Date: Tue, 11 Jul 2000 13:48:30 -0400 (EDT) From: Robert Watson <rwatson@freebsd.org> To: Dmitry Brodsky <dima@cs.ubc.ca> Cc: freebsd-fs@freebsd.org Subject: Re: Meta-Data & stackable FS Message-ID: <Pine.NEB.3.96L.1000711134206.35768G-100000@fledge.watson.org> In-Reply-To: <200007102330.QAA06084@cascade.cs.ubc.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 10 Jul 2000, Dmitry Brodsky wrote: > I am in the process of creating a stackable fs by using the nullfs as > a template. One thing I am having a hard time figuring out is a good > way to instantiate additional meta-data with each file. Is that > possible? What is the best way of going about doing that? If this is non-persistent meta-data, you can store it off of the vnode in your layer via v_data. However, if it's persistent, you'll need to store it elsewhere, and have a way to map the persistent data back onto the vnodes in future instantiations. For some stacked file systems I've looked at, persistent storage of layering data is done via a transformation on the underlying vnode. In others, namespace escapes are used to provide storage. Both of these techniques have substantial limitations for per-file attribute data (such as ACLs). In the TrustedBSD framework, we've introduced two additional VOP's to get and set extended attributes on files, and support exists in UFS/FFS to back extended attributes. Stacked file systems sitting on top of UFS can store meta-data for a vnode under the layer in EA's off of that vnode. This addresses maintaining a correspondence between the meta-data and nodes, as the namespace quirks of UNIX (hard links, et al) and page/block alignment memory mapping requirements make the other methods at best uncomfortable. One thing to keep in mind with stacked file systems and the UFS attribute mechanism is that locking is a substantial concern: as not all locks can be acquired in advance in the current framework, it's easy to get into deadlock situations. There are also a couple of bugs in the EA support in 5.0, including some refcount issues on the vnodes. Kirk has simple fixes for that, and may already have committed them. In essence, vrele's are missing from some of the syscall wrappers, resulting in vnode leaking. Robert N M Watson robert@fledge.watson.org http://www.watson.org/~robert/ PGP key fingerprint: AF B5 5F FF A6 4A 79 37 ED 5F 55 E9 58 04 6A B1 TIS Labs at Network Associates, Safeport Network Services 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?Pine.NEB.3.96L.1000711134206.35768G-100000>