OK
test argptest: OK
test createdelete: OK
test linkunlink: OK
test linktest: OK
test unlinkread: OK
test concreate: OK
test subdir: OK
test fourfiles: OK
test sharedfd: OK
test exectest: OK
test bigargtest: OK
test bigwrite: OK
test bsstest: OK
test sbrkbasic: OK
test sbrkmuch: OK
test kernmem: MAXVA: 0x0000004000000000, va: 0x0000000080000000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008000c000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080018000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080024000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080030000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008003d000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080049000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080055000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080061000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008006d000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008007a000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080086000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080092000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008009e000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800aa000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800b7000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800c3000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800cf000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800db000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800e7000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000800f4000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080100000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008010c000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080118000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080124000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080131000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008013d000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080149000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080155000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080162000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008016e000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008017a000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080186000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x0000000080192000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x000000008019f000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000801ab000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000801b7000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000801c3000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000801cf000, current_max: 0x0000003fffffe000
usertrap(): not find vma
MAXVA: 0x0000004000000000, va: 0x00000000801dc000, current_max: 0x0000003fffffe000
usertrap(): not find vma
OK
test sbrkfail: MAXVA: 0x0000004000000000, va: 0x0000000000010000, current_max: 0x0000003fffffe000
usertrap(): not find vma
OK
test sbrkarg: OK
test validatetest: OK
test stacktest: MAXVA: 0x0000004000000000, va: 0x000000000000d000, current_max: 0x0000003fffffe000
usertrap(): not find vma
OK
test opentest: OK
test writetest: OK
test writebig: OK
test createtest: OK
test openiput: OK
test exitiput: OK
test iput: OK
test mem: OK
test pipe1: OK
test preempt: kill... wait... OK
test exitwait: OK
test rmdot: OK
test fourteen: OK
test bigfile: OK
test dirfile: OK
test iref: OK
test forktest: OK
test bigdir: OK
ALL TESTS PASSED
$ qemu-system-riscv64: terminating on signal 15 from pid 12749 (make)
应该是没啥问题了！OK，起飞🛫！
1404
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
