Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 07 Aug 2002 17:42:23 -0700
From:      Terry Lambert <tlambert2@mindspring.com>
To:        Kelly Yancey <kbyanc@posi.net>
Cc:        "David E. Cross" <crossd@cs.rpi.edu>, fs@FreeBSD.ORG, guptar@cs.rpi.edu
Subject:   Re: vnodes (UFS journaling)?
Message-ID:  <3D51BE6F.298F0AF@mindspring.com>
References:  <20020807170230.Y6264-100000@gateway.posi.net>

next in thread | previous in thread | raw e-mail | index | archive | help
Kelly Yancey wrote:
> On Wed, 7 Aug 2002, Terry Lambert wrote:
> > "David E. Cross" wrote:
> > > Ok, we've noticed an "unfortunate" side-effect of our work. We leak data
> > > (part of the journal shows up in other files).  I think it is because
> >
> > What is it that you are trying to do with the vnode?
> >
> > -- Terry
> >
> 
>   I believe he is trying to keep a journal, likely as a file on the
> filesystem that is being journalled.  I don't actually know, but that's
> what I would guess.

I understand that he's working on a journalling FS.

I'm wondering what reaccessing vnodes has to do with journalling?

A vnode is an in-core abstract representation of a file in
a file system, that has a non-abstract real file association
in the FS itself.  A vnode doesn't know about the journalling
taking place in the underlying FS -- and *shouldn't* know
about it.  The VFS interface layer is intended as a method of
abstracting *exacly* this sort of implementation detail.

I'm wondering why journalling would ever be visible via a vnode
interaction, as a system artifact, when all the journaling
operations are encapsulated in the FS implementation proper, at
the non-abstract layer (in FFS terms, journalling would be a
per inode operation, not per vnode operation).

The *only* place you would really make the underlying structure
visible at all, IMO, would be in the addition of VNOPs to add a
transactioning interface for user applications.  Even then,
though, the operations would *still* be abstract.

It *seems* like they are trying to implement the journals above
the vnode layer, which makes no sense at all... so the question
is *why* this is a *vnode* issue at all, ever, under *any*
possible interpretation of the problem that's being solved?  If
they are having data corruption problems, it's lower down.

The question, as he asked it, is a total non-sequitur, given
his introductory text.  It's like asking "Is it shorter to New
York, or by bus?".

-- Terry

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-fs" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3D51BE6F.298F0AF>