Minix块高速缓存

参照《操作系统:设计与实现》文中代码与文字很多出自此书。
Minix块高速缓存如下图所示:
 

 
相关数据结构如下:
此数据结构在/src/fs/buf.h中
struct buf {
  /* Data portion of the buffer. */
  union {/*公用体,可用来普通存储数据,目录,i节点,位图*/
    char b__data[BLOCK_SIZE];       /* ordinary user data */
    struct direct b__dir[NR_DIR_ENTRIES];    /* directory block */
    zone1_t b__v1_ind[V1_INDIRECTS];      /* V1 indirect block */
    zone_t  b__v2_ind[V2_INDIRECTS];      /* V2 indirect block */
    d1_inode b__v1_ino[V1_INODES_PER_BLOCK]; /* V1 inode block */
    d2_inode b__v2_ino[V2_INODES_PER_BLOCK]; /* V2 inode block */
    bitchunk_t b__bitmap[BITMAP_CHUNKS];     /* bit map block */
  } b;
  /* Header portion of the buffer. */
  struct buf *b_next;  /* used to link all free bufs in a chain */
  struct buf *b_prev;  /* used to link all free bufs the other way */
  struct buf *b_hash;  /* used to link bufs on hash chains ,用于连接属于同一
hash的链
                       */
  block_t b_blocknr;  /* block number of its (minor) device */
  dev_t b_dev;   /* major | minor device where block resides */
  char b_dirt;   /* CLEAN or DIRTY */
  char b_count;   /* number of users of this buffer */
} buf[NR_BUFS];/*目前此数目为80*/
 
struct buf *buf_hash[NR_BUF_HASH]; /* the buffer hash table 哈希表,大小为128
个*/
struct buf *front; /* points to least recently used free block
指向最近不久不会使用的
                   块*/
struct buf *rear; /* points to most recently used free block
指向最近不久会使用的块*/
 
当文件系统需要新的块时,先去上述缓存区中查找,如果找到则将块的指针返回,如果未找到,则front中替换出一个块出去,从磁盘上加载一个块,放入此缓冲区中。当文件系统用完该块,系统先将该块挂入front和rear指针指向的空闲缓冲区中。目前普通数据块的大小为1024
字节。
文件系统需要一块时会调用get_block(),文件系统不需要该块时调用put_block()
将块放入空闲缓冲区,调用rm_lru()函数将块从空闲缓冲区中删除。
具体代码如下: 代码位于/src/fs/cache.c
PUBLIC struct buf *get_block(dev, block, only_search)
register dev_t dev;  /* on which device is the block? */
register block_t block;  /* which block is wanted? */
int only_search;  /* if NO_READ, don't read, else act normal */
{
/* Check to see if the requested block is in the block cache.  If so, return
 * a pointer to it.  If not, evict some other block and fetch it (unless
 * 'only_search' is 1).  All the blocks in the cache that are not in use
 * are linked together in a chain, with 'front' pointing to the least recently
 * used block and 'rear' to the most recently used block.  If 'only_search' is
 * 1, the block being requested will be overwritten in its entirety, so it is
 * only necessary to see if it is in the cache; if it is not, any free buffer
 * will do.  It is not necessary to actually read the block in from disk.
 * If 'only_search' is PREFETCH, the block need not be read from the disk,
 * and the device is not to be marked on the block, so callers can tell if
 * the block returned is valid.
 * In addition to the LRU chain, there is also a hash chain to link together
 * blocks whose block numbers end with the same bit strings, for fast lookup.
 */
  int b;
  register struct buf *bp, *prev_ptr;
  /* Search the hash chain for (dev, block). Do_read() can use
   * get_block(NO_DEV ...) to get an unnamed block to fill with zeros when
   * someone wants to read from a hole in a file, in which case this search
   * is skipped
   */
  if (dev != NO_DEV) {
 b = (int) block & HASH_MASK; /*此处HASH_MASK的值为127,即1111111*/
 bp = buf_hash[b]; /*取到其hash表中的表项入口*/
 while (bp != NIL_BUF) {
  if (bp->b_blocknr == block && bp->b_dev == dev) {
   /* Block needed has been found. */
   if (bp->b_count == 0) rm_lru(bp);
/*如果此表项在空闲缓冲链表中,则将其从其中删除*/
   bp->b_count++; /* record that block is in use 使用计数器加1*/
   return(bp);
  } else {
   /* This block is not the one sought. ,遍历同属一hash
值的链表,查找需要的块*/
   bp = bp->b_hash; /* move to next block on hash chain */
  }
 }
  }
  /* Desired block is not on available chain.  Take oldest block ('front'). */
  if ((bp = front) == NIL_BUF) panic("all buffers in use", NR_BUFS);
  rm_lru(bp);
/*未找到需要的块,如果空闲链表不为空,则从其中摘除一项,用于接下来加载新的块*/
  /* Remove the block that was just taken from its hash chain. */
  b = (int) bp->b_blocknr & HASH_MASK;
  prev_ptr = buf_hash[b];
  if (prev_ptr == bp) {/*将摘除的空闲块从其目前所属的hash表中摘除*/
 buf_hash[b] = bp->b_hash;
  } else {
 /* The block just taken is not on the front of its hash chain. */
 while (prev_ptr->b_hash != NIL_BUF)
  if (prev_ptr->b_hash == bp) {
   prev_ptr->b_hash = bp->b_hash; /* found it,查找摘除的空闲块的具体位置 */
   break;
  } else {
   prev_ptr = prev_ptr->b_hash; /* keep looking */
  }
  }
  /* If the block taken is dirty, make it clean by writing it to the disk.
   * Avoid hysteresis by flushing all other dirty blocks for the same device.
   */
  if (bp->b_dev != NO_DEV) {
 if (bp->b_dirt == DIRTY) flushall(bp->b_dev);/*如果空闲块为脏,则回写到磁盘*/
#if ENABLE_CACHE2
 put_block2(bp);
#endif
  }
  /* Fill in block's parameters and add it to the hash chain where it goes. */
  bp->b_dev = dev;  /* fill in device number */
  bp->b_blocknr = block; /* fill in block number */
  bp->b_count++;  /* record that block is being used */
  b = (int) bp->b_blocknr & HASH_MASK;
  bp->b_hash = buf_hash[b];/*将原hash链中的第一块链接到当前块后面*/
  buf_hash[b] = bp;  /* add to hash list 加入到hash表中*/
  /* Go get the requested block unless searching or prefetching. */
  if (dev != NO_DEV) {
#if ENABLE_CACHE2
 if (get_block2(bp, only_search)) /* in 2nd level cache */;
 else
#endif
 if (only_search == PREFETCH) bp->b_dev = NO_DEV;
 else
 if (only_search == NORMAL) rw_block(bp, READING);
  }
  return(bp);   /* return the newly acquired block */
}

/*===========================================================================*
 *    put_block         *
 *===========================================================================*/
PUBLIC void put_block(bp, block_type)/*将块加入到LRU链中*/
register struct buf *bp; /* pointer to the buffer to be released */
int block_type;   /* INODE_BLOCK, DIRECTORY_BLOCK, or whatever */
{
/* Return a block to the list of available blocks.   Depending on 'block_type'
 * it may be put on the front or rear of the LRU chain.  Blocks that are
 * expected to be needed again shortly (e.g., partially full data blocks)
 * go on the rear; blocks that are unlikely to be needed again shortly
 * (e.g., full data blocks) go on the front.  Blocks whose loss can hurt
 * the integrity of the file system (e.g., inode blocks) are written to
 * disk immediately if they are dirty.
 */
  if (bp == NIL_BUF) return; /* it is easier to check here than in caller */
  bp->b_count--;  /* there is one use fewer now */
  if (bp->b_count != 0) return; /* block is still in use */
  bufs_in_use--;  /* one fewer block buffers in use */
  /* Put this block back on the LRU chain.  If the ONE_SHOT bit is set in
   * 'block_type', the block is not likely to be needed again shortly, so put
   * it on the front of the LRU chain where it will be the first one to be
   * taken when a free buffer is needed later.
   */
  if (block_type & ONE_SHOT) {
 /* Block probably won't be needed quickly. Put it on front of chain.
    * It will be the next block to be evicted from the cache.
    */
 bp->b_prev = NIL_BUF;
 bp->b_next = front;
 if (front == NIL_BUF)
  rear = bp; /* LRU chain was empty */
 else
  front->b_prev = bp;
 front = bp;
  } else {
 /* Block probably will be needed quickly.  Put it on rear of chain.
    * It will not be evicted from the cache for a long time.
    */
 bp->b_prev = rear;
 bp->b_next = NIL_BUF;
 if (rear == NIL_BUF)
  front = bp;
 else
  rear->b_next = bp;
 rear = bp;
  }
  /* Some blocks are so important (e.g., inodes, indirect blocks) that they
   * should be written to the disk immediately to avoid messing up the file
   * system in the event of a crash.
   */
  if ((block_type & WRITE_IMMED) && bp->b_dirt==DIRTY && bp->b_dev != NO_DEV)
 rw_block(bp, WRITING);
}
 
PRIVATE void rm_lru(bp)/*将块从LRU链中删除,简单的链表操作*/
struct buf *bp;
{
/* Remove a block from its LRU chain. */
  struct buf *next_ptr, *prev_ptr;
  bufs_in_use++;
  next_ptr = bp->b_next; /* successor on LRU chain */
  prev_ptr = bp->b_prev; /* predecessor on LRU chain */
  if (prev_ptr != NIL_BUF)
 prev_ptr->b_next = next_ptr;
  else
 front = next_ptr; /* this block was at front of chain */
  if (next_ptr != NIL_BUF)
 next_ptr->b_prev = prev_ptr;
  else
 rear = prev_ptr; /* this block was at rear of chain */
}

 

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值