romfs文件系统注释

/*
 * ROMFS file system, Linux implementation
 *
 * Copyright (C) 1997-1999  Janos Farkas <chexum@shadow.banki.hu>
 *
 * Using parts of the minix filesystem
 * Copyright (C) 1991, 1992  Linus Torvalds
 *
 * and parts of the affs filesystem additionally
 * Copyright (C) 1993  Ray Burr
 * Copyright (C) 1996  Hans-Joachim Widmaier
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 *
 * Changes
 *     Changed for 2.1.19 modules
 * Jan 1997   Initial release
 * Jun 1997   2.1.43+ changes
 *     Proper page locking in readpage
 *     Changed to work with 2.1.45+ fs
 * Jul 1997   Fixed follow_link
 *   2.1.47
 *     lookup shouldn't return -ENOENT
 *     from Horst von Brand:
 *       fail on wrong checksum
 *       double unlock_super was possible
 *       correct namelen for statfs
 *     spotted by Bill Hawes:
 *       readlink shouldn't iput()
 * Jun 1998 2.1.106  from Avery Pennarun: glibc scandir()
 *       exposed a problem in readdir
 *   2.1.107  code-freeze spellchecker run
 * Aug 1998   2.1.118+ VFS changes
 * Sep 1998 2.1.122  another VFS change (follow_link)
 * Apr 1999 2.2.7  no more EBADF checking in
 *       lookup/readdir, use ERR_PTR
 * Jun 1999 2.3.6  d_alloc_root use changed
 *   2.3.9  clean up usage of ENOENT/negative
 *       dentries in lookup
 *     clean up page flags setting
 *       (error, uptodate, locking) in
 *       in readpage
 *     use init_special_inode for
 *       fifos/sockets (and streamline) in
 *       read_inode, fix _ops table order
 * Aug 1999 2.3.16  __initfunc() => __init change
 * Oct 1999 2.3.24  page->owner hack obsoleted
 * Nov 1999 2.3.27  2.3.25+ page->offset => index change
 */

/* todo:
 * - see Documentation/filesystems/romfs.txt
 * - use allocated, not stack memory for file names?
 * - considering write access...
 * - network (tftp) files?
 * - merge back some _op tables
 */

/*
 * Sorry about some optimizations and for some goto's.  I just wanted
 * to squeeze some more bytes out of this code.. :)
 */

#include <linux/module.h>
#include <linux/types.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/romfs_fs.h>
#include <linux/fs.h>
#include <linux/locks.h>
#include <linux/init.h>
#include <linux/smp_lock.h>

#include <asm/uaccess.h>

//校验数据
static __s32
romfs_checksum(void *data, int size)
{
 __s32 sum, *ptr;

 sum = 0; ptr = data;
 size>>=2;
 while (size>0) {
  sum += ntohl(*ptr++);
  size--;
 }
 return sum;
}

static struct super_operations romfs_ops;

static struct super_block *
romfs_read_super(struct super_block *s, void *data, int silent)
{
 struct buffer_head *bh;
 kdev_t dev = s->s_dev;
 struct romfs_super_block *rsb;
 int sz;

 /* I would parse the options here, but there are none.. :) */

 set_blocksize(dev, ROMBSIZE);
 s->s_blocksize = ROMBSIZE;
 s->s_blocksize_bits = ROMBSBITS;
 s->u.generic_sbp = (void *) 0;
 s->s_maxbytes = 0xFFFFFFFF;

 bh = sb_bread(s, 0);
 if (!bh) {
  /* XXX merge with other printk? */
                printk ("romfs: unable to read superblock/n");
  goto outnobh;
 }

 rsb = (struct romfs_super_block *)bh->b_data;
 sz = ntohl(rsb->size);
 if (rsb->word0 != ROMSB_WORD0 || rsb->word1 != ROMSB_WORD1
    || sz < ROMFH_SIZE) {
  if (!silent)
   printk ("VFS: Can't find a romfs filesystem on dev "
    "%s./n", kdevname(dev));
  goto out;
 }
 if (romfs_checksum(rsb, min_t(int, sz, 512))) {
  printk ("romfs: bad initial checksum on dev "
   "%s./n", kdevname(dev));
  goto out;
 }

 s->s_magic = ROMFS_MAGIC;
 //文件系统的最大容量
 s->u.romfs_sb.s_maxsize = sz;

 s->s_flags |= MS_RDONLY;

 /* Find the start of the fs */
 sz = (ROMFH_SIZE +
       strnlen(rsb->name, ROMFS_MAXFN) + 1 + ROMFH_PAD)
      & ROMFH_MASK;

 brelse(bh);

 s->s_op = &romfs_ops;
 //iget 获取inode
 //root节点的no实际上是offset，不是0开始的
 //d_alloc_root  分配该inode对应的dentry.
 s->s_root = d_alloc_root(iget(s, sz));

 if (!s->s_root)
  goto outnobh;

 /* Ehrhm; sorry.. :)  And thanks to Hans-Joachim Widmaier  :) */
 if (0) {
out:
  brelse(bh);
outnobh:
  s = NULL;
 }

 return s;
}

/* That's simple too. */

static int
romfs_statfs(struct super_block *sb, struct statfs *buf)
{
 buf->f_type = ROMFS_MAGIC;
 buf->f_bsize = ROMBSIZE;
 buf->f_bfree = buf->f_bavail = buf->f_ffree;
 buf->f_blocks = (sb->u.romfs_sb.s_maxsize+ROMBSIZE-1)>>ROMBSBITS;
 buf->f_namelen = ROMFS_MAXFN;
 return 0;
}

/* some helper routines */

//获取文件名长度，不超过128字节
static int
romfs_strnlen(struct inode *i, unsigned long offset, unsigned long count)
{
 struct buffer_head *bh;
 unsigned long avail, maxsize, res;

 //文件系统越界检查
 maxsize = i->i_sb->u.romfs_sb.s_maxsize;
 if (offset >= maxsize)
  return -1;

 /* strnlen is almost always valid */
 if (count > maxsize || offset+count > maxsize)
  count = maxsize-offset;

 //由于采用offset寻址，没有block的说法
 //这里通过 offset>>ROMBSBITS (10bit)，强制将offset换算为
 //对应的block。 bh = sb_bread(i->i_sb, offset>>ROMBSBITS);
 if (!bh)
  return -1;  /* error */
 
 //计算第一个block中,offset位置距离block边界
 //剩余的字节数
 avail = ROMBSIZE - (offset & ROMBMASK);
 maxsize = min_t(unsigned long, count, avail);
 //为了得到文件名的长度
 res = strnlen(((char *)bh->b_data)+(offset&ROMBMASK), maxsize);
 brelse(bh);

 if (res < maxsize)
  return res;  /* found all of it */

 while (res < count) {
  offset += maxsize;

  bh = sb_bread(i->i_sb, offset>>ROMBSBITS);
  if (!bh)
   return -1;
  maxsize = min_t(unsigned long, count - res, ROMBSIZE);
  avail = strnlen(bh->b_data, maxsize);
  res += avail;
  brelse(bh);
  if (avail < maxsize)
   return res;
 }
 return res;
}

/*
从offset位置，拷贝count数据，返回到dest
count 的值不一定小于1k，可以一次读取多个block
*/
static int
romfs_copyfrom(struct inode *i, void *dest, unsigned long offset, unsigned long count)
{
 struct buffer_head *bh;
 unsigned long avail, maxsize, res;
 
       //文件系统越界检查
 maxsize = i->i_sb->u.romfs_sb.s_maxsize;
 if (offset >= maxsize || count > maxsize || offset+count>maxsize)
  return -1;

 //由于采用offset寻址，没有block的说法
 //这里通过 offset>>ROMBSBITS (10bit)，强制将offset换算为
 //对应的block。
 bh = sb_bread(i->i_sb, offset>>ROMBSBITS);
 if (!bh)
  return -1;  /* error */

 //计算第一个block中,offset位置距离block边界
 //剩余的字节数
 avail = ROMBSIZE - (offset & ROMBMASK);
 //求count,和avail的最小值
 maxsize = min_t(unsigned long, count, avail);
 //拷贝数据
 memcpy(dest, ((char *)bh->b_data) + (offset & ROMBMASK), maxsize);
 brelse(bh);
 //处理完第一个block

 //完成maxsize个数据拷贝
 res = maxsize;   /* all of it */

 //若count的数据无法从第一个block中全部获取
 //执行下面的循环
 while (res < count) {
  //偏移指针
  offset += maxsize;
  dest += maxsize;

  bh = sb_bread(i->i_sb, offset>>ROMBSBITS);
  if (!bh)
   return -1;
  //判断剩余的字节count-res是否超过一个block的大小
  maxsize = min_t(unsigned long, count - res, ROMBSIZE);
  memcpy(dest, bh->b_data, maxsize);
  brelse(bh);
  res += maxsize;
 }
 return res;
}

static unsigned char romfs_dtype_table[] = {
 DT_UNKNOWN, DT_DIR, DT_REG, DT_LNK, DT_BLK, DT_CHR, DT_SOCK, DT_FIFO
};

//文件系统对目录的读操作
static int
romfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
 struct inode *i = filp->f_dentry->d_inode;
 struct romfs_inode ri;
 unsigned long offset, maxoff;
 int j, ino, nextfh;
 int stored = 0;
 char fsname[ROMFS_MAXFN]; /* XXX dynamic? */

  //文件系统越界检查
 maxoff = i->i_sb->u.romfs_sb.s_maxsize;

 offset = filp->f_pos;
 if (!offset) {
  //获取目录节点偏移
  offset = i->i_ino & ROMFH_MASK;
  //读取目录节点信息
  if (romfs_copyfrom(i, &ri, offset, ROMFH_SIZE) <= 0)
   return stored;
  //找到目录对应的偏移位置
  offset = ntohl(ri.spec) & ROMFH_MASK;
 }

 /* Not really failsafe, but we are read-only... */
 for(;;) {
  if (!offset || offset >= maxoff) {
   offset = maxoff;
   filp->f_pos = offset;
   return stored;
  }
  filp->f_pos = offset;

  /* Fetch inode info */
  if (romfs_copyfrom(i, &ri, offset, ROMFH_SIZE) <= 0)
   //错误返回
   return stored;

  j = romfs_strnlen(i, offset+ROMFH_SIZE, sizeof(fsname)-1);
  if (j < 0)
   return stored;

  fsname[j]=0;
  //获取文件名
  romfs_copyfrom(i, fsname, offset+ROMFH_SIZE, j);

  ino = offset;
  nextfh = ntohl(ri.next);
  //处理硬链接
  if ((nextfh & ROMFH_TYPE) == ROMFH_HRD)
   ino = ntohl(ri.spec);
  //返回用户区，文件名信息
  if (filldir(dirent, fsname, j, offset, ino,
       romfs_dtype_table[nextfh & ROMFH_TYPE]) < 0) {
       //错误返回
   return stored;
  }
  //目录里面的元素个数
  stored++;
  offset = nextfh & ROMFH_MASK;
 }
}

//目录节点查找函数
//可以查找该目录下面的文件，目录等。。
static struct dentry *
romfs_lookup(struct inode *dir, struct dentry *dentry)
{
 unsigned long offset, maxoff;
 int fslen, res;
 struct inode *inode;
 char fsname[ROMFS_MAXFN]; /* XXX dynamic? */
 struct romfs_inode ri;
 const char *name;  /* got from dentry */
 int len;

 res = -EACCES;   /* placeholder for "no data here" */
 //目录节点 在文件系统的offset
 //可以通过ino换算
 offset = dir->i_ino & ROMFH_MASK;
 //从文件系统读取对应目录文件头数据16字节
 if (romfs_copyfrom(dir, &ri, offset, ROMFH_SIZE) <= 0)
  goto out;

 //maxoff表示文件系统的大小
 maxoff = dir->i_sb->u.romfs_sb.s_maxsize;

 //目录的偏移起始地址
 offset = ntohl(ri.spec) & ROMFH_MASK;

 /* OK, now find the file whose name is in "dentry" in the
  * directory specified by "dir".  */

 name = dentry->d_name.name;
 len = dentry->d_name.len;
 //根据dentry查找对应的inode

 for(;;) {
  //越界检查
  if (!offset || offset >= maxoff)
   goto out0;
 
  if (romfs_copyfrom(dir, &ri, offset, ROMFH_SIZE) <= 0)
   goto out;

  /* try to match the first 16 bytes of name */
  //获取文件名长度
  fslen = romfs_strnlen(dir, offset+ROMFH_SIZE, ROMFH_SIZE);
 
  if (len < ROMFH_SIZE) { //小于128字节
   if (len == fslen) {
    /* both are shorter, and same size */
    romfs_copyfrom(dir, fsname, offset+ROMFH_SIZE, len+1);
    //判断是否一致，正确则退出for
    if (strncmp (name, fsname, len) == 0)
     break;
   }
  } else if (fslen >= ROMFH_SIZE) {
   /* both are longer; XXX optimize max size */
   fslen = romfs_strnlen(dir, offset+ROMFH_SIZE, sizeof(fsname)-1);
   if (len == fslen) {
    romfs_copyfrom(dir, fsname, offset+ROMFH_SIZE, len+1);
    if (strncmp(name, fsname, len) == 0)
     break;
   }
  }
  //无法匹配则继续下一个
  /* next entry */
  offset = ntohl(ri.next) & ROMFH_MASK;
 }

 /* Hard link handling */
 if ((ntohl(ri.next) & ROMFH_TYPE) == ROMFH_HRD)
  offset = ntohl(ri.spec) & ROMFH_MASK;

 if ((inode = iget(dir->i_sb, offset)))
  goto outi;

 /*
  * it's a bit funky, _lookup needs to return an error code
  * (negative) or a NULL, both as a dentry.  ENOENT should not
  * be returned, instead we need to create a negative dentry by
  * d_add(dentry, NULL); and return 0 as no error.
  * (Although as I see, it only matters on writable file
  * systems).
  */

out0: inode = NULL;
outi: res = 0;
 d_add (dentry, inode);

out: return ERR_PTR(res);
}

/*
 * Ok, we do readpage, to be able to execute programs.  Unfortunately,
 * we can't use bmap, since we may have looser alignments.
 */

static int
romfs_readpage(struct file *file, struct page * page)
{
 struct inode *inode = page->mapping->host;
 unsigned long offset, avail, readlen;
 void *buf;
 int result = -EIO;

 page_cache_get(page);
 lock_kernel();
 buf = kmap(page);
 if (!buf)
  goto err_out;

 /* 32 bit warning -- but not for us :) */
 offset = page->index << PAGE_CACHE_SHIFT;

 //inode->i_size只有普通文件＆符号链接文件
 if (offset < inode->i_size) {
  //avail表明剩余的字节数
  avail = inode->i_size-offset;
  //读取一页还是avail
  readlen = min_t(unsigned long, avail, PAGE_SIZE);
 
  if (romfs_copyfrom(inode, buf, inode->u.romfs_i.i_dataoffset+offset, readlen) == readlen) {
   //页面尾部填充零
   if (readlen < PAGE_SIZE) {
    memset(buf + readlen,0,PAGE_SIZE-readlen);
   }
   SetPageUptodate(page);
   result = 0;
  }
 }
 //错误处理
 if (result) {
  memset(buf, 0, PAGE_SIZE);
  SetPageError(page);
 }
 flush_dcache_page(page);

 UnlockPage(page);

 kunmap(page);
err_out:
 page_cache_release(page);
 unlock_kernel();

 return result;
}

/* Mapping from our types to the kernel */

static struct address_space_operations romfs_aops = {
 readpage: romfs_readpage
};

static struct file_operations romfs_dir_operations = {
 read:  generic_read_dir,
 readdir: romfs_readdir,
};

static struct inode_operations romfs_dir_inode_operations = {
 lookup:  romfs_lookup,
};

//权限映射表
static mode_t romfs_modemap[] =
{
 0, S_IFDIR+0644, S_IFREG+0644, S_IFLNK+0777,
 S_IFBLK+0600, S_IFCHR+0600, S_IFSOCK+0644, S_IFIFO+0644
};

static void
romfs_read_inode(struct inode *i)
{
 int nextfh, ino;
 struct romfs_inode ri;

 ino = i->i_ino & ROMFH_MASK;
 i->i_mode = 0;

 /* Loop for finding the real hard link */
 for(;;) {
  //根据ino作为offset获取16字节的文件头
  if (romfs_copyfrom(i, &ri, ino, ROMFH_SIZE) <= 0) {
   printk("romfs: read error for inode 0x%x/n", ino);
   return;
  }
  /* XXX: do romfs_checksum here too (with name) */

  nextfh = ntohl(ri.next);
  //当前文件节点是否是硬链接
  if ((nextfh & ROMFH_TYPE) != ROMFH_HRD)
   break;
  //如果是硬链接，spec.info保存对应的地址
  //由于ino和offset一致，所以重新读取节点
  ino = ntohl(ri.spec) & ROMFH_MASK;
 }

 //到此只是得到了文件头信息
 i->i_nlink = 1;  /* Hard to decide.. */
 i->i_size = ntohl(ri.size);
 i->i_mtime = i->i_atime = i->i_ctime = 0;
 i->i_uid = i->i_gid = 0;

        /* Precalculate the data offset */
 //获取文件名的长度
        ino = romfs_strnlen(i, ino+ROMFH_SIZE, ROMFS_MAXFN);
  //ino非负，表明没有错误
        if (ino >= 0)
   //计算从offset位置到数据区的偏移ino=16+name+pad
   //仍然16字节对其
                ino = ((ROMFH_SIZE+ino+1+ROMFH_PAD)&ROMFH_MASK);
        else
                ino = 0;

 //metasize就是该文件除数据区外的信息长度
        i->u.romfs_i.i_metasize = ino;
 //dataoffset就是offset+偏移ino,表明数据区的位置
        i->u.romfs_i.i_dataoffset = ino+(i->i_ino&ROMFH_MASK);

        /* Compute permissions */
 //判断节点类型，文件?目录 ?....
 //赋予相应的权限
        ino = romfs_modemap[nextfh & ROMFH_TYPE];
 /* only "normal" files have ops */
 switch (nextfh & ROMFH_TYPE) {
  case 1: //目录
   i->i_size = i->u.romfs_i.i_metasize;
   //目录节点的操作函数
   i->i_op = &romfs_dir_inode_operations;
  
   i->i_fop = &romfs_dir_operations;
   //是否有执行权限
   if (nextfh & ROMFH_EXEC)
    ino |= S_IXUGO;
   i->i_mode = ino;
   break;
  case 2: //普通文件
 
   i->i_fop = &generic_ro_fops;
  
   i->i_data.a_ops = &romfs_aops;
   //是否有执行权限
   if (nextfh & ROMFH_EXEC)
    ino |= S_IXUGO;
   i->i_mode = ino;
   break;
  case 3://符号链接
 
   i->i_op = &page_symlink_inode_operations;
  
   i->i_data.a_ops = &romfs_aops;
   i->i_mode = ino | S_IRWXUGO;
   break;
  default: //其他由内核处理
   /* depending on MBZ for sock/fifos */
   nextfh = ntohl(ri.spec);
   nextfh = kdev_t_to_nr(MKDEV(nextfh>>16,nextfh&0xffff));
   init_special_inode(i, ino, nextfh);
 }
}

static struct super_operations romfs_ops = {
 read_inode: romfs_read_inode,
 statfs:  romfs_statfs,
};

static DECLARE_FSTYPE_DEV(romfs_fs_type, "romfs", romfs_read_super);

static int __init init_romfs_fs(void)
{
 return register_filesystem(&romfs_fs_type);
}

static void __exit exit_romfs_fs(void)
{
 unregister_filesystem(&romfs_fs_type);
}

/* Yes, works even as a module... :) */

EXPORT_NO_SYMBOLS;

module_init(init_romfs_fs)
module_exit(exit_romfs_fs)
MODULE_LICENSE("GPL");