You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
riscv-pke/kernel/rfs.c

744 lines
23 KiB

/*
* RFS (Ramdisk File System) is a customized simple file system installed in the
* RAM disk. added @lab4_1.
* Layout of the file system:
*
* ******** RFS MEM LAYOUT (112 BLOCKS) ****************
* superblock | disk inodes | bitmap | free blocks *
* 1 block | 10 blocks | 1 | 100 *
* *****************************************************
*
* The disk layout of rfs is similar to the fs in xv6.
*/
#include "rfs.h"
#include "pmm.h"
#include "ramdev.h"
#include "spike_interface/spike_utils.h"
#include "util/string.h"
#include "vfs.h"
/**** vinode inteface ****/
const struct vinode_ops rfs_i_ops = {
.viop_read = rfs_read,
.viop_write = rfs_write,
.viop_create = rfs_create,
.viop_lseek = rfs_lseek,
.viop_disk_stat = rfs_disk_stat,
.viop_lookup = rfs_lookup,
.viop_readdir = rfs_readdir,
.viop_mkdir = rfs_mkdir,
.viop_write_back_vinode = rfs_write_back_vinode,
.viop_hook_opendir = rfs_hook_opendir,
.viop_hook_closedir = rfs_hook_closedir,
};
/**** rfs utility functions ****/
//
// register rfs to the fs list supported by PKE.
//
int register_rfs() {
struct file_system_type *fs_type = (struct file_system_type *)alloc_page();
fs_type->type_num = RFS_TYPE;
fs_type->get_superblock = rfs_get_superblock;
for (int i = 0; i < MAX_SUPPORTED_FS; i++) {
if (fs_list[i] == NULL) {
fs_list[i] = fs_type;
return 0;
}
}
return -1;
}
//
// format "dev" with rfs. note the "dev" should be a ram disk device.
//
int rfs_format_dev(struct device *dev) {
struct rfs_device *rdev = rfs_device_list[dev->dev_id];
// ** first, format the superblock
// build a new superblock
struct super_block *super = (struct super_block *)rdev->iobuffer;
super->magic = RFS_MAGIC;
super->size =
1 + RFS_MAX_INODE_BLKNUM + 1 + RFS_MAX_INODE_BLKNUM * RFS_DIRECT_BLKNUM;
// only direct index blocks
super->nblocks = RFS_MAX_INODE_BLKNUM * RFS_DIRECT_BLKNUM;
super->ninodes = RFS_BLKSIZE / RFS_INODESIZE * RFS_MAX_INODE_BLKNUM;
// write the superblock to RAM Disk0
if (rfs_w1block(rdev, RFS_BLK_OFFSET_SUPER) != 0) // write to device
panic("RFS: failed to write superblock!\n");
// ** second, set up the inodes and write them to RAM disk
// build an empty inode disk block which has RFS_BLKSIZE/RFS_INODESIZE(=32)
// disk inodes
struct rfs_dinode *p_dinode = (struct rfs_dinode *)rdev->iobuffer;
for (int i = 0; i < RFS_BLKSIZE / RFS_INODESIZE; ++i) {
p_dinode->size = 0;
p_dinode->type = R_FREE;
p_dinode->nlinks = 0;
p_dinode->blocks = 0;
p_dinode = (struct rfs_dinode *)((char *)p_dinode + RFS_INODESIZE);
}
// write RFS_MAX_INODE_BLKNUM(=10) empty inode disk blocks to RAM Disk0
for (int inode_block = 0; inode_block < RFS_MAX_INODE_BLKNUM; ++inode_block) {
if (rfs_w1block(rdev, RFS_BLK_OFFSET_INODE + inode_block) != 0)
panic("RFS: failed to initialize empty inodes!\n");
}
// build root directory inode (ino = 0)
struct rfs_dinode root_dinode;
root_dinode.size = 0;
root_dinode.type = R_DIR;
root_dinode.nlinks = 1;
root_dinode.blocks = 1;
root_dinode.addrs[0] = RFS_BLK_OFFSET_FREE;
// write root directory inode to RAM Disk0 (ino = 0)
if (rfs_write_dinode(rdev, &root_dinode, 0) != 0) {
sprint("RFS: failed to write root inode!\n");
return -1;
}
// ** third, write freemap to disk
int *freemap = (int *)rdev->iobuffer;
memset(freemap, 0, RFS_BLKSIZE);
freemap[0] = 1; // the first data block is used for root directory
// write the bitmap to RAM Disk0
if (rfs_w1block(rdev, RFS_BLK_OFFSET_BITMAP) != 0) { // write to device
sprint("RFS: failed to write bitmap!\n");
return -1;
}
sprint("RFS: format %s done!\n", dev->dev_name);
return 0;
}
// ** Note: If you use the following four functions interchangeably,
// ** be sure to watch out for IOBUFFER BEING OVERWRITTEN !!!
//
// call ramdisk_read via the device structure.
// read the "n_block"^th block from RAM disk to the iobuffer of rfs_dev.
//
int rfs_r1block(struct rfs_device *rfs_dev, int n_block) {
return dop_read(rfs_dev, n_block);
}
//
// call ramdisk_write via the device structure.
// write iobuffer of rfs_dev to RAM disk at the "n_block"^th block.
//
int rfs_w1block(struct rfs_device *rfs_dev, int n_block) {
return dop_write(rfs_dev, n_block);
}
//
// read disk inode from RAM disk
//
struct rfs_dinode *rfs_read_dinode(struct rfs_device *rdev, int n_inode) {
int n_block = n_inode / (RFS_BLKSIZE / RFS_INODESIZE) + RFS_BLK_OFFSET_INODE;
int offset = n_inode % (RFS_BLKSIZE / RFS_INODESIZE);
// call ramdisk_read defined in dev.c
if (dop_read(rdev, n_block) != 0) return NULL;
struct rfs_dinode *dinode = (struct rfs_dinode *)alloc_page();
memcpy(dinode, (char *)rdev->iobuffer + offset * RFS_INODESIZE,
sizeof(struct rfs_dinode));
return dinode;
}
//
// write disk inode to RAM disk.
// note: we need first read the "disk" block containing the "n_inode"^th inode,
// modify it, and write the block back to "disk" eventually.
//
int rfs_write_dinode(struct rfs_device *rdev, const struct rfs_dinode *dinode,
int n_inode) {
int n_block = n_inode / (RFS_BLKSIZE / RFS_INODESIZE) + RFS_BLK_OFFSET_INODE;
int offset = n_inode % (RFS_BLKSIZE / RFS_INODESIZE);
// call ramdisk_read defined in dev.c
dop_read(rdev, n_block);
memcpy(rdev->iobuffer + offset * RFS_INODESIZE, dinode,
sizeof(struct rfs_dinode));
// call ramdisk_write defined in dev.c
int ret = dop_write(rdev, n_block);
return ret;
}
//
// allocate a block from RAM disk
//
int rfs_alloc_block(struct super_block *sb) {
int free_block = -1;
// think of s_fs_info as freemap information
int *freemap = (int *)sb->s_fs_info;
for (int block = 0; block < sb->nblocks; ++block) {
if (freemap[block] == 0) { // find a free block
freemap[block] = 1;
free_block = RFS_BLK_OFFSET_FREE + block;
break;
}
}
if (free_block == -1) panic("rfs_alloc_block: no more free block!\n");
return free_block;
}
//
// free a block in RAM disk
//
int rfs_free_block(struct super_block *sb, int block_num) {
int *freemap = (int *)sb->s_fs_info;
freemap[block_num - RFS_BLK_OFFSET_FREE] = 0;
return 0;
}
//
// add a new directory entry to a directory
//
int rfs_add_direntry(struct vinode *dir, const char *name, int inum) {
if (dir->type != DIR_I) {
sprint("rfs_add_direntry: not a directory!\n");
return -1;
}
struct rfs_device *rdev = rfs_device_list[dir->sb->s_dev->dev_id];
int n_block = dir->addrs[dir->size / RFS_BLKSIZE];
if (rfs_r1block(rdev, n_block) != 0) {
sprint("rfs_add_direntry: failed to read block %d!\n", n_block);
return -1;
}
// prepare iobuffer
char *addr = (char *)rdev->iobuffer + dir->size % RFS_BLKSIZE;
struct rfs_direntry *p_direntry = (struct rfs_direntry *)addr;
p_direntry->inum = inum;
strcpy(p_direntry->name, name);
// write the modified (parent) directory block back to disk
if (rfs_w1block(rdev, n_block) != 0) {
sprint("rfs_add_direntry: failed to write block %d!\n", n_block);
return -1;
}
// update its parent dir state
dir->size += sizeof(struct rfs_direntry);
// write the parent dir inode back to disk
if (rfs_write_back_vinode(dir) != 0) {
sprint("rfs_add_direntry: failed to write back parent dir inode!\n");
return -1;
}
return 0;
}
//
// alloc a new (and empty) vinode
//
struct vinode *rfs_alloc_vinode(struct super_block *sb) {
struct vinode *vinode = default_alloc_vinode(sb);
vinode->i_ops = &rfs_i_ops;
return vinode;
}
//
// convert vfs inode to disk inode, and write it back to disk
//
int rfs_write_back_vinode(struct vinode *vinode) {
// copy vinode info to disk inode
struct rfs_dinode dinode;
dinode.size = vinode->size;
dinode.nlinks = vinode->nlinks;
dinode.blocks = vinode->blocks;
dinode.type = vinode->type;
for (int i = 0; i < RFS_DIRECT_BLKNUM; ++i) {
dinode.addrs[i] = vinode->addrs[i];
}
struct rfs_device *rdev = rfs_device_list[vinode->sb->s_dev->dev_id];
if (rfs_write_dinode(rdev, &dinode, vinode->inum) != 0) {
sprint("rfs_free_write_back_inode: failed to write back disk inode!\n");
return -1;
}
return 0;
}
//
// update vinode info by reading disk inode
//
int rfs_update_vinode(struct vinode *vinode) {
struct rfs_device *rdev = rfs_device_list[vinode->sb->s_dev->dev_id];
struct rfs_dinode *dinode = rfs_read_dinode(rdev, vinode->inum);
if (dinode == NULL) {
sprint("rfs_update_vinode: failed to read disk inode!\n");
return -1;
}
vinode->size = dinode->size;
vinode->nlinks = dinode->nlinks;
vinode->blocks = dinode->blocks;
vinode->type = dinode->type;
for (int i = 0; i < RFS_DIRECT_BLKNUM; ++i) {
vinode->addrs[i] = dinode->addrs[i];
}
free_page(dinode);
return 0;
}
/**** vfs-rfs file interface functions ****/
//
// read the content (for "len") of a file ("f_inode"), and copy the content
// to "r_buf".
//
ssize_t rfs_read(struct vinode *f_inode, char *r_buf, ssize_t len,
int *offset) {
// obtain disk inode from vfs inode
if (f_inode->size < *offset)
panic("rfs_read:offset should less than file size!");
if (f_inode->size < (*offset + len)) len = f_inode->size - *offset;
char buffer[len + 1];
// compute how many blocks we need to read
int align = *offset % RFS_BLKSIZE;
int block_offset = *offset / RFS_BLKSIZE;
int buf_offset = 0;
int readtimes = (align + len) / RFS_BLKSIZE;
int remain = (align + len) % RFS_BLKSIZE;
struct rfs_device *rdev = rfs_device_list[f_inode->sb->s_dev->dev_id];
// read first block
rfs_r1block(rdev, f_inode->addrs[block_offset]);
int first_block_len = (readtimes == 0 ? len : RFS_BLKSIZE - align);
memcpy(buffer + buf_offset, rdev->iobuffer + align, first_block_len);
buf_offset += first_block_len;
block_offset++;
readtimes--;
// readtimes < 0 means that the file has only one block (and not full),
// so our work is done
// otherwise...
if (readtimes >= 0) {
// read in complete blocks
while (readtimes != 0) {
rfs_r1block(rdev, f_inode->addrs[block_offset]);
memcpy(buffer + buf_offset, rdev->iobuffer, RFS_BLKSIZE);
buf_offset += RFS_BLKSIZE;
block_offset++;
readtimes--;
}
// read in the remaining data
if (remain > 0) {
rfs_r1block(rdev, f_inode->addrs[block_offset]);
memcpy(buffer + buf_offset, rdev->iobuffer, remain);
}
}
buffer[len] = '\0';
strcpy(r_buf, buffer);
*offset += len;
return len;
}
//
// write the content of "w_buf" (lengthed "len") to a file ("f_inode").
//
ssize_t rfs_write(struct vinode *f_inode, const char *w_buf, ssize_t len,
int *offset) {
if (f_inode->size < *offset) {
panic("rfs_write:offset should less than file size!");
}
// compute how many blocks we need to write
int align = *offset % RFS_BLKSIZE;
int writetimes = (len + align) / RFS_BLKSIZE;
int remain = (len + align) % RFS_BLKSIZE;
int buf_offset = 0;
int block_offset = *offset / RFS_BLKSIZE;
struct rfs_device *rdev = rfs_device_list[f_inode->sb->s_dev->dev_id];
// write first block
if (align != 0) {
rfs_r1block(rdev, f_inode->addrs[block_offset]);
int first_block_len = (writetimes == 0 ? len : RFS_BLKSIZE - align);
memcpy(rdev->iobuffer + align, w_buf, first_block_len);
rfs_w1block(rdev, f_inode->addrs[block_offset]);
buf_offset += first_block_len;
block_offset++;
writetimes--;
}
// writetimes < 0 means that the file has only one block (and not full),
// so our work is done
// otherwise...
if (writetimes >= 0) {
// write complete blocks
while (writetimes != 0) {
if (block_offset == f_inode->blocks) { // need to create new block
// allocate a free block for the file
f_inode->addrs[block_offset] = rfs_alloc_block(f_inode->sb);
f_inode->blocks++;
}
memcpy(rdev->iobuffer, w_buf + buf_offset, RFS_BLKSIZE);
rfs_w1block(rdev, f_inode->addrs[block_offset]);
buf_offset += RFS_BLKSIZE;
block_offset++;
writetimes--;
}
// write the remaining data
if (remain > 0) {
if (block_offset == f_inode->blocks) {
f_inode->addrs[block_offset] = rfs_alloc_block(f_inode->sb);
++f_inode->blocks;
}
memcpy(rdev->iobuffer, w_buf + buf_offset, remain);
rfs_w1block(rdev, f_inode->addrs[block_offset]);
}
}
// update file size
f_inode->size =
(f_inode->size < *offset + len ? *offset + len : f_inode->size);
*offset += len;
return len;
}
//
// lookup a directory entry("sub_dentry") under "parent".
// note that this is a one level lookup ,and the vfs layer will call this
// function several times until the final file is found.
// return: if found, return its vinode, otherwise return NULL
//
struct vinode *rfs_lookup(struct vinode *parent, struct dentry *sub_dentry) {
struct rfs_direntry *p_direntry = NULL;
struct vinode *child_vinode = NULL;
int total_direntrys = parent->size / sizeof(struct rfs_direntry);
int one_block_direntrys = RFS_BLKSIZE / sizeof(struct rfs_direntry);
struct rfs_device *rdev = rfs_device_list[parent->sb->s_dev->dev_id];
// browse the dir entries contained in a directory file
for (int i = 0; i < total_direntrys; ++i) {
if (i % one_block_direntrys == 0) { // read in the disk block at boundary
rfs_r1block(rdev, parent->addrs[i / one_block_direntrys]);
p_direntry = (struct rfs_direntry *)rdev->iobuffer;
}
if (strcmp(p_direntry->name, sub_dentry->name) == 0) { // found
child_vinode = rfs_alloc_vinode(parent->sb);
child_vinode->inum = p_direntry->inum;
if (rfs_update_vinode(child_vinode) != 0)
panic("rfs_lookup: read inode failed!");
break;
}
++p_direntry;
}
return child_vinode;
}
//
// create a file with "sub_dentry->name" at directory "parent" in rfs.
// return the vfs inode of the file being created.
//
struct vinode *rfs_create(struct vinode *parent, struct dentry *sub_dentry) {
struct rfs_device *rdev = rfs_device_list[parent->sb->s_dev->dev_id];
// ** find a free disk inode to store the file that is going to be created
struct rfs_dinode *free_dinode = NULL;
int free_inum = 0;
for (int i = 0; i < (RFS_BLKSIZE / RFS_INODESIZE * RFS_MAX_INODE_BLKNUM);
++i) {
free_dinode = rfs_read_dinode(rdev, i);
if (free_dinode->type == R_FREE) { // found
free_inum = i;
break;
}
free_page(free_dinode);
}
if (free_dinode == NULL)
panic("rfs_create: no more free disk inode, we cannot create file.\n" );
// initialize the states of the file being created
// TODO (lab4_1): implement the code for populating the disk inode (free_dinode)
// of a new file being created.
// hint: members of free_dinode to be filled are:
// size, should be zero for a new file.
// type, see kernel/rfs.h and find the type for a rfs file.
// nlinks, i.e., the number of links.
// blocks, i.e., its block count.
// Note: DO NOT DELETE CODE BELOW PANIC.
panic("You need to implement the code of populating a disk inode in lab4_1.\n" );
// DO NOT REMOVE ANY CODE BELOW.
// allocate a free block for the file
free_dinode->addrs[0] = rfs_alloc_block(parent->sb);
// ** write the disk inode of file being created to disk
rfs_write_dinode(rdev, free_dinode, free_inum);
free_page(free_dinode);
// ** build vfs inode according to dinode
struct vinode *new_vinode = rfs_alloc_vinode(parent->sb);
new_vinode->inum = free_inum;
rfs_update_vinode(new_vinode);
// ** append the new file as a direntry to its parent dir
int result = rfs_add_direntry(parent, sub_dentry->name, free_inum);
if (result == -1) {
sprint("rfs_create: rfs_add_direntry failed");
return NULL;
}
return new_vinode;
}
//
// there are two types of seek (specify by whence): LSEEK_SET, SEEK_CUR
// LSEEK_SET: set the file pointer to the offset
// LSEEK_CUR: set the file pointer to the current offset plus the offset
// return 0 if success, otherwise return -1
//
int rfs_lseek(struct vinode *f_inode, ssize_t new_offset, int whence, int *offset) {
int file_size = f_inode->size;
switch (whence) {
case LSEEK_SET:
if (new_offset < 0 || new_offset > file_size) {
sprint("rfs_lseek: invalid offset!\n");
return -1;
}
*offset = new_offset;
break;
case LSEEK_CUR:
if (*offset + new_offset < 0 || *offset + new_offset > file_size) {
sprint("rfs_lseek: invalid offset!\n");
return -1;
}
*offset += new_offset;
break;
default:
sprint("rfs_lseek: invalid whence!\n");
return -1;
}
return 0;
}
//
// read disk inode information from disk
//
int rfs_disk_stat(struct vinode *vinode, struct istat *istat) {
struct rfs_device *rdev = rfs_device_list[vinode->sb->s_dev->dev_id];
struct rfs_dinode *dinode = rfs_read_dinode(rdev, vinode->inum);
if (dinode == NULL) {
sprint("rfs_disk_stat: read dinode failed!\n");
return -1;
}
istat->st_inum = 1;
istat->st_inum = vinode->inum; // get inode number from vinode
istat->st_size = dinode->size;
istat->st_type = dinode->type;
istat->st_nlinks = dinode->nlinks;
istat->st_blocks = dinode->blocks;
free_page(dinode);
return 0;
}
//
// when a directory is opened, the contents of the directory file are read
// into the memory for directory read operations
//
int rfs_hook_opendir(struct vinode *dir_vinode, struct dentry *dentry) {
// allocate space and read the contents of the dir block into memory
void *pdire = NULL;
void *previous = NULL;
struct rfs_device *rdev = rfs_device_list[dir_vinode->sb->s_dev->dev_id];
// read-in the directory file, store all direntries in dir cache.
for (int i = dir_vinode->blocks - 1; i >= 0; i--) {
previous = pdire;
pdire = alloc_page();
if (previous != NULL && previous - pdire != RFS_BLKSIZE)
panic("rfs_hook_opendir: memory discontinuity");
rfs_r1block(rdev, dir_vinode->addrs[i]);
memcpy(pdire, rdev->iobuffer, RFS_BLKSIZE);
}
// save the pointer to the directory block in the vinode
struct rfs_dir_cache *dir_cache = (struct rfs_dir_cache *)alloc_page();
dir_cache->block_count = dir_vinode->blocks;
dir_cache->dir_base_addr = (struct rfs_direntry *)pdire;
dir_vinode->i_fs_info = dir_cache;
return 0;
}
//
// when a directory is closed, the memory space allocated for the directory
// block is freed
//
int rfs_hook_closedir(struct vinode *dir_vinode, struct dentry *dentry) {
struct rfs_dir_cache *dir_cache =
(struct rfs_dir_cache *)dir_vinode->i_fs_info;
// reclaim the dir cache
for (int i = 0; i < dir_cache->block_count; ++i) {
free_page((char *)dir_cache->dir_base_addr + i * RFS_BLKSIZE);
}
return 0;
}
//
// read a directory entry from the directory "dir", and the "offset" indicate
// the position of the entry to be read. if offset is 0, the first entry is read,
// if offset is 1, the second entry is read, and so on.
// return: 0 on success, -1 when there are no more entry (end of the list).
//
int rfs_readdir(struct vinode *dir_vinode, struct dir *dir, int *offset) {
int total_direntrys = dir_vinode->size / sizeof(struct rfs_direntry);
int one_block_direntrys = RFS_BLKSIZE / sizeof(struct rfs_direntry);
int direntry_index = *offset;
if (direntry_index >= total_direntrys) {
// no more direntry
return -1;
}
// reads a directory entry from the directory cache stored in vfs inode.
struct rfs_dir_cache *dir_cache =
(struct rfs_dir_cache *)dir_vinode->i_fs_info;
struct rfs_direntry *p_direntry = dir_cache->dir_base_addr + direntry_index;
// TODO (lab4_2): implement the code to read a directory entry.
// hint: in the above code, we had found the directory entry that located at the
// *offset, and used p_direntry to point it.
// in the remaining processing, we need to return our discovery.
// the method of returning is to popular proper members of "dir", more specifically,
// dir->name and dir->inum.
// note: DO NOT DELETE CODE BELOW PANIC.
panic("You need to implement the code for reading a directory entry of rfs in lab4_2.\n" );
// DO NOT DELETE CODE BELOW.
(*offset)++;
return 0;
}
//
// make a new direntry named "sub_dentry->name" under the directory "parent",
// return the vfs inode of subdir being created.
//
struct vinode *rfs_mkdir(struct vinode *parent, struct dentry *sub_dentry) {
struct rfs_device *rdev = rfs_device_list[parent->sb->s_dev->dev_id];
// ** find a free disk inode to store the file that is going to be created
struct rfs_dinode *free_dinode = NULL;
int free_inum = 0;
for (int i = 0; i < (RFS_BLKSIZE / RFS_INODESIZE * RFS_MAX_INODE_BLKNUM); i++) {
free_dinode = rfs_read_dinode(rdev, i);
if (free_dinode->type == R_FREE) { // found
free_inum = i;
break;
}
free_page(free_dinode);
}
if (free_dinode == NULL)
panic( "rfs_mkdir: no more free disk inode, we cannot create directory.\n" );
// initialize the states of the file being created
free_dinode->size = 0;
free_dinode->type = R_DIR;
free_dinode->nlinks = 1;
free_dinode->blocks = 1;
// allocate a free block for the file
free_dinode->addrs[0] = rfs_alloc_block(parent->sb);
// ** write the disk inode of file being created to disk
rfs_write_dinode(rdev, free_dinode, free_inum);
free_page(free_dinode);
// ** add a direntry to the directory
int result = rfs_add_direntry(parent, sub_dentry->name, free_inum);
if (result == -1) {
sprint("rfs_mkdir: rfs_add_direntry failed");
return NULL;
}
// ** allocate a new vinode
struct vinode *sub_vinode = rfs_alloc_vinode(parent->sb);
sub_vinode->inum = free_inum;
rfs_update_vinode(sub_vinode);
return sub_vinode;
}
/**** vfs-rfs file system type interface functions ****/
struct super_block *rfs_get_superblock(struct device *dev) {
struct rfs_device *rdev = rfs_device_list[dev->dev_id];
// read super block from ramdisk
if (rfs_r1block(rdev, RFS_BLK_OFFSET_SUPER) != 0)
panic("RFS: failed to read superblock!\n");
struct rfs_superblock d_sb;
memcpy(&d_sb, rdev->iobuffer, sizeof(struct rfs_superblock));
// set the data for the vfs super block
struct super_block *sb = alloc_page();
sb->magic = d_sb.magic;
sb->size = d_sb.size;
sb->nblocks = d_sb.nblocks;
sb->ninodes = d_sb.ninodes;
sb->s_dev = dev;
if( sb->magic != RFS_MAGIC )
panic("rfs_get_superblock: wrong ramdisk device!\n");
// build root dentry and root inode
struct vinode *root_inode = rfs_alloc_vinode(sb);
root_inode->inum = 0;
rfs_update_vinode(root_inode);
struct dentry *root_dentry = alloc_vfs_dentry("/", root_inode, NULL);
sb->s_root = root_dentry;
// save the bitmap in the s_fs_info field
if (rfs_r1block(rdev, RFS_BLK_OFFSET_BITMAP) != 0)
panic("RFS: failed to read bitmap!\n");
void *bitmap = alloc_page();
memcpy(bitmap, rdev->iobuffer, RFS_BLKSIZE);
sb->s_fs_info = bitmap;
return sb;
}