Re: 2.6.0-test6-mm2

[viro@parcelfarce.linux.theplanet.co.uk]

On Thu, Oct 02, 2003 at 09:04:32PM +0400, Nikita Danilov wrote:
> Andrew Morton writes:
>  > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test6/2.6.0-test6-mm2/
>  > 
>  > . A large series of VFS patches from Al Viro which replace usage of
>  >   file->f_dentry->d_inode->i_mapping with the new file->f_mapping.
>  > 
>  >   This is mainly so we can get disk hot removal right.
> 
> What consequences does this have for (out-of-the-tree) file systems,
> beyond s/->f_dentry->d_inode->i_mapping/->f_mapping/g ?

None.  It only matters for block device inodes.  Out-of-tree fs is free to
do whatever it does with inodes of regular files/directories/etc.  

In quite a few cases you can cut down on dereferencing that way, but that's
covered by what you've mentioned.

If you take a look at the patchset you'll see
	* change of method prototypes in block devices (aka "you don't need
to start with bdev = inode->i_bdev, you get it from arguments" - check RD1--RD6
and you'll see)
	* a lot of places in mm/* that got aforementioned search-and-replace
treatment
	* very few changes in fs code, most of them of the same variety (same
search-and-replace)
	* couple of helper functions changed their prototypes:
		generic_write_checks() lost "inode" argument
		generic_osync_inode(), OTOH, got explicit address_space one.

That's it.