Subject: Re: diffs for UVM/UBC improvements available
To: Frank van der Linden <firstname.lastname@example.org>
From: Chuck Silvers <email@example.com>
Date: 05/22/2001 08:42:20
On Tue, May 22, 2001 at 03:00:38PM +0200, Frank van der Linden wrote:
> On Tue, May 22, 2001 at 02:38:33AM -0700, Chuck Silvers wrote:
> > - move the vnode lock from struct vnode back into the fs-specific data.
> > the layered filesystems don't have a separate lock for their vnodes,
> > so the lock isn't really a vnode property.
> With that change you have taken a piece of shared code and re-duplicated
> it into all FS code (except the stacked code). I'm not sure I see the
> point in this. Also, the idea of having a lock element in the vnode
> structure was, as I recall (Bill Studenmund did it and will know) to
> be able to use the same lock for all layers of a layered filesystem.
> I.e. the upper layers just have pointers to the lower layer's lock.
> How did you do this now that this no longer exist? Have you tested
> the known-working layered filesystems (null and layer)?
I talked with Bill about this a few months back and moving the lock
back out of the vnode was his idea, actually. the way I've got it here,
nullfs and friends still use the same lock as the lower filesystem,
they just get to it via code rather than a pointer.
the idea was that since I'm changing struct vnode again now,
so I might as well make all the changes at once.
> I think I agree with removing VOP_ISLOCKED, arguably any code that
> needs it has problems. But better think about this some more.
ok, well, let me know if you can think of any problems with removing it.
this will certainly be necessary for MP eventually, so since I'm messing
with the locking VOPs above it seemed best just to remove this now.