approach on getting nullfs to work again

Joerg Sonnenberger joerg at britannica.bec.de
Thu Feb 10 08:49:28 PST 2005


On Wed, Feb 09, 2005 at 09:38:15PM -0800, Matthew Dillon wrote:
>     In fact, *ALL* VOP or VFS ops that specify a file path have the same
>     ability to distinguish between namespaces as VOP ops that pass namecache
>     pointers, because path ops will also have access to the related namecache
>     pointer.

Matt, shouldn't we fully separate vnode operation affection only the vnode
(open, close, read, write) from those affection / depending on both
the name space and the vnode?

The first kind just has to be looked up and doesn't have to be handled
by nullfs at all. For unionfs the situation would be a bit more complicated,
but the underlying filesystem wouldn't have to be aware of unionfs for this
operations.

The second type is more difficult to handle, because the semantic of the
operation (e.g. mknod or remove or mountctl) can be different for nullfs.
This operations should be accessed by the mount point of the current
name cache entry, not the vnode. It would not change the situation
dramatically for normal filesystems, but allow stackable filesystems to
just change what they need to change.

Joerg





More information about the Kernel mailing list