Subject: Re: VFS LKMs & pool_init() panic
To: None <dolecek@ics.muni.cz>
From: Bill Studenmund <wrstuden@nas.nasa.gov>
List: tech-kern
Date: 03/13/2000 11:21:32
On Mon, 13 Mar 2000, PER4MANCE, J. Dolecek wrote:

> Bill Studenmund wrote:
> > No, it (running an old LKM) would bite them squarely all the time. :-)
> > You're talking about (or SHOULD be talking about :-) adding a VFS routine.
> > These routines are exported by explicit jump vectors in the vfsops
> > structure. By adding one, you move everything after it, which includes the
> > handle to the VOP descripter table. :-)
> 
> Oh, that sucks :( Why is the handle last in list ? vfs_opv_descs,

Historical behavior I guess. But please don't change it. I don't think
this is a place we need to diverge from the other BSD's (see below).

> vfs_refcount & vfs_list should probably be moved to the start of
> structure
> and vfs_name should be constified. Would ease adding next vfs routine if
> there
> would ever be need.

I think the case where we don't need the vfs routine for normal operation
is rare, so I'd vote for not making it easy to use lkms for an out-of-date
vfs interface. :-)

We've already radically changed the VFS interface between 1.4 and 1.5, so
we wouldn't win anything by being compatible in this way. All we might do
is let people load lkm's which would bomb on unload. That doesn't sound
good to me. :-)

Take care,

Bill