File system buffer

buffer cache, page cache, page buffer.

这篇文章介绍了buffer早期的演进。

http://bulma.net/impresion.phtml?nIdNoticia=1154

缺失的图可以在这里找

http://blog.sina.com.cn/s/blog_43653df6010003t6.html

 

buffer cache接口物理设备，被VM和IO子系统共享。

早期，page cache建立在buffer cache之上，用于VM子系统，它和buffer cache是分离的。因此OS需要处理两者的同步。

然后，linux把文件数据只保存在page cache中（或者是page cache和buffer cache共享同样的内存）来避免double copy。

然后linus决定统一使用page cache作为缓存，他提到

it will be one less thing that uses the buffer cache as a ''cache'' (I want people to think of, and use, the buffer cache as an _IO_ entity, not a cache).

 

released 2.4.10, which finally unified the page and buffer cache。

buffer cache包含在page buffer中，page buffer属于page cache。

什么时候使用buffer cache，什么时候使用page cache呢？

可以这样理解，buffer cache要求的是物理上的连续（一个磁盘块），page cache要求的是逻辑上的连续（文件）。

在2.4 ext2中，对小文件使用4k的page buffer，这也可能是ext文件系统在小文件上低效的原因之一（猜测），不知道ReiserFS是否有不同的缓存策略。

 

Kernel function	System call	Cache	I/O operation
bread( )	None	Buffer	Read an Ext2 superblock
bread( )	None	Buffer	Read an Ext2 inode
generic_file_read( )	getdents( )	Page	Read an Ext2 directory
generic_file_read( )	read( )	Page	Read an Ext2 regular file
generic_file_write( )	write( )	Page	Write an Ext2 regular file
generic_file_read( )	read( )	Page	Read a block device file
generic_file_write( )	write( )	Page	Write a block device file
filemap_nopage( )	None	Page	Access a memory-mapped file
brw_page( )	None	Page	Access to swapped-out page

 另外有看到（最开始的链接），buffer cache只能映射到内核空间，因此page buffer用途更广。

 

Starting from stable version 2.4.10, the buffer cache does not really exist anymore. In fact, for reasons of efficiency, block buffers are no longer allocated individually; instead, they are stored in dedicated pages called "buffer pages ," which are kept in the page cache.

Formally, a buffer page is a page of data associated with additional descriptors called "buffer heads ," whose main purpose is to quickly locate the disk address of each individual block in the page.