On Wed, Jul 23, 2008 at 04:53:17PM -0500, Russell Cattelan wrote:
> Christoph Hellwig wrote:
> >Dave complained today that the fate of bhv_vnode_t isn't entirely clear
> >yet, so I've prepared these two patches to kill it in a minimally
> >invasive way. While it causes churn in a lot of areas it does not
> >affect the generated code at all.
> >
> >
> I know a bunch of stuff has gone in that is not very portable, which is
> fine since
> they can be dealt with individually since they are not that intrusive.
>
> Changing bhv_vnode_t to struct inode throughout the code is a pretty big
> change and would be a major pain to work around.
Have you actually looed at the patches? The only places where we use
struct inode outside of linux-2.6/ are:
- xfs_finish_reclaim:
Distangles the xfs_inode from Linux inode. Per defintion
OS-specific.
- xfs_sync_inodes:
Sync code that is quite OS specific. Dave will move it
to linux-2.6/ pretty soon.
- quota/xfs_qm_syscalls.c:
Similar sync code.
- xfs_acl.c:
ACL code with some OS dependencies, and pretty dead with my
pending patch to use the generic ACL code.
And no, it's not actually a big change.
|