On Wednesday 10 April 2002 09:52 am, Steve Best wrote:
> A very large part of extendfs functionality is in the file system,
> like 100%. Doing this is user space is going to have some challenges,
> like there several items that aren't exported by the file system
> (i.e. creating a transaction, .....) 

The transaction stuff is meaningless in user-space, and the I/O is 
handled completely differently as well.  Rather than trying to port the 
kernel code to user space, it would be more logical to use the kernel 
code as a roadmap to rewrite the same function in user space.  I'm sure 
there would be some useful code that can be found in the jfsutils 
source to do some of the lower-level functions, i.e. reading the inode 
allocation map pages via the imap inode's xtree.

Here is a version of
> jfs_extendfs.c that I'm in the process of porting. It still doesn't
> compile yet, hopefully it will soon.
>
> http://oss.software.ibm.com/developer/opensource/jfs/project/pub/jfs_
>extendfswip.c
>
> Does the extendfs functionality need to be in user space? or can this
> functionality be in the file system and just be called from your user
> space program?

This is a good question.  Once the functionality is in the kernel, is 
there a need for it to be implemented in user space?  Note that at 
least for now, the kernel function is only going to allow the volume to 
grow, not to shrink.
-- 
David Kleikamp
IBM Linux Technology Center
_______________________________________________
Jfs-discussion mailing list
[EMAIL PROTECTED]
http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion

Reply via email to