Re: VFS KPI was Re: [OpenAFS-devel] Re: AFS ... or equivalent ...
Scott Long <[email protected]> Thu, 17 Jan 2008 10:58:08 -0700
| Newsgroups | gmane.os.freebsd.devel.file-systems,gmane.os.freebsd.devel.afs,gmane.comp.file-systems.openafs.devel,gmane.comp.file-systems.openafs.freebsd |
|---|---|
| Message-ID | <[email protected]> |
Rick Macklem wrote: > > > On Thu, 17 Jan 2008, Scott Long wrote: > >> >> Both Solaris and OSX seem to have found the path out of the VFS locking >> woods, and it would indeed be really nice if FreeBSD could follow suit. > Yes, I didn't mention Solaris (because I've never looked at it), but > being able to easily port file system code from OpenSolaris sounds like > it might be a big win. ZFS is an obvious example, but there is also > stuff like pNFS in the pipe that would be nice. (I'm going to try and > get my nfsv4 code into FreeBSD, but it will only be nfsv4.0 and I have > no urge to do 4.1/pNFS.) And others mentioned that there is quite a bit > of momentum w.r.t. OpenAFS in Solaris. > > rick > It's not just about portability, though. I know that some will argue with me on this, but the current VFS API is prohibitively difficult to use properly if you're trying to write, port, or debug a filesystem. You have to worry about interactions between vnode locks, vnode interlocks, and mount locks even before you start locking your own filesystem-private structures. Vectors like VOP_RECLAIM are a delicate and unforgiving minefield. Gaining better portability would of course be a big win, but an even bigger win would be gaining the ability of mere mortal programmers to understand and get involved in filesystems in FreeBSD. Scott _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-fs To unsubscribe, send any mail to "[email protected]"