Do we need DCACHE_WHITEOUT_TYPE?

Do we need DCACHE_WHITEOUT_TYPE?

[Goldwyn Rodrigues]

Hi,

There are some DCACHE_*_TYPE which are not used. I assume they were
added during the development of overlayfs/unionfs and do not seem to
make sense to keep.

Should we revert the following commits?

155e35d4daa8 ("VFS: Introduce inode-getting helpers for layered/unioned fs environments")
e7f7d2253c05 ("VFS: Add a whiteout dentry type")
df1a085af1f6 ("VFS: Add a fallthrough flag for marking virtual dentries")

Are there any more commits related or during this merge which are not
relevant anymore?

Regards,

-- 
Goldwyn

Re: Do we need DCACHE_WHITEOUT_TYPE?

[David Howells]

Goldwyn Rodrigues <rgoldwyn@suse.de> wrote:

> There are some DCACHE_*_TYPE which are not used. I assume they were
> added during the development of overlayfs/unionfs and do not seem to
> make sense to keep.

Yes, Al started passing the patches for unionmounts upstream, then Linus went
and plonked overlayfs in there which broke unionmounts in a number of ways, so
these can be cleaned up.

(Though I suggest that overlayfs should probably set the DCACHE_WHITEOUT_TYPE
on whiteouts).

> Should we revert the following commits?
> 
> 155e35d4daa8 ("VFS: Introduce inode-getting helpers for layered/unioned fs environments")
> e7f7d2253c05 ("VFS: Add a whiteout dentry type")
> df1a085af1f6 ("VFS: Add a fallthrough flag for marking virtual dentries")

The last one can definitely be reverted.  The second one possibly, though
overlayfs should possibly be using it.  The first one, I'm not sure - parts of
that overlayfs might be using.

> Are there any more commits related or during this merge which are not
> relevant anymore?

Possibly; I'll try and find some time to go through them.

David