Re: [PATCH 1/5] Forking ext4 filesystem from ext3 filesystem

Andrew Morton <[email protected]> Thu, 10 Aug 2006 13:27:20 -0700
Newsgroups gmane.comp.file-systems.ext2.devel,gmane.linux.file-systems,gmane.linux.kernel
Message-ID <[email protected]>
On Thu, 10 Aug 2006 16:13:33 -0400
Jeff Garzik <[email protected]> wrote:

> The sooner we kill buffer heads and use submit_bio(), the better :)

A buffer_head is a caching entity and a bio is an IO container.  They're
quite separate concepts.

A buffer_head is the kernel's sole abstraction of a disk block. 
Filesystems use disk blocks a lot, and they need such an abstraction.

If one was to replace buffer_heads with direct-to-BIO operations then the
filesytem would need to internally track the mapping from

	page+offset+length -> disk block

and it would need to internally track the page+offset+length<->disk block
coherency state and it would need to internally perform serialisation of
access to each page+offset+length hunk of pagecache and etc and etc and
etc.  Create a data structure with which to do all that and voila,
buffer_heads reinvented.

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642