Re: [prepatch] address_space-based writeback

[Christoph Hellwig]

On Mon, Apr 29, 2002 at 01:34:02PM +0100, Anton Altaparmakov wrote:
> Basically one could just change iget4 into two functions: iget calling 
> fs->read_inode (with read_inode2 removed) and iget4 without the 
> ->read_inode and ->read_inode2 and returning a locked inode instead.
> 
> That would make it fs specific.
> 
> If we wanted to make it generic we do need a special method in the 
> operations, but wait, we already have read_inode2! So perhaps it isn't as 
> much of a hack after all...
> 
> If you wanted to get rid of the hackish nature of the beast, one could just 
> remove ->read_inode and rename ->read_inode2 to ->read_inode. Then change 
> all fs to accept two dummy parameters in their ->read_inode declaration...
> 
> If that would be an acceptable approach I would be happy to do a patch to 
> convert all in kernel file systems in 2.5.x.

Please take a look at icreate & surrounding code in the 2.5 XFS tree.
The code needs some cleanup, but I think the API is exactly what we want.

	Christoph